Received-SPF: pass (google.com: domain of linux-kernel+bounces-16059-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: <20240103102553.17a19cea@gandalf.local.home> <CAHk-=whrRobm82kcjwj625bZrdK+vvEo0B5PBzP+hVaBcHUkJA@mail.gmail.com>
 <CAHk-=wjVdGkjDXBbvLn2wbZnqP4UsH46E3gqJ9m7UG6DpX2+WA@mail.gmail.com>
 <20240103145306.51f8a4cd@gandalf.local.home> <CAHk-=wg=tnnsTjnzTs8xRQOBLvw4ceKe7=yxfzNtx4Z9gb-xJw@mail.gmail.com>
 <CAHk-=wh5kkk2+JAv_D1fm8t1SOpTQyb4n7zuMuVSBG094HH7gA@mail.gmail.com> <20240103170556.28df7163@gandalf.local.home>
In-Reply-To: <20240103170556.28df7163@gandalf.local.home>
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Wed, 3 Jan 2024 14:14:08 -0800
Message-ID: <CAHk-=wiWbqy-hd-Xo6dwx=6dH+gMyvf_wFtih7eEpBkHyigzJA@mail.gmail.com>
Subject: Re: [PATCH] eventfs: Stop using dcache_readdir() for getdents()
To: Steven Rostedt <rostedt@goodmis.org>
Cc: LKML <linux-kernel@vger.kernel.org>, 
	Linux Trace Kernel <linux-trace-kernel@vger.kernel.org>, Masami Hiramatsu <mhiramat@kernel.org>, 
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>, Al Viro <viro@zeniv.linux.org.uk>, 
	Christian Brauner <brauner@kernel.org>, linux-fsdevel@vger.kernel.org, 
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Content-Type: text/plain; charset="UTF-8"

On Wed, 3 Jan 2024 at 14:04, Steven Rostedt <rostedt@goodmis.org> wrote:
>
> I actually have something almost working too. Here's the WIP. It only works
> for tracefs, and now eventfs needs to be updated as the "events" directory
> no longer has the right ownership. So I need a way to link the eventfs
> entries to use the tracefs default conditionally.

Ack. So the ->getattr() and ->permission() thing is a bit more
targeted to "look at modes", and is probably better just for that
reason.

Doing it in d_revalidate() is a bit hacky, and impacts path lookup a
bit even when not necessary. But it's still a lot less evil than
walking the dentry tree manually.

So that d_revalidate() patch was more of a "I think you can make it
smaller by just hooking in at this layer"). So d_revalidate ends up
with a smaller patch, I think, but it has the problem that now you
*have* to be able to deal with things in RCU context.

In contrast, doing it in ->getattr() and ->permission() ends up
meaning you can use sleeping locks etc if you need to serialize, for
example. So it's a bit more specific to the whole issue of "deal with
modes and owndership", but it is *also* a bit more flexible in how you
can then do it.

Anyway, your patch looks fine from a quick scan.

                  Linus