From mboxrd@z Thu Jan 1 00:00:00 1970 From: Steven Whitehouse Date: Fri, 20 Apr 2007 09:18:30 +0100 Subject: [Cluster-devel] [GFS2] Patch to fix mmap of stuffed files Message-ID: <1177057110.1636.370.camel@quoit.chygwyn.com> List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hi, If a stuffed file is mmaped and a page fault is generated at some offset above the initial page, we need to create a zero page to hang the buffer heads off before we can unstuff the file. This is a fix for bz #236087 Steve. diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c index 90c2879..30c1562 100644 --- a/fs/gfs2/ops_address.c +++ b/fs/gfs2/ops_address.c @@ -197,7 +197,19 @@ static int stuffed_readpage(struct gfs2_inode *ip, struct page *page) void *kaddr; int error; - BUG_ON(page->index); + /* + * Due to the order of unstuffing files and ->nopage(), we can be + * asked for a zero page in the case of a stuffed file being extended, + * so we need to supply one here. It doesn't happen often. + */ + if (unlikely(page->index)) { + kaddr = kmap_atomic(page, KM_USER0); + memset(kaddr, 0, PAGE_CACHE_SIZE); + kunmap_atomic(kaddr, KM_USER0); + flush_dcache_page(page); + SetPageUptodate(page); + return 0; + } error = gfs2_meta_inode_buffer(ip, &dibh); if (error) @@ -208,9 +220,8 @@ static int stuffed_readpage(struct gfs2_inode *ip, struct page *page) ip->i_di.di_size); memset(kaddr + ip->i_di.di_size, 0, PAGE_CACHE_SIZE - ip->i_di.di_size); kunmap_atomic(kaddr, KM_USER0); - + flush_dcache_page(page); brelse(dibh); - SetPageUptodate(page); return 0;