Re: [PATCH 6/6] efivarfs: fix error on write to new variable leaving remnants

[Al Viro <viro@xxxxxxxxxxxxxxxxxx>]

On Mon, Dec 23, 2024 at 11:04:58PM -0500, James Bottomley wrote:

> +static int efivarfs_file_release(struct inode *inode, struct file *file)
> +{
> +	if (i_size_read(inode) == 0)
> +		simple_recursive_removal(file->f_path.dentry, NULL);
> +
> +	return 0;
> +}

What happens if you have

	fd = creat(name, 0700);
	fd2 = open(name, O_RDONLY);
	close(fd2);
	write(fd, "barf", 4);

or, better yet, if open()/close() pair happens in an unrelated thread
poking around?

I mean, having that logics in ->release() feels very awkward...

For that matter, what about
	fd = creat(name, 0700);
	fd2 = open(name, O_RDWR);
	close(fd);
	write(fd2, "barf", 4);

I'm not asking about the implementation; what behaviour do you want
to see in userland?