Merge home.transmeta.com:/home/torvalds/v2.5/morton

into home.transmeta.com:/home/torvalds/v2.5/linux

fs/block_dev.c

@@ -195,11 +195,15 @@ static loff_t block_llseek(struct file *file, loff_t offset, int origin)
 
 static int __block_fsync(struct inode * inode)
 {
-	int ret;
-
-	filemap_fdatasync(inode->i_mapping);
-	ret = sync_buffers(inode->i_rdev, 1);
-	filemap_fdatawait(inode->i_mapping);
+	int ret, err;
+
+	ret = filemap_fdatasync(inode->i_mapping);
+	err = sync_buffers(inode->i_rdev, 1);
+	if (err && !ret)
+		ret = err;
+	err = filemap_fdatawait(inode->i_mapping);
+	if (err && !ret)
+		ret = err;
 
 	return ret;
 }

fs/buffer.c

@@ -410,9 +410,9 @@ asmlinkage long sys_fsync(unsigned int fd)
 	struct file * file;
 	struct dentry * dentry;
 	struct inode * inode;
-	int err;
+	int ret, err;
 
-	err = -EBADF;
+	ret = -EBADF;
 	file = fget(fd);
 	if (!file)
 		goto out;
@@ -420,21 +420,27 @@ asmlinkage long sys_fsync(unsigned int fd)
 	dentry = file->f_dentry;
 	inode = dentry->d_inode;
 
-	err = -EINVAL;
-	if (!file->f_op || !file->f_op->fsync)
+	ret = -EINVAL;
+	if (!file->f_op || !file->f_op->fsync) {
+		/* Why?  We can still call filemap_fdatasync */
 		goto out_putf;
+	}
 
 	/* We need to protect against concurrent writers.. */
 	down(&inode->i_sem);
-	filemap_fdatasync(inode->i_mapping);
+	ret = filemap_fdatasync(inode->i_mapping);
 	err = file->f_op->fsync(file, dentry, 0);
-	filemap_fdatawait(inode->i_mapping);
+	if (err && !ret)
+		ret = err;
+	err = filemap_fdatawait(inode->i_mapping);
+	if (err && !ret)
+		ret = err;
 	up(&inode->i_sem);
 
 out_putf:
 	fput(file);
 out:
-	return err;
+	return ret;
 }
 
 asmlinkage long sys_fdatasync(unsigned int fd)
@@ -442,9 +448,9 @@ asmlinkage long sys_fdatasync(unsigned int fd)
 	struct file * file;
 	struct dentry * dentry;
 	struct inode * inode;
-	int err;
+	int ret, err;
 
-	err = -EBADF;
+	ret = -EBADF;
 	file = fget(fd);
 	if (!file)
 		goto out;
@@ -452,20 +458,24 @@ asmlinkage long sys_fdatasync(unsigned int fd)
 	dentry = file->f_dentry;
 	inode = dentry->d_inode;
 
-	err = -EINVAL;
+	ret = -EINVAL;
 	if (!file->f_op || !file->f_op->fsync)
 		goto out_putf;
 
 	down(&inode->i_sem);
-	filemap_fdatasync(inode->i_mapping);
+	ret = filemap_fdatasync(inode->i_mapping);
 	err = file->f_op->fsync(file, dentry, 1);
-	filemap_fdatawait(inode->i_mapping);
+	if (err && !ret)
+		ret = err;
+	err = filemap_fdatawait(inode->i_mapping);
+	if (err && !ret)
+		ret = err;
 	up(&inode->i_sem);
 
 out_putf:
 	fput(file);
 out:
-	return err;
+	return ret;
 }
 
 /* After several hours of tedious analysis, the following hash
@@ -1431,6 +1441,7 @@ static int __block_write_full_page(struct inode *inode, struct page *page, get_b
 	int err, i;
 	unsigned long block;
 	struct buffer_head *bh, *head;
+	int need_unlock;
 
 	if (!PageLocked(page))
 		BUG();
@@ -1486,8 +1497,34 @@ static int __block_write_full_page(struct inode *inode, struct page *page, get_b
 	return 0;
 
 out:
+	/*
+	 * ENOSPC, or some other error.  We may already have added some
+	 * blocks to the file, so we need to write these out to avoid
+	 * exposing stale data.
+	 */
 	ClearPageUptodate(page);
-	UnlockPage(page);
+	bh = head;
+	need_unlock = 1;
+	/* Recovery: lock and submit the mapped buffers */
+	do {
+		if (buffer_mapped(bh)) {
+			lock_buffer(bh);
+			set_buffer_async_io(bh);
+			need_unlock = 0;
+		}
+		bh = bh->b_this_page;
+	} while (bh != head);
+	do {
+		struct buffer_head *next = bh->b_this_page;
+		if (buffer_mapped(bh)) {
+			set_bit(BH_Uptodate, &bh->b_state);
+			clear_bit(BH_Dirty, &bh->b_state);
+			submit_bh(WRITE, bh);
+		}
+		bh = next;
+	} while (bh != head);
+	if (need_unlock)
+		UnlockPage(page);
 	return err;
 }
 
@@ -1518,6 +1555,7 @@ static int __block_prepare_write(struct inode *inode, struct page *page,
 			continue;
 		if (block_start >= to)
 			break;
+		clear_bit(BH_New, &bh->b_state);
 		if (!buffer_mapped(bh)) {
 			err = get_block(inode, block, bh, 1);
 			if (err)
@@ -1552,12 +1590,35 @@ static int __block_prepare_write(struct inode *inode, struct page *page,
 	 */
 	while(wait_bh > wait) {
 		wait_on_buffer(*--wait_bh);
-		err = -EIO;
 		if (!buffer_uptodate(*wait_bh))
-			goto out;
+			return -EIO;
 	}
 	return 0;
 out:
+	/*
+	 * Zero out any newly allocated blocks to avoid exposing stale
+	 * data.  If BH_New is set, we know that the block was newly
+	 * allocated in the above loop.
+	 */
+	bh = head;
+	block_start = 0;
+	do {
+		block_end = block_start+blocksize;
+		if (block_end <= from)
+			goto next_bh;
+		if (block_start >= to)
+			break;
+		if (buffer_new(bh)) {
+			if (buffer_uptodate(bh))
+				printk(KERN_ERR "%s: zeroing uptodate buffer!\n", __FUNCTION__);
+			memset(kaddr+block_start, 0, bh->b_size);
+			set_bit(BH_Uptodate, &bh->b_state);
+			mark_buffer_dirty(bh);
+		}
+next_bh:
+		block_start = block_end;
+		bh = bh->b_this_page;
+	} while (bh != head);
 	return err;
 }
 
@@ -1954,6 +2015,48 @@ int block_write_full_page(struct page *page, get_block_t *get_block)
 	goto done;
 }
 
+/*
+ * Commence writeout of all the buffers against a page.  The
+ * page must be locked.   Returns zero on success or a negative
+ * errno.
+ */
+int writeout_one_page(struct page *page)
+{
+	struct buffer_head *bh, *head = page->buffers;
+
+	if (!PageLocked(page))
+		BUG();
+	bh = head;
+	do {
+		if (buffer_locked(bh) || !buffer_dirty(bh) || !buffer_uptodate(bh))
+			continue;
+
+		bh->b_flushtime = jiffies;
+		ll_rw_block(WRITE, 1, &bh);	
+	} while ((bh = bh->b_this_page) != head);
+	return 0;
+}
+EXPORT_SYMBOL(writeout_one_page);
+
+/*
+ * Wait for completion of I/O of all buffers against a page.  The page
+ * must be locked.  Returns zero on success or a negative errno.
+ */
+int waitfor_one_page(struct page *page)
+{
+	int error = 0;
+	struct buffer_head *bh, *head = page->buffers;
+
+	bh = head;
+	do {
+		wait_on_buffer(bh);
+		if (buffer_req(bh) && !buffer_uptodate(bh))
+			error = -EIO;
+	} while ((bh = bh->b_this_page) != head);
+	return error;
+}
+EXPORT_SYMBOL(waitfor_one_page);
+
 sector_t generic_block_bmap(struct address_space *mapping, sector_t block,
 			    get_block_t *get_block)
 {

fs/ext2/dir.c

@@ -52,8 +52,13 @@ static int ext2_commit_chunk(struct page *page, unsigned from, unsigned to)
 	int err = 0;
 	dir->i_version = ++event;
 	page->mapping->a_ops->commit_write(NULL, page, from, to);
-	if (IS_SYNC(dir))
-		err = waitfor_one_page(page);
+	if (IS_SYNC(dir)) {
+		int err2;
+		err = writeout_one_page(page);
+		err2 = waitfor_one_page(page);
+		if (err == 0)
+			err = err2;
+	}
 	return err;
 }
 

fs/minix/dir.c

@@ -36,8 +36,13 @@ static int dir_commit_chunk(struct page *page, unsigned from, unsigned to)
 	struct inode *dir = (struct inode *)page->mapping->host;
 	int err = 0;
 	page->mapping->a_ops->commit_write(NULL, page, from, to);
-	if (IS_SYNC(dir))
-		err = waitfor_one_page(page);
+	if (IS_SYNC(dir)) {
+		int err2;
+		err = writeout_one_page(page);
+		err2 = waitfor_one_page(page);
+		if (err == 0)
+			err = err2;
+	}
 	return err;
 }
 
@@ -236,10 +241,10 @@ int minix_delete_entry(struct minix_dir_entry *de, struct page *page)
 
 	lock_page(page);
 	err = mapping->a_ops->prepare_write(NULL, page, from, to);
-	if (err)
-		BUG();
-	de->inode = 0;
-	err = dir_commit_chunk(page, from, to);
+	if (err == 0) {
+		de->inode = 0;
+		err = dir_commit_chunk(page, from, to);
+	}
 	UnlockPage(page);
 	dir_put_page(page);
 	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
@@ -336,10 +341,10 @@ void minix_set_link(struct minix_dir_entry *de, struct page *page,
 
 	lock_page(page);
 	err = page->mapping->a_ops->prepare_write(NULL, page, from, to);
-	if (err)
-		BUG();
-	de->inode = inode->i_ino;
-	err = dir_commit_chunk(page, from, to);
+	if (err == 0) {
+		de->inode = inode->i_ino;
+		err = dir_commit_chunk(page, from, to);
+	}
 	UnlockPage(page);
 	dir_put_page(page);
 	dir->i_mtime = dir->i_ctime = CURRENT_TIME;

fs/nfs/file.c

@@ -244,6 +244,7 @@ nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
 {
 	struct inode * inode = filp->f_dentry->d_inode;
 	int	status = 0;
+	int	status2;
 
 	dprintk("NFS: nfs_lock(f=%s/%ld, t=%x, fl=%x, r=%Ld:%Ld)\n",
 			inode->i_sb->s_id, inode->i_ino,
@@ -278,11 +279,15 @@ nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
 	 * Flush all pending writes before doing anything
 	 * with locks..
 	 */
-	filemap_fdatasync(inode->i_mapping);
+	status = filemap_fdatasync(inode->i_mapping);
 	down(&inode->i_sem);
-	status = nfs_wb_all(inode);
+	status2 = nfs_wb_all(inode);
+	if (status2 && !status)
+		status = status2;
 	up(&inode->i_sem);
-	filemap_fdatawait(inode->i_mapping);
+	status2 = filemap_fdatawait(inode->i_mapping);
+	if (status2 && !status)
+		status = status2;
 	if (status < 0)
 		return status;
 

fs/sysv/ChangeLog

+Thu Feb 14 2002  Andrew Morton  <akpm@zip.com.au>
+
+	* dir_commit_chunk(): call writeout_one_page() as well as
+	  waitfor_one_page() for IS_SYNC directories, so that we
+	  actually do sync the directory. (forward-port from 2.4).
+
 Thu Feb  7 2002  Alexander Viro  <viro@math.psu.edu>
 
 	* super.c: switched to ->get_sb()
@@ -97,3 +103,4 @@ Fri Oct 26 2001  Christoph Hellwig  <hch@caldera.de>
 	Remove symlink faking.	Noone really wants to use these as
 	linux filesystems and native OSes don't support it anyway.
 
+

fs/sysv/dir.c

@@ -42,8 +42,13 @@ static int dir_commit_chunk(struct page *page, unsigned from, unsigned to)
 	int err = 0;
 
 	page->mapping->a_ops->commit_write(NULL, page, from, to);
-	if (IS_SYNC(dir))
-		err = waitfor_one_page(page);
+	if (IS_SYNC(dir)) {
+		int err2;
+		err = writeout_one_page(page);
+		err2 = waitfor_one_page(page);
+		if (err == 0)
+			err = err2;
+	}
 	return err;
 }
 

include/linux/fs.h

@@ -1230,8 +1230,8 @@ static inline int fsync_inode_data_buffers(struct inode *inode)
 	return fsync_buffers_list(&inode->i_dirty_data_buffers);
 }
 extern int inode_has_buffers(struct inode *);
-extern void filemap_fdatasync(struct address_space *);
-extern void filemap_fdatawait(struct address_space *);
+extern int filemap_fdatasync(struct address_space *);
+extern int filemap_fdatawait(struct address_space *);
 extern void sync_supers(kdev_t);
 extern int bmap(struct inode *, int);
 extern int notify_change(struct dentry *, struct iattr *);
@@ -1440,8 +1440,9 @@ sector_t generic_block_bmap(struct address_space *, sector_t, get_block_t *);
 int generic_commit_write(struct file *, struct page *, unsigned, unsigned);
 int block_truncate_page(struct address_space *, loff_t, get_block_t *);
 extern int generic_direct_IO(int, struct inode *, struct kiobuf *, unsigned long, int, get_block_t *);
+extern int waitfor_one_page(struct page *);
+extern int writeout_one_page(struct page *);
 
-extern int waitfor_one_page(struct page*);
 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
 extern int file_read_actor(read_descriptor_t * desc, struct page *page, unsigned long offset, unsigned long size);
 extern ssize_t generic_file_read(struct file *, char *, size_t, loff_t *);

kernel/ksyms.c

@@ -213,7 +213,6 @@ EXPORT_SYMBOL(cont_prepare_write);
 EXPORT_SYMBOL(generic_commit_write);
 EXPORT_SYMBOL(block_truncate_page);
 EXPORT_SYMBOL(generic_block_bmap);
-EXPORT_SYMBOL(waitfor_one_page);
 EXPORT_SYMBOL(generic_file_read);
 EXPORT_SYMBOL(do_generic_file_read);
 EXPORT_SYMBOL(generic_file_write);

mm/filemap.c

@@ -449,41 +449,6 @@ static inline struct page * __find_page_nolock(struct address_space *mapping, un
 	return page;
 }
 
-/*
- * By the time this is called, the page is locked and
- * we don't have to worry about any races any more.
- *
- * Start the IO..
- */
-static int writeout_one_page(struct page *page)
-{
-	struct buffer_head *bh, *head = page->buffers;
-
-	bh = head;
-	do {
-		if (buffer_locked(bh) || !buffer_dirty(bh) || !buffer_uptodate(bh))
-			continue;
-
-		bh->b_flushtime = jiffies;
-		ll_rw_block(WRITE, 1, &bh);	
-	} while ((bh = bh->b_this_page) != head);
-	return 0;
-}
-
-int waitfor_one_page(struct page *page)
-{
-	int error = 0;
-	struct buffer_head *bh, *head = page->buffers;
-
-	bh = head;
-	do {
-		wait_on_buffer(bh);
-		if (buffer_req(bh) && !buffer_uptodate(bh))
-			error = -EIO;
-	} while ((bh = bh->b_this_page) != head);
-	return error;
-}
-
 static int do_buffer_fdatasync(struct list_head *head, unsigned long start, unsigned long end, int (*fn)(struct page *))
 {
 	struct list_head *curr;
@@ -577,8 +542,9 @@ EXPORT_SYMBOL(fail_writepage);
  *      @mapping: address space structure to write
  *
  */
-void filemap_fdatasync(struct address_space * mapping)
+int filemap_fdatasync(struct address_space * mapping)
 {
+	int ret = 0;
 	int (*writepage)(struct page *) = mapping->a_ops->writepage;
 
 	spin_lock(&pagecache_lock);
@@ -598,8 +564,11 @@ void filemap_fdatasync(struct address_space * mapping)
 		lock_page(page);
 
 		if (PageDirty(page)) {
+			int err;
 			ClearPageDirty(page);
-			writepage(page);
+			err = writepage(page);
+			if (err && !ret)
+				ret = err;
 		} else
 			UnlockPage(page);
 
@@ -607,6 +576,7 @@ void filemap_fdatasync(struct address_space * mapping)
 		spin_lock(&pagecache_lock);
 	}
 	spin_unlock(&pagecache_lock);
+	return ret;
 }
 
 /**
@@ -616,8 +586,10 @@ void filemap_fdatasync(struct address_space * mapping)
  *      @mapping: address space structure to wait for
  *
  */
-void filemap_fdatawait(struct address_space * mapping)
+int filemap_fdatawait(struct address_space * mapping)
 {
+	int ret = 0;
+
 	spin_lock(&pagecache_lock);
 
         while (!list_empty(&mapping->locked_pages)) {
@@ -633,11 +605,14 @@ void filemap_fdatawait(struct address_space * mapping)
 		spin_unlock(&pagecache_lock);
 
 		___wait_on_page(page);
+		if (PageError(page))
+			ret = -EIO;
 
 		page_cache_release(page);
 		spin_lock(&pagecache_lock);
 	}
 	spin_unlock(&pagecache_lock);
+	return ret;
 }
 
 /*
@@ -1514,12 +1489,14 @@ static ssize_t generic_file_direct_IO(int rw, struct file * filp, char * buf, si
 		goto out_free;
 
 	/*
-	 * Flush to disk exlusively the _data_, metadata must remains
+	 * Flush to disk exclusively the _data_, metadata must remain
 	 * completly asynchronous or performance will go to /dev/null.
 	 */
-	filemap_fdatasync(mapping);
-	retval = fsync_inode_data_buffers(inode);
-	filemap_fdatawait(mapping);
+	retval = filemap_fdatasync(mapping);
+	if (retval == 0)
+		retval = fsync_inode_data_buffers(inode);
+	if (retval == 0)
+		retval = filemap_fdatawait(mapping);
 	if (retval < 0)
 		goto out_free;
 
@@ -2136,26 +2113,45 @@ int generic_file_mmap(struct file * file, struct vm_area_struct * vma)
  * The msync() system call.
  */
 
+/*
+ * MS_SYNC syncs the entire file - including mappings.
+ *
+ * MS_ASYNC initiates writeout of just the dirty mapped data.
+ * This provides no guarantee of file integrity - things like indirect
+ * blocks may not have started writeout.  MS_ASYNC is primarily useful
+ * where the application knows that it has finished with the data and
+ * wishes to intelligently schedule its own I/O traffic.
+ */
 static int msync_interval(struct vm_area_struct * vma,
 	unsigned long start, unsigned long end, int flags)
 {
+	int ret = 0;
 	struct file * file = vma->vm_file;
+
 	if (file && (vma->vm_flags & VM_SHARED)) {
-		int error;
-		error = filemap_sync(vma, start, end-start, flags);
+		ret = filemap_sync(vma, start, end-start, flags);
 
-		if (!error && (flags & MS_SYNC)) {
+		if (!ret && (flags & (MS_SYNC|MS_ASYNC))) {
 			struct inode * inode = file->f_dentry->d_inode;
+
 			down(&inode->i_sem);
-			filemap_fdatasync(inode->i_mapping);
-			if (file->f_op && file->f_op->fsync)
-				error = file->f_op->fsync(file, file->f_dentry, 1);
-			filemap_fdatawait(inode->i_mapping);
+			ret = filemap_fdatasync(inode->i_mapping);
+			if (flags & MS_SYNC) {
+				int err;
+
+				if (file->f_op && file->f_op->fsync) {
+					err = file->f_op->fsync(file, file->f_dentry, 1);
+					if (err && !ret)
+						ret = err;
+				}
+				err = filemap_fdatawait(inode->i_mapping);
+				if (err && !ret)
+					ret = err;
+			}
 			up(&inode->i_sem);
 		}
-		return error;
 	}
-	return 0;
+	return ret;
 }
 
 asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
@@ -2999,7 +2995,7 @@ generic_file_write(struct file *file,const char *buf,size_t count, loff_t *ppos)
 		kaddr = kmap(page);
 		status = mapping->a_ops->prepare_write(file, page, offset, offset+bytes);
 		if (status)
-			goto unlock;
+			goto sync_failure;
 		page_fault = __copy_from_user(kaddr+offset, buf, bytes);
 		flush_dcache_page(page);
 		status = mapping->a_ops->commit_write(file, page, offset, offset+bytes);
@@ -3024,6 +3020,7 @@ generic_file_write(struct file *file,const char *buf,size_t count, loff_t *ppos)
 		if (status < 0)
 			break;
 	} while (count);
+done:
 	*ppos = pos;
 
 	if (cached_page)
@@ -3046,6 +3043,18 @@ generic_file_write(struct file *file,const char *buf,size_t count, loff_t *ppos)
 	status = -EFAULT;
 	goto unlock;
 
+sync_failure:
+	/*
+	 * If blocksize < pagesize, prepare_write() may have instantiated a
+	 * few blocks outside i_size.  Trim these off again.
+	 */
+	kunmap(page);
+	UnlockPage(page);
+	page_cache_release(page);
+	if (pos + bytes > inode->i_size)
+		vmtruncate(inode, inode->i_size);
+	goto done;
+
 o_direct:
 	written = generic_file_direct_IO(WRITE, file, (char *) buf, count, pos);
 	if (written > 0) {