[PATCH 1/3] NFS: Fix documenting comment for nfs_revalidate_file_size()

[PATCH 1/3] NFS: Fix documenting comment for nfs_revalidate_file_size()

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 63940a7a70be..d02a63af9c15 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -89,7 +89,7 @@ nfs_file_release(struct inode *inode, struct file *filp)
 EXPORT_SYMBOL_GPL(nfs_file_release);
 
 /**
- * nfs_revalidate_size - Revalidate the file size
+ * nfs_revalidate_file_size - Revalidate the file size
  * @inode: pointer to inode struct
  * @filp: pointer to struct file
  *
-- 
2.29.2

[PATCH 2/3] NFS: Optimise sparse writes past the end of file

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

If we're doing a write, and the entire page lies beyond the end-of-file,
then we can assume the write can be extended to cover the beginning of
the page, since we know the data in that region will be all zeros.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/file.c  |  4 +---
 fs/nfs/write.c | 20 ++++++++++++--------
 2 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index d02a63af9c15..02795a01c7ef 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -626,13 +626,11 @@ ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
 	/*
 	 * O_APPEND implies that we must revalidate the file length.
 	 */
-	if (iocb->ki_flags & IOCB_APPEND) {
+	if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) {
 		result = nfs_revalidate_file_size(inode, file);
 		if (result)
 			goto out;
 	}
-	if (iocb->ki_pos > i_size_read(inode))
-		nfs_revalidate_mapping(inode, file->f_mapping);
 
 	since = filemap_sample_wb_err(file->f_mapping);
 	nfs_start_io_write(inode);
diff --git a/fs/nfs/write.c b/fs/nfs/write.c
index 639c34fec04a..6193350356a8 100644
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1278,19 +1278,21 @@ bool nfs_ctx_key_to_expire(struct nfs_open_context *ctx, struct inode *inode)
  * the PageUptodate() flag. In this case, we will need to turn off
  * write optimisations that depend on the page contents being correct.
  */
-static bool nfs_write_pageuptodate(struct page *page, struct inode *inode)
+static bool nfs_write_pageuptodate(struct page *page, struct inode *inode,
+				   unsigned int pagelen)
 {
 	struct nfs_inode *nfsi = NFS_I(inode);
 
 	if (nfs_have_delegated_attributes(inode))
 		goto out;
-	if (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
+	if (nfsi->cache_validity &
+	    (NFS_INO_REVAL_PAGECACHE | NFS_INO_INVALID_SIZE))
 		return false;
 	smp_rmb();
-	if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags))
+	if (test_bit(NFS_INO_INVALIDATING, &nfsi->flags) && pagelen != 0)
 		return false;
 out:
-	if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
+	if (nfsi->cache_validity & NFS_INO_INVALID_DATA && pagelen != 0)
 		return false;
 	return PageUptodate(page) != 0;
 }
@@ -1310,7 +1312,8 @@ is_whole_file_wrlock(struct file_lock *fl)
  * If the file is opened for synchronous writes then we can just skip the rest
  * of the checks.
  */
-static int nfs_can_extend_write(struct file *file, struct page *page, struct inode *inode)
+static int nfs_can_extend_write(struct file *file, struct page *page,
+				struct inode *inode, unsigned int pagelen)
 {
 	int ret;
 	struct file_lock_context *flctx = inode->i_flctx;
@@ -1318,7 +1321,7 @@ static int nfs_can_extend_write(struct file *file, struct page *page, struct ino
 
 	if (file->f_flags & O_DSYNC)
 		return 0;
-	if (!nfs_write_pageuptodate(page, inode))
+	if (!nfs_write_pageuptodate(page, inode, pagelen))
 		return 0;
 	if (NFS_PROTO(inode)->have_delegation(inode, FMODE_WRITE))
 		return 1;
@@ -1356,6 +1359,7 @@ int nfs_updatepage(struct file *file, struct page *page,
 	struct nfs_open_context *ctx = nfs_file_open_context(file);
 	struct address_space *mapping = page_file_mapping(page);
 	struct inode	*inode = mapping->host;
+	unsigned int	pagelen = nfs_page_length(page);
 	int		status = 0;
 
 	nfs_inc_stats(inode, NFSIOS_VFSUPDATEPAGE);
@@ -1366,8 +1370,8 @@ int nfs_updatepage(struct file *file, struct page *page,
 	if (!count)
 		goto out;
 
-	if (nfs_can_extend_write(file, page, inode)) {
-		count = max(count + offset, nfs_page_length(page));
+	if (nfs_can_extend_write(file, page, inode, pagelen)) {
+		count = max(count + offset, pagelen);
 		offset = 0;
 	}
 
-- 
2.29.2

[PATCH 3/3] NFS: Always clear an invalid mapping when attempting a buffered write

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

If the page cache is invalid, then we can't do read-modify-write, so
ensure that we do clear it when we know it is invalid.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/file.c          |  2 +
 fs/nfs/inode.c         | 97 +++++++++++++++++++++++-------------------
 include/linux/nfs_fs.h |  1 +
 3 files changed, 57 insertions(+), 43 deletions(-)

diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 02795a01c7ef..03fd1dcc96bd 100644
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -632,6 +632,8 @@ ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
 			goto out;
 	}
 
+	nfs_clear_invalid_mapping(file->f_mapping);
+
 	since = filemap_sample_wb_err(file->f_mapping);
 	nfs_start_io_write(inode);
 	result = generic_write_checks(iocb, from);
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 522aa10a1a3e..2533053d764a 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -1257,55 +1257,19 @@ static int nfs_invalidate_mapping(struct inode *inode, struct address_space *map
 	return 0;
 }
 
-bool nfs_mapping_need_revalidate_inode(struct inode *inode)
-{
-	return nfs_check_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE) ||
-		NFS_STALE(inode);
-}
-
-int nfs_revalidate_mapping_rcu(struct inode *inode)
-{
-	struct nfs_inode *nfsi = NFS_I(inode);
-	unsigned long *bitlock = &nfsi->flags;
-	int ret = 0;
-
-	if (IS_SWAPFILE(inode))
-		goto out;
-	if (nfs_mapping_need_revalidate_inode(inode)) {
-		ret = -ECHILD;
-		goto out;
-	}
-	spin_lock(&inode->i_lock);
-	if (test_bit(NFS_INO_INVALIDATING, bitlock) ||
-	    (nfsi->cache_validity & NFS_INO_INVALID_DATA))
-		ret = -ECHILD;
-	spin_unlock(&inode->i_lock);
-out:
-	return ret;
-}
-
 /**
- * nfs_revalidate_mapping - Revalidate the pagecache
- * @inode: pointer to host inode
+ * nfs_clear_invalid_mapping - Conditionally clear a mapping
  * @mapping: pointer to mapping
+ *
+ * If the NFS_INO_INVALID_DATA inode flag is set, clear the mapping.
  */
-int nfs_revalidate_mapping(struct inode *inode,
-		struct address_space *mapping)
+int nfs_clear_invalid_mapping(struct address_space *mapping)
 {
+	struct inode *inode = mapping->host;
 	struct nfs_inode *nfsi = NFS_I(inode);
 	unsigned long *bitlock = &nfsi->flags;
 	int ret = 0;
 
-	/* swapfiles are not supposed to be shared. */
-	if (IS_SWAPFILE(inode))
-		goto out;
-
-	if (nfs_mapping_need_revalidate_inode(inode)) {
-		ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
-		if (ret < 0)
-			goto out;
-	}
-
 	/*
 	 * We must clear NFS_INO_INVALID_DATA first to ensure that
 	 * invalidations that come in while we're shooting down the mappings
@@ -1336,8 +1300,8 @@ int nfs_revalidate_mapping(struct inode *inode,
 
 	set_bit(NFS_INO_INVALIDATING, bitlock);
 	smp_wmb();
-	nfsi->cache_validity &= ~(NFS_INO_INVALID_DATA|
-			NFS_INO_DATA_INVAL_DEFER);
+	nfsi->cache_validity &=
+		~(NFS_INO_INVALID_DATA | NFS_INO_DATA_INVAL_DEFER);
 	spin_unlock(&inode->i_lock);
 	trace_nfs_invalidate_mapping_enter(inode);
 	ret = nfs_invalidate_mapping(inode, mapping);
@@ -1350,6 +1314,53 @@ int nfs_revalidate_mapping(struct inode *inode,
 	return ret;
 }
 
+bool nfs_mapping_need_revalidate_inode(struct inode *inode)
+{
+	return nfs_check_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE) ||
+		NFS_STALE(inode);
+}
+
+int nfs_revalidate_mapping_rcu(struct inode *inode)
+{
+	struct nfs_inode *nfsi = NFS_I(inode);
+	unsigned long *bitlock = &nfsi->flags;
+	int ret = 0;
+
+	if (IS_SWAPFILE(inode))
+		goto out;
+	if (nfs_mapping_need_revalidate_inode(inode)) {
+		ret = -ECHILD;
+		goto out;
+	}
+	spin_lock(&inode->i_lock);
+	if (test_bit(NFS_INO_INVALIDATING, bitlock) ||
+	    (nfsi->cache_validity & NFS_INO_INVALID_DATA))
+		ret = -ECHILD;
+	spin_unlock(&inode->i_lock);
+out:
+	return ret;
+}
+
+/**
+ * nfs_revalidate_mapping - Revalidate the pagecache
+ * @inode: pointer to host inode
+ * @mapping: pointer to mapping
+ */
+int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
+{
+	/* swapfiles are not supposed to be shared. */
+	if (IS_SWAPFILE(inode))
+		return 0;
+
+	if (nfs_mapping_need_revalidate_inode(inode)) {
+		int ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
+		if (ret < 0)
+			return ret;
+	}
+
+	return nfs_clear_invalid_mapping(mapping);
+}
+
 static bool nfs_file_has_writers(struct nfs_inode *nfsi)
 {
 	struct inode *inode = &nfsi->vfs_inode;
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index 681ed98e4ba8..0b0d67017177 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -387,6 +387,7 @@ extern int nfs_open(struct inode *, struct file *);
 extern int nfs_attribute_cache_expired(struct inode *inode);
 extern int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode);
 extern int __nfs_revalidate_inode(struct nfs_server *, struct inode *);
+extern int nfs_clear_invalid_mapping(struct address_space *mapping);
 extern bool nfs_mapping_need_revalidate_inode(struct inode *inode);
 extern int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping);
 extern int nfs_revalidate_mapping_rcu(struct inode *inode);
-- 
2.29.2