Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <87edkce118.wl-tiwai@suse.de> <20230809143801.GA693@lst.de>
In-Reply-To: <20230809143801.GA693@lst.de>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed, 9 Aug 2023 08:48:11 -0700
Message-ID: <CAHk-=wiyWOaPtOJ1PTdERswXV9m7W_UkPV-HE0kbpr48mbnrEA@mail.gmail.com>
Subject: Re: [PATCH RFC] Introduce uniptr_t as a generic "universal" pointer
To:     Christoph Hellwig <hch@lst.de>
Cc:     Takashi Iwai <tiwai@suse.de>, linux-kernel@vger.kernel.org,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Mark Brown <broonie@kernel.org>, netdev@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, 9 Aug 2023 at 07:38, Christoph Hellwig <hch@lst.de> wrote:
>
> The original set_fs removal series did that as uptr_t, and Linus
> hated it with passion.  I somehow doubt he's going to like it more now.

Christoph is right. I do hate this. The whole "pass a pointer that is
either user or kernel" concept is wrong.

Now, if it was some kind of extended pointer that also included the
length of the area and had a way to deal with updating the pointer
sanely, maybe that would be a different thing.

And it should guarantee that in the case of a user pointer it had gone
through access_ok().

And it also allowed the other common cases like having a raw page
array, along with a unified interface to copy and update this kind of
pointer either as a source or a destination, that would be a different
thing.

But this kind of "if (uniptr_is_kernel(src))" special case thing is
just garbage and *not* acceptable.

And oh, btw, we already *have* that extended kind of unipointer thing.

It's called "struct iov_iter".

And yes, it's a very complicated thing, exactly because it handles way
more cases than that uniptr_t. It's a *real* unified pointer of many
different types.

Those iov_iter things are often so complicated that you really don't
want to use them, but if you really want a uniptr, that is what you
should do. It comes with a real cost, but it does come with real
advantages, one of which is "this is extensively tested
nfrastructure".

                  Linus