linux-kernel - Re: [GIT PULL] overlayfs updates for 6.13

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-=whoEWma-c-ZTj=fpXtD+1EyYimW4TwqDV9FeUVVfzwang@mail.gmail.com>
Date: Sat, 23 Nov 2024 09:29:05 -0800
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Christian Brauner <brauner@...nel.org>
Cc: Amir Goldstein <amir73il@...il.com>, Miklos Szeredi <miklos@...redi.hu>, linux-kernel@...r.kernel.org, 
	linux-fsdevel@...r.kernel.org, linux-unionfs@...r.kernel.org
Subject: Re: [GIT PULL] overlayfs updates for 6.13

On Sat, 23 Nov 2024 at 04:06, Christian Brauner <brauner@...nel.org> wrote:
>
> So just to clarify when that issue was brought up I realized that the
> cred bump was a big deal for overlayfs but from a quick grep I didn't
> think for any of the other cases it really mattered that much.

Oh, I agree. It's probably not really a performance issue anywhere
else. I don't think this has really ever come up before.

So my "please convert everything to one single new model" is not
because I think that would help performance, but because I really hate
having two differently flawed models when I think one would do.

We have other situations where we really do have two or more different
interfaces for the "same" thing, with very special rules: things like
fget() vs fget_raw() vs fget_task() (and similar issues wrt fdget).

But I think those other situations have more _reason_ for them.

The whole "override_creds()" thing is _already_ such a special
operation, that I hate seeing two subtly different versions of the
interface, both with their own quirks.

Because the old interface really isn't some "perfectly tailored"
thing. Yes, the performance implications were a surprise to me and I
hadn't seen that before, but the "refcounting isn't wonderful" was
_not_ really a big surprise at all.

                        Linus