Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Mime-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset=UTF-8
Cc:     <kernel-team@fb.com>, <mhiramat@kernel.org>
Subject: Re: [PATCH bpf-next] lib/strncpy_from_user.c: Don't overcopy bytes
 after NUL terminator
From:   "Daniel Xu" <dxu@dxuuu.xyz>
To:     "Daniel Borkmann" <daniel@iogearbox.net>, <bpf@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <ast@kernel.org>
Date:   Wed, 04 Nov 2020 18:21:57 -0800
Message-Id: <C6UYZS8KPY7V.23QPWOJX7R50T@maharaja>
In-Reply-To: <7d1a34fa-2475-0958-37fe-ed416249bc4b@iogearbox.net>
Precedence: bulk

On Wed Nov 4, 2020 at 2:36 PM PST, Daniel Borkmann wrote:
> On 11/4/20 9:18 PM, Daniel Xu wrote:
> > On Wed Nov 4, 2020 at 8:24 AM PST, Daniel Borkmann wrote:
> >> On 11/4/20 3:29 AM, Daniel Xu wrote:
> >>> do_strncpy_from_user() may copy some extra bytes after the NUL
> >>> terminator into the destination buffer. This usually does not matter =
for
> >>> normal string operations. However, when BPF programs key BPF maps wit=
h
> >>> strings, this matters a lot.
> >>>
> >>> A BPF program may read strings from user memory by calling the
> >>> bpf_probe_read_user_str() helper which eventually calls
> >>> do_strncpy_from_user(). The program can then key a map with the
> >>> resulting string. BPF map keys are fixed-width and string-agnostic,
> >>> meaning that map keys are treated as a set of bytes.
> >>>
> >>> The issue is when do_strncpy_from_user() overcopies bytes after the N=
UL
> >>> terminator, it can result in seemingly identical strings occupying
> >>> multiple slots in a BPF map. This behavior is subtle and totally
> >>> unexpected by the user.
> >>>
> >>> This commit uses the proper word-at-a-time APIs to avoid overcopying.
> >>>
> >>> Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
> >>
> >> It looks like this is a regression from the recent refactoring of the
> >> mem probing
> >> util functions?
> >=20
> > I think it was like this from the beginning, at 6ae08ae3dea2 ("bpf: Add
> > probe_read_{user, kernel} and probe_read_{user, kernel}_str helpers").
> > The old bpf_probe_read_str() used the kernel's byte-by-byte copying
> > routine. bpf_probe_read_user_str() started using strncpy_from_user()
> > which has been doing the long-sized strides since ~2012 or earlier.
> >=20
> > I tried to build and test the kernel at that commit but it seems my
> > compiler is too new to build that old code. Bunch of build failures.
> >=20
> > I assume the refactor you're referring to is 8d92db5c04d1 ("bpf: rework
> > the compat kernel probe handling").
>
> Ah I see, it was just reusing 3d7081822f7f ("uaccess: Add non-pagefault
> user-space
> read functions"). Potentially it might be safer choice to just rework
> the
> strncpy_from_user_nofault() to mimic strncpy_from_kernel_nofault() in
> that
> regard?

I'm a little reluctant to do that b/c it would introduce less efficient,
duplicated code. The word-at-a-time API already has the zero_bytemask()
API so it's clear that it was designed to handle this issue -- we're not
really hacking anything here.

I'll send out a V2 with the selftest shortly. Happy to change things
after that.

Thanks,
Daniel