X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=fs%2Fnfs%2Fread.c;h=a9c26521a9e2d8a13965e364dae9ac8392cecdb1;hb=97bf2856c6014879bd04983a3e9dfcdac1e7fe85;hp=05eb43fadf8e4296d3e6d41ed4a161c40e9e8efb;hpb=76828883507a47dae78837ab5dec5a5b4513c667;p=linux-2.6.git

diff --git a/fs/nfs/read.c b/fs/nfs/read.c
index 05eb43fad..a9c26521a 100644
--- a/fs/nfs/read.c
+++ b/fs/nfs/read.c
@@ -15,7 +15,6 @@
  * within the RPC code when root squashing is suspected.
  */
 
-#include <linux/config.h>
 #include <linux/time.h>
 #include <linux/kernel.h>
 #include <linux/errno.h>
@@ -31,36 +30,58 @@
 
 #include <asm/system.h>
 
+#include "internal.h"
+#include "iostat.h"
+
 #define NFSDBG_FACILITY		NFSDBG_PAGECACHE
 
 static int nfs_pagein_one(struct list_head *, struct inode *);
-static void nfs_readpage_result_partial(struct nfs_read_data *, int);
-static void nfs_readpage_result_full(struct nfs_read_data *, int);
+static const struct rpc_call_ops nfs_read_partial_ops;
+static const struct rpc_call_ops nfs_read_full_ops;
 
-static kmem_cache_t *nfs_rdata_cachep;
-mempool_t *nfs_rdata_mempool;
+static struct kmem_cache *nfs_rdata_cachep;
+static mempool_t *nfs_rdata_mempool;
 
 #define MIN_POOL_READ	(32)
 
-void nfs_readdata_release(void *data)
+struct nfs_read_data *nfs_readdata_alloc(size_t len)
 {
-        nfs_readdata_free(data);
+	unsigned int pagecount = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	struct nfs_read_data *p = mempool_alloc(nfs_rdata_mempool, GFP_NOFS);
+
+	if (p) {
+		memset(p, 0, sizeof(*p));
+		INIT_LIST_HEAD(&p->pages);
+		p->npages = pagecount;
+		if (pagecount <= ARRAY_SIZE(p->page_array))
+			p->pagevec = p->page_array;
+		else {
+			p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_NOFS);
+			if (!p->pagevec) {
+				mempool_free(p, nfs_rdata_mempool);
+				p = NULL;
+			}
+		}
+	}
+	return p;
 }
 
-static
-unsigned int nfs_page_length(struct inode *inode, struct page *page)
+static void nfs_readdata_rcu_free(struct rcu_head *head)
 {
-	loff_t i_size = i_size_read(inode);
-	unsigned long idx;
+	struct nfs_read_data *p = container_of(head, struct nfs_read_data, task.u.tk_rcu);
+	if (p && (p->pagevec != &p->page_array[0]))
+		kfree(p->pagevec);
+	mempool_free(p, nfs_rdata_mempool);
+}
 
-	if (i_size <= 0)
-		return 0;
-	idx = (i_size - 1) >> PAGE_CACHE_SHIFT;
-	if (page->index > idx)
-		return 0;
-	if (page->index != idx)
-		return PAGE_CACHE_SIZE;
-	return 1 + ((i_size - 1) & (PAGE_CACHE_SIZE - 1));
+static void nfs_readdata_free(struct nfs_read_data *rdata)
+{
+	call_rcu_bh(&rdata->task.u.tk_rcu, nfs_readdata_rcu_free);
+}
+
+void nfs_readdata_release(void *data)
+{
+        nfs_readdata_free(data);
 }
 
 static
@@ -72,6 +93,35 @@ int nfs_return_empty_page(struct page *page)
 	return 0;
 }
 
+static void nfs_readpage_truncate_uninitialised_page(struct nfs_read_data *data)
+{
+	unsigned int remainder = data->args.count - data->res.count;
+	unsigned int base = data->args.pgbase + data->res.count;
+	unsigned int pglen;
+	struct page **pages;
+
+	if (data->res.eof == 0 || remainder == 0)
+		return;
+	/*
+	 * Note: "remainder" can never be negative, since we check for
+	 * 	this in the XDR code.
+	 */
+	pages = &data->args.pages[base >> PAGE_CACHE_SHIFT];
+	base &= ~PAGE_CACHE_MASK;
+	pglen = PAGE_CACHE_SIZE - base;
+	for (;;) {
+		if (remainder <= pglen) {
+			memclear_highpage_flush(*pages, base, remainder);
+			break;
+		}
+		memclear_highpage_flush(*pages, base, pglen);
+		pages++;
+		remainder -= pglen;
+		pglen = PAGE_CACHE_SIZE;
+		base = 0;
+	}
+}
+
 /*
  * Read a page synchronously.
  */
@@ -80,12 +130,12 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode,
 {
 	unsigned int	rsize = NFS_SERVER(inode)->rsize;
 	unsigned int	count = PAGE_CACHE_SIZE;
-	int		result;
+	int result = -ENOMEM;
 	struct nfs_read_data *rdata;
 
-	rdata = nfs_readdata_alloc(1);
+	rdata = nfs_readdata_alloc(count);
 	if (!rdata)
-		return -ENOMEM;
+		goto out_unlock;
 
 	memset(rdata, 0, sizeof(*rdata));
 	rdata->flags = (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
@@ -112,7 +162,7 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode,
 		rdata->args.offset = page_offset(page) + rdata->args.pgbase;
 
 		dprintk("NFS: nfs_proc_read(%s, (%s/%Ld), %Lu, %u)\n",
-			NFS_SERVER(inode)->hostname,
+			NFS_SERVER(inode)->nfs_client->cl_hostname,
 			inode->i_sb->s_id,
 			(long long)NFS_FILEID(inode),
 			(unsigned long long)rdata->args.pgbase,
@@ -133,6 +183,8 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode,
 		}
 		count -= result;
 		rdata->args.pgbase += result;
+		nfs_add_stats(inode, NFSIOS_SERVERREADBYTES, result);
+
 		/* Note: result == 0 should only happen if we're caching
 		 * a write that extends the file and punches a hole.
 		 */
@@ -143,16 +195,17 @@ static int nfs_readpage_sync(struct nfs_open_context *ctx, struct inode *inode,
 	NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME;
 	spin_unlock(&inode->i_lock);
 
-	if (count)
-		memclear_highpage_flush(page, rdata->args.pgbase, count);
-	SetPageUptodate(page);
-	if (PageError(page))
-		ClearPageError(page);
+	if (rdata->res.eof || rdata->res.count == rdata->args.count) {
+		SetPageUptodate(page);
+		if (rdata->res.eof && count != 0)
+			memclear_highpage_flush(page, rdata->args.pgbase, count);
+	}
 	result = 0;
 
 io_error:
-	unlock_page(page);
 	nfs_readdata_free(rdata);
+out_unlock:
+	unlock_page(page);
 	return result;
 }
 
@@ -163,7 +216,7 @@ static int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
 	struct nfs_page	*new;
 	unsigned int len;
 
-	len = nfs_page_length(inode, page);
+	len = nfs_page_length(page);
 	if (len == 0)
 		return nfs_return_empty_page(page);
 	new = nfs_create_request(ctx, inode, page, 0, len);
@@ -196,9 +249,11 @@ static void nfs_readpage_release(struct nfs_page *req)
  * Set up the NFS read request struct
  */
 static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
+		const struct rpc_call_ops *call_ops,
 		unsigned int count, unsigned int offset)
 {
 	struct inode		*inode;
+	int flags;
 
 	data->req	  = req;
 	data->inode	  = inode = req->wb_context->dentry->d_inode;
@@ -216,6 +271,9 @@ static void nfs_read_rpcsetup(struct nfs_page *req, struct nfs_read_data *data,
 	data->res.eof     = 0;
 	nfs_fattr_init(&data->fattr);
 
+	/* Set up the initial task struct. */
+	flags = RPC_TASK_ASYNC | (IS_SWAPFILE(inode)? NFS_RPC_SWAPFLAGS : 0);
+	rpc_init_task(&data->task, NFS_CLIENT(inode), flags, call_ops, data);
 	NFS_PROTO(inode)->read_setup(data);
 
 	data->task.tk_cookie = (unsigned long)inode;
@@ -250,9 +308,7 @@ static void nfs_execute_read(struct nfs_read_data *data)
 	sigset_t oldset;
 
 	rpc_clnt_sigmask(clnt, &oldset);
-	lock_kernel();
 	rpc_execute(&data->task);
-	unlock_kernel();
 	rpc_clnt_sigunmask(clnt, &oldset);
 }
 
@@ -274,25 +330,25 @@ static int nfs_pagein_multi(struct list_head *head, struct inode *inode)
 	struct nfs_page *req = nfs_list_entry(head->next);
 	struct page *page = req->wb_page;
 	struct nfs_read_data *data;
-	unsigned int rsize = NFS_SERVER(inode)->rsize;
-	unsigned int nbytes, offset;
+	size_t rsize = NFS_SERVER(inode)->rsize, nbytes;
+	unsigned int offset;
 	int requests = 0;
 	LIST_HEAD(list);
 
 	nfs_list_remove_request(req);
 
 	nbytes = req->wb_bytes;
-	for(;;) {
-		data = nfs_readdata_alloc(1);
+	do {
+		size_t len = min(nbytes,rsize);
+
+		data = nfs_readdata_alloc(len);
 		if (!data)
 			goto out_bad;
 		INIT_LIST_HEAD(&data->pages);
 		list_add(&data->pages, &list);
 		requests++;
-		if (nbytes <= rsize)
-			break;
-		nbytes -= rsize;
-	}
+		nbytes -= len;
+	} while(nbytes != 0);
 	atomic_set(&req->wb_complete, requests);
 
 	ClearPageError(page);
@@ -303,14 +359,15 @@ static int nfs_pagein_multi(struct list_head *head, struct inode *inode)
 		list_del_init(&data->pages);
 
 		data->pagevec[0] = page;
-		data->complete = nfs_readpage_result_partial;
 
 		if (nbytes > rsize) {
-			nfs_read_rpcsetup(req, data, rsize, offset);
+			nfs_read_rpcsetup(req, data, &nfs_read_partial_ops,
+					rsize, offset);
 			offset += rsize;
 			nbytes -= rsize;
 		} else {
-			nfs_read_rpcsetup(req, data, nbytes, offset);
+			nfs_read_rpcsetup(req, data, &nfs_read_partial_ops,
+					nbytes, offset);
 			nbytes = 0;
 		}
 		nfs_execute_read(data);
@@ -339,7 +396,7 @@ static int nfs_pagein_one(struct list_head *head, struct inode *inode)
 	if (NFS_SERVER(inode)->rsize < PAGE_CACHE_SIZE)
 		return nfs_pagein_multi(head, inode);
 
-	data = nfs_readdata_alloc(NFS_SERVER(inode)->rpages);
+	data = nfs_readdata_alloc(NFS_SERVER(inode)->rsize);
 	if (!data)
 		goto out_bad;
 
@@ -356,8 +413,7 @@ static int nfs_pagein_one(struct list_head *head, struct inode *inode)
 	}
 	req = nfs_list_entry(data->pages.next);
 
-	data->complete = nfs_readpage_result_full;
-	nfs_read_rpcsetup(req, data, count, 0);
+	nfs_read_rpcsetup(req, data, &nfs_read_full_ops, count, 0);
 
 	nfs_execute_read(data);
 	return 0;
@@ -388,26 +444,74 @@ nfs_pagein_list(struct list_head *head, int rpages)
 	return error;
 }
 
+/*
+ * This is the callback from RPC telling us whether a reply was
+ * received or some error occurred (timeout or socket shutdown).
+ */
+int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data)
+{
+	int status;
+
+	dprintk("%s: %4d, (status %d)\n", __FUNCTION__, task->tk_pid,
+			task->tk_status);
+
+	status = NFS_PROTO(data->inode)->read_done(task, data);
+	if (status != 0)
+		return status;
+
+	nfs_add_stats(data->inode, NFSIOS_SERVERREADBYTES, data->res.count);
+
+	if (task->tk_status == -ESTALE) {
+		set_bit(NFS_INO_STALE, &NFS_FLAGS(data->inode));
+		nfs_mark_for_revalidate(data->inode);
+	}
+	spin_lock(&data->inode->i_lock);
+	NFS_I(data->inode)->cache_validity |= NFS_INO_INVALID_ATIME;
+	spin_unlock(&data->inode->i_lock);
+	return 0;
+}
+
+static int nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data)
+{
+	struct nfs_readargs *argp = &data->args;
+	struct nfs_readres *resp = &data->res;
+
+	if (resp->eof || resp->count == argp->count)
+		return 0;
+
+	/* This is a short read! */
+	nfs_inc_stats(data->inode, NFSIOS_SHORTREAD);
+	/* Has the server at least made some progress? */
+	if (resp->count == 0)
+		return 0;
+
+	/* Yes, so retry the read at the end of the data */
+	argp->offset += resp->count;
+	argp->pgbase += resp->count;
+	argp->count -= resp->count;
+	rpc_restart_call(task);
+	return -EAGAIN;
+}
+
 /*
  * Handle a read reply that fills part of a page.
  */
-static void nfs_readpage_result_partial(struct nfs_read_data *data, int status)
+static void nfs_readpage_result_partial(struct rpc_task *task, void *calldata)
 {
+	struct nfs_read_data *data = calldata;
 	struct nfs_page *req = data->req;
 	struct page *page = req->wb_page;
  
-	if (status >= 0) {
-		unsigned int request = data->args.count;
-		unsigned int result = data->res.count;
-
-		if (result < request) {
-			memclear_highpage_flush(page,
-						data->args.pgbase + result,
-						request - result);
-		}
-	} else
-		SetPageError(page);
+	if (nfs_readpage_result(task, data) != 0)
+		return;
 
+	if (likely(task->tk_status >= 0)) {
+		nfs_readpage_truncate_uninitialised_page(data);
+		if (nfs_readpage_retry(task, data) != 0)
+			return;
+	}
+	if (unlikely(task->tk_status < 0))
+		SetPageError(page);
 	if (atomic_dec_and_test(&req->wb_complete)) {
 		if (!PageError(page))
 			SetPageUptodate(page);
@@ -415,68 +519,68 @@ static void nfs_readpage_result_partial(struct nfs_read_data *data, int status)
 	}
 }
 
-/*
- * This is the callback from RPC telling us whether a reply was
- * received or some error occurred (timeout or socket shutdown).
- */
-static void nfs_readpage_result_full(struct nfs_read_data *data, int status)
+static const struct rpc_call_ops nfs_read_partial_ops = {
+	.rpc_call_done = nfs_readpage_result_partial,
+	.rpc_release = nfs_readdata_release,
+};
+
+static void nfs_readpage_set_pages_uptodate(struct nfs_read_data *data)
 {
 	unsigned int count = data->res.count;
-
-	while (!list_empty(&data->pages)) {
-		struct nfs_page *req = nfs_list_entry(data->pages.next);
-		struct page *page = req->wb_page;
-		nfs_list_remove_request(req);
-
-		if (status >= 0) {
-			if (count < PAGE_CACHE_SIZE) {
-				if (count < req->wb_bytes)
-					memclear_highpage_flush(page,
-							req->wb_pgbase + count,
-							req->wb_bytes - count);
-				count = 0;
-			} else
-				count -= PAGE_CACHE_SIZE;
-			SetPageUptodate(page);
-		} else
-			SetPageError(page);
-		nfs_readpage_release(req);
-	}
+	unsigned int base = data->args.pgbase;
+	struct page **pages;
+
+	if (data->res.eof)
+		count = data->args.count;
+	if (unlikely(count == 0))
+		return;
+	pages = &data->args.pages[base >> PAGE_CACHE_SHIFT];
+	base &= ~PAGE_CACHE_MASK;
+	count += base;
+	for (;count >= PAGE_CACHE_SIZE; count -= PAGE_CACHE_SIZE, pages++)
+		SetPageUptodate(*pages);
+	if (count == 0)
+		return;
+	/* Was this a short read? */
+	if (data->res.eof || data->res.count == data->args.count)
+		SetPageUptodate(*pages);
 }
 
 /*
  * This is the callback from RPC telling us whether a reply was
  * received or some error occurred (timeout or socket shutdown).
  */
-void nfs_readpage_result(struct rpc_task *task, void *calldata)
+static void nfs_readpage_result_full(struct rpc_task *task, void *calldata)
 {
 	struct nfs_read_data *data = calldata;
-	struct nfs_readargs *argp = &data->args;
-	struct nfs_readres *resp = &data->res;
-	int status = task->tk_status;
-
-	dprintk("NFS: %4d nfs_readpage_result, (status %d)\n",
-		task->tk_pid, status);
-
-	/* Is this a short read? */
-	if (task->tk_status >= 0 && resp->count < argp->count && !resp->eof) {
-		/* Has the server at least made some progress? */
-		if (resp->count != 0) {
-			/* Yes, so retry the read at the end of the data */
-			argp->offset += resp->count;
-			argp->pgbase += resp->count;
-			argp->count -= resp->count;
-			rpc_restart_call(task);
+
+	if (nfs_readpage_result(task, data) != 0)
+		return;
+	/*
+	 * Note: nfs_readpage_retry may change the values of
+	 * data->args. In the multi-page case, we therefore need
+	 * to ensure that we call nfs_readpage_set_pages_uptodate()
+	 * first.
+	 */
+	if (likely(task->tk_status >= 0)) {
+		nfs_readpage_truncate_uninitialised_page(data);
+		nfs_readpage_set_pages_uptodate(data);
+		if (nfs_readpage_retry(task, data) != 0)
 			return;
-		}
-		task->tk_status = -EIO;
 	}
-	spin_lock(&data->inode->i_lock);
-	NFS_I(data->inode)->cache_validity |= NFS_INO_INVALID_ATIME;
-	spin_unlock(&data->inode->i_lock);
-	data->complete(data, status);
+	while (!list_empty(&data->pages)) {
+		struct nfs_page *req = nfs_list_entry(data->pages.next);
+
+		nfs_list_remove_request(req);
+		nfs_readpage_release(req);
+	}
 }
 
+static const struct rpc_call_ops nfs_read_full_ops = {
+	.rpc_call_done = nfs_readpage_result_full,
+	.rpc_release = nfs_readdata_release,
+};
+
 /*
  * Read a page over NFS.
  * We read the page synchronously in the following case:
@@ -491,6 +595,9 @@ int nfs_readpage(struct file *file, struct page *page)
 
 	dprintk("NFS: nfs_readpage (%p %ld@%lu)\n",
 		page, PAGE_CACHE_SIZE, page->index);
+	nfs_inc_stats(inode, NFSIOS_VFSREADPAGE);
+	nfs_add_stats(inode, NFSIOS_READPAGES, 1);
+
 	/*
 	 * Try to flush any pending writes to the file..
 	 *
@@ -502,10 +609,15 @@ int nfs_readpage(struct file *file, struct page *page)
 	if (error)
 		goto out_error;
 
+	error = -ESTALE;
+	if (NFS_STALE(inode))
+		goto out_error;
+
 	if (file == NULL) {
+		error = -EBADF;
 		ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
 		if (ctx == NULL)
-			return -EBADF;
+			goto out_error;
 	} else
 		ctx = get_nfs_open_context((struct nfs_open_context *)
 				file->private_data);
@@ -540,7 +652,7 @@ readpage_async_filler(void *data, struct page *page)
 	unsigned int len;
 
 	nfs_wb_page(inode, page);
-	len = nfs_page_length(inode, page);
+	len = nfs_page_length(page);
 	if (len == 0)
 		return nfs_return_empty_page(page);
 	new = nfs_create_request(desc->ctx, inode, page, 0, len);
@@ -564,12 +676,16 @@ int nfs_readpages(struct file *filp, struct address_space *mapping,
 	};
 	struct inode *inode = mapping->host;
 	struct nfs_server *server = NFS_SERVER(inode);
-	int ret;
+	int ret = -ESTALE;
 
 	dprintk("NFS: nfs_readpages (%s/%Ld %d)\n",
 			inode->i_sb->s_id,
 			(long long)NFS_FILEID(inode),
 			nr_pages);
+	nfs_inc_stats(inode, NFSIOS_VFSREADPAGES);
+
+	if (NFS_STALE(inode))
+		goto out;
 
 	if (filp == NULL) {
 		desc.ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
@@ -582,13 +698,15 @@ int nfs_readpages(struct file *filp, struct address_space *mapping,
 	if (!list_empty(&head)) {
 		int err = nfs_pagein_list(&head, server->rpages);
 		if (!ret)
+			nfs_add_stats(inode, NFSIOS_READPAGES, err);
 			ret = err;
 	}
 	put_nfs_open_context(desc.ctx);
+out:
 	return ret;
 }
 
-int nfs_init_readpagecache(void)
+int __init nfs_init_readpagecache(void)
 {
 	nfs_rdata_cachep = kmem_cache_create("nfs_read_data",
 					     sizeof(struct nfs_read_data),
@@ -597,10 +715,8 @@ int nfs_init_readpagecache(void)
 	if (nfs_rdata_cachep == NULL)
 		return -ENOMEM;
 
-	nfs_rdata_mempool = mempool_create(MIN_POOL_READ,
-					   mempool_alloc_slab,
-					   mempool_free_slab,
-					   nfs_rdata_cachep);
+	nfs_rdata_mempool = mempool_create_slab_pool(MIN_POOL_READ,
+						     nfs_rdata_cachep);
 	if (nfs_rdata_mempool == NULL)
 		return -ENOMEM;
 
@@ -610,6 +726,5 @@ int nfs_init_readpagecache(void)
 void nfs_destroy_readpagecache(void)
 {
 	mempool_destroy(nfs_rdata_mempool);
-	if (kmem_cache_destroy(nfs_rdata_cachep))
-		printk(KERN_INFO "nfs_read_data: not all structures were freed\n");
+	kmem_cache_destroy(nfs_rdata_cachep);
 }