Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:1 as permitted sender) client-ip=2620:137:e000::3:1;
MIME-Version: 1.0
References: <20231118025748.2778044-1-mmaurer@google.com> <CAK7LNAQt8fy5+vSwpd1aXfzjzeZ5hiyW7EW9SW7pbG2eTJZAOA@mail.gmail.com>
 <CAGSQo00hyCTVsqHtrzKBBPvuH38z5yRm_4jzdi00C0RV+8APwQ@mail.gmail.com> <2023112314-tubby-eligibly-007a@gregkh>
In-Reply-To: <2023112314-tubby-eligibly-007a@gregkh>
From:   Masahiro Yamada <masahiroy@kernel.org>
Date:   Thu, 23 Nov 2023 20:38:45 +0900
Message-ID: <CAK7LNAT-OcaCi6tqPRgZxPXOV6u+YbaO_0RxtfmrVXPzdrio0Q@mail.gmail.com>
Subject: Re: [PATCH v2 0/5] MODVERSIONS + RUST Redux
To:     Greg KH <gregkh@linuxfoundation.org>
Cc:     Matthew Maurer <mmaurer@google.com>,
        Nick Desaulniers <ndesaulniers@google.com>,
        Miguel Ojeda <ojeda@kernel.org>, Gary Guo <gary@garyguo.net>,
        Luis Chamberlain <mcgrof@kernel.org>,
        Nathan Chancellor <nathan@kernel.org>,
        Nicolas Schier <nicolas@fjasle.eu>,
        linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org,
        linux-modules@vger.kernel.org, linux-kbuild@vger.kernel.org,
        rust-for-linux@vger.kernel.org, Laura Abbott <laura@labbott.name>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Nov 23, 2023 at 6:05=E2=80=AFPM Greg KH <gregkh@linuxfoundation.org=
> wrote:
>
> On Wed, Nov 22, 2023 at 01:04:09PM -0800, Matthew Maurer wrote:
> > > So, even if you enable CONFIG_MODVERSIONS,
> > > nothing is checked for Rust.
> > > Genksyms computes a CRC from "int foo", and
> > > the module subsystem confirms it is a "int"
> > > variable.
> > >
> > > We know this check always succeeds.
> > >
> > > Why is this useful?
> > The reason this is immediately useful is that it allows us to have Rust
> > in use with a kernel where C modules are able to benefit from MODVERSIO=
NS
> > checking. The check would effectively be a no-op for now, as you have c=
orrectly
> > determined, but we could refine it to make it more restrictive later.
> > Since the
> > existing C approach errs on the side of "it could work" rather than "it=
 will
> > work", I thought being more permissive was the correct initial solution=
.
>
> But it's just providing "fake" information to the CRC checker, which
> means that the guarantee of a ABI check is not true at all.
>
> So the ask for the user of "ensure that the ABI checking is correct" is
> being circumvented here, and any change in the rust side can not be
> detected at all.
>
> The kernel is a "whole", either an option works for it, or it doesn't,
> and you are splitting that guarantee here by saying "modversions will
> only work for a portion of the kernel, not the whole thing" which is
> going to cause problems for when people expect it to actually work
> properly.
>
> So, I'd strongly recommend fixing this for the rust code if you wish to
> allow modversions to be enabled at all.
>
> > With regards to future directions that likely won't work for loosening =
it:
> > Unfortunately, the .rmeta format itself is not stable, so I wouldn't wa=
nt to
> > teach genksyms to open it up and split out the pieces for specific func=
tions.
> > Extending genksyms to parse Rust would also not solve the situation -
> > layouts are allowed to differ across compiler versions or even (in rare
> > cases) seemingly unrelated code changes.
>
> What do you mean by "layout" here?  Yes, the crcs can be different
> across compiler versions and seemingly unrelated code changes (genksyms
> is VERY fragile) but that's ok, that's not what you are checking here.
> You want to know if the rust function signature changes or not from the
> last time you built the code, with the same compiler and options, that's
> all you are verifying.
>
> > Future directions that might work for loosening it:
> > * Generating crcs from debuginfo + compiler + flags
> > * Adding a feature to the rust compiler to dump this information. This
> > is likely to
> >   get pushback because Rust's current stance is that there is no abilit=
y to load
> >   object code built against a different library.
>
> Why not parse the function signature like we do for C?
>
> > Would setting up Rust symbols so that they have a crc built out of .rme=
ta be
> > sufficient for you to consider this useful? If not, can you help me und=
erstand
> > what level of precision would be required?
>
> What exactly does .rmeta have to do with the function signature?  That's
> all you care about here.


rmeta is generated per crate.

CRC is computed per symbol.

They have different granularity.
It is weird to refuse a module for incompatibility
of a symbol that it is not using at all.


> thanks,
>
> greg k-h


--
Best Regards
Masahiro Yamada