linux-kernel - Re: [GIT PULL] overlayfs update for 6.16

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAHk-=wgH174aR4HnpmV7yVYVjS7VmSRC31md5di7_Cr_v0Afqg@mail.gmail.com>
Date: Thu, 5 Jun 2025 12:34:19 -0700
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Miklos Szeredi <miklos@...redi.hu>
Cc: overlayfs <linux-unionfs@...r.kernel.org>, linux-fsdevel@...r.kernel.org, 
	linux-kernel@...r.kernel.org
Subject: Re: [GIT PULL] overlayfs update for 6.16

On Thu, 5 Jun 2025 at 07:51, Miklos Szeredi <miklos@...redi.hu> wrote:
>
> - The above fix contains a cast to non-const, which is not actually
> needed.  So add the necessary helpers postfixed with _c that allow the
> cast to be removed (touches vfs files but only in trivial ways)

Grr.

I despise those "trivial ways".

In particular, I despise how this renames the backing_file_user_path()
helper to something actively *worse*.

The "_c()" makes no sense as a name. Yes, I realize that the "c"
stands for "const", but it still makes absolutely zero sense, since
everybody wants the const version.

The only user of the non-const version is the *ointernal*
implementation that is never exported to other modules, and that could
have the special name.

Although I suspect it doesn't even need it, it could just use the
backing_file(f) macro directly and that should just be moved to
internal.h, and then the 'const'ness would come from the argument as
required.

In fact, most of the _internal_ vfs users don't even want the
non-const version, ie as far as I can tell the user in
file_get_write_access() would be perfectly happy with the const
version too.

So you made the *normal* case have an odd name, and then kept the old
sane name for the case nobody else really wants to see.

If anything, the internal non-const version is the one that should be
renamed (and *not* using some crazy "_nc()" postfix nasty crud). Not
the one that gets exported and that everybody wants.

So I could fix up that last commit to not hate it, but honestly, I
don't want that broken state in the kernel in the first place.

Please do that thing properly. Not this hacky and bass-ackwards way.

             Linus