Re: [PATCH 1/2] kernel/acct.c: saner struct file treatment

[Al Viro <viro@zeniv.linux.org.uk>]

On Thu, Sep 25, 2025 at 02:28:16PM +0200, Jan Kara wrote:

> This is mostly harmless (just the returned error code changed) but it is a
> side effect of one thing I'd like to discuss: The original code uses
> file_open_name(O_WRONLY|O_APPEND|O_LARGEFILE) to open the file to write
> data to. The new code uses dentry_open() for that. Now one important
> difference between these two is that dentry_open() doesn't end up calling
> may_open() on the path and hence now acct_on() fails to check file
> permissions which looks like a bug? Am I missing something Al?

You are not; a bug it is.  FWIW, I suspect that the right approach would
be to keep file_open_name(), do all checks on result of that, then use
	mnt = mnt_clone_internal(&original_file->f_path);
and from that point on same as now - opening the file to be used with
dentry_open(), etc.  Original file would get dropped in the end of acct_on().
I'll put together something along those lines and post it.