NFS: Avoid races between writebacks and truncation

Currently, there is no serialisation between NFS asynchronous writebacks
and truncation at the page level due to the fact that nfs_sync_inode()
cannot lock the pages that it is about to write out.

This means that it is possible to be flushing out data (and calling something
like set_page_writeback()) while the page cache is busy evicting the page.
Oops...

Use the hooks provided in try_to_release_page() to ensure that dirty pages
are always written back to storage before we evict them.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

diff --git a/fs/nfs/file.c b/fs/nfs/file.c
index 7a79fbe9f5394c67dd53b01526d37e29fae0ce4e..387809f2d188db64fc7fc699a651d79af4be9cfa 100644 (file)
--- a/fs/nfs/file.c
+++ b/fs/nfs/file.c
@@ -316,6 +316,17 @@ static int nfs_commit_write(struct file *file, struct page *page, unsigned offse
        return status;
 }
 
+static int nfs_invalidate_page(struct page *page, unsigned long offset)
+{
+       /* FIXME: we really should cancel any unstarted writes on this page */
+       return 1;
+}
+
+static int nfs_release_page(struct page *page, gfp_t gfp)
+{
+       return !nfs_wb_page(page->mapping->host, page);
+}
+
 struct address_space_operations nfs_file_aops = {
        .readpage = nfs_readpage,
        .readpages = nfs_readpages,
@@ -324,6 +335,8 @@ struct address_space_operations nfs_file_aops = {
        .writepages = nfs_writepages,
        .prepare_write = nfs_prepare_write,
        .commit_write = nfs_commit_write,
+       .invalidatepage = nfs_invalidate_page,
+       .releasepage = nfs_release_page,
 #ifdef CONFIG_NFS_DIRECTIO
        .direct_IO = nfs_direct_IO,
 #endif

diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index d53857b148e249933ac2ff9813e07bdb445a63dc..d6e076c9dbe13b91ab47f598b97625614140750f 100644 (file)
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -85,6 +85,10 @@ nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
        atomic_set(&req->wb_complete, 0);
        req->wb_index   = page->index;
        page_cache_get(page);
+       BUG_ON(PagePrivate(page));
+       BUG_ON(!PageLocked(page));
+       BUG_ON(page->mapping->host != inode);
+       SetPagePrivate(page);
        req->wb_offset  = offset;
        req->wb_pgbase  = offset;
        req->wb_bytes   = count;
@@ -147,8 +151,10 @@ void nfs_clear_page_writeback(struct nfs_page *req)
  */
 void nfs_clear_request(struct nfs_page *req)
 {
-       if (req->wb_page) {
-               page_cache_release(req->wb_page);
+       struct page *page = req->wb_page;
+       if (page != NULL) {
+               ClearPagePrivate(page);
+               page_cache_release(page);
                req->wb_page = NULL;
        }
 }