Received-SPF: pass (google.com: domain of linux-kernel+bounces-43377-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240126150209.367ff402@gandalf.local.home> <CAHk-=wgZEHwFRgp2Q8_-OtpCtobbuFPBmPTZ68qN3MitU-ub=Q@mail.gmail.com>
 <20240126162626.31d90da9@gandalf.local.home> <CAHk-=wj8WygQNgoHerp-aKyCwFxHeyKMguQszVKyJfi-=Yfadw@mail.gmail.com>
 <CAHk-=whNfNti-mn6vhL-v-WZnn0i7ZAbwSf_wNULJeyanhPOgg@mail.gmail.com>
 <8547159a-0b28-4d75-af02-47fc450785fa@efficios.com> <CAHk-=whAG6TM6PgH0YnsRe6U=RzL+JMvCi=_f0Bhw+q_7SSZuw@mail.gmail.com>
 <29be300d-00c4-4759-b614-2523864c074b@efficios.com> <CAHk-=wjpyv+fhxzV+XEQgsC+-HaouKT7Ns8qT31jkpN_Jm84_g@mail.gmail.com>
 <3120f1f0-eaf8-4058-9a65-bdbee28c68c9@efficios.com>
In-Reply-To: <3120f1f0-eaf8-4058-9a65-bdbee28c68c9@efficios.com>
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Mon, 29 Jan 2024 10:58:30 -0800
Message-ID: <CAHk-=wg8BrZEzjJ5kUyZzHPZmFqH6ooMN1gRBCofxxCfucgjaw@mail.gmail.com>
Subject: Re: [PATCH] eventfs: Have inodes have unique inode numbers
To: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Steven Rostedt <rostedt@goodmis.org>, LKML <linux-kernel@vger.kernel.org>, 
	Linux Trace Devel <linux-trace-devel@vger.kernel.org>, Masami Hiramatsu <mhiramat@kernel.org>, 
	Christian Brauner <brauner@kernel.org>, Ajay Kaher <ajay.kaher@broadcom.com>, 
	Geert Uytterhoeven <geert@linux-m68k.org>, linux-fsdevel <linux-fsdevel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"

On Mon, 29 Jan 2024 at 08:00, Mathieu Desnoyers
<mathieu.desnoyers@efficios.com> wrote:
>
> This breaks "cp -aH" and "cp -aL".

Do we care? Do we have a user that cares? Has anybody ever hit it?

Why would you ever do anything like that to tracefs filesystem?

In other words: my point is that tracefs just isn't a regular
filesystem. Never was, never will be.

And people should be *aware* of that. We should not say "hey, if it
doesn't work like a normal filesystem, it's a bug".

Try "cp -aL" on /proc, and guess what? It won't work all that well
either. For entirely *different* reasons. You'll get some variation of
"Input/output error"s, and insanely big files and quite possibly
you'll end up with recursive copying as you try to copy the file that
is /proc/self/fd/<output>.

It's just a nonsensical operation to do, and if somebody says "I can't
copy /proc on my system" it's a PEBKAC, not a kernel problem.

The "no regressions" rule is not about made-up "if I do this, behavior changes".

The "no regressions" rule is about *users*.

If you have an actual user that has been doing insane things, and we
change something, and now the insane thing no longer works, at that
point it's a regression, and we'll sigh, and go "Users are insane" and
have to fix it.

But if you have some random test that now behaves differently, it's
not a regression. It's a *warning* sign, sure: tests are useful.

So tests can show when something user-visible changed, and as such
they are a "there be monsters here" sign that maybe some user
experience will hit it too.

So I agree that "just use the same inode number" changes behavior. I
agree that it can be a bit of a danger. But these "look, I can see a
difference" isn't an argument.

And honestly, I have now spent *days* looking at tracefs, and I'm
finding core fundamental bugs that would cause actual oopses and/or
wild pointer accesses.

All of which makes me go "this code needs to be simpler and *cleaner*
and stop making problems".

In other words: tracefs is such a complete mess that I do not care one
*whit* about "cp -aL". I care about "this is actual kernel
instability".

           Linus