From: Alexei Starovoitov <alexei.starovoitov@gmail.com>
To: David Vernet <void@manifault.com>
Cc: bpf <bpf@vger.kernel.org>, Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Andrii Nakryiko <andrii@kernel.org>,
Martin KaFai Lau <martin.lau@linux.dev>,
Song Liu <song@kernel.org>, Yonghong Song <yhs@fb.com>,
John Fastabend <john.fastabend@gmail.com>,
KP Singh <kpsingh@kernel.org>,
Stanislav Fomichev <sdf@google.com>, Hao Luo <haoluo@google.com>,
Jiri Olsa <jolsa@kernel.org>, LKML <linux-kernel@vger.kernel.org>,
Kernel Team <kernel-team@fb.com>, Tejun Heo <tj@kernel.org>,
Kumar Kartikeya Dwivedi <memxor@gmail.com>
Subject: Re: [PATCH bpf-next v6 1/3] bpf: Allow trusted pointers to be passed to KF_TRUSTED_ARGS kfuncs
Date: Tue, 1 Nov 2022 13:22:39 -0700 [thread overview]
Message-ID: <CAADnVQ+KZcFZdC=W_qZ3kam9yAjORtpN-9+Ptg_Whj-gRxCZNQ@mail.gmail.com> (raw)
In-Reply-To: <Y2FhXC/s5GUkbr9P@maniforge.dhcp.thefacebook.com>
On Tue, Nov 1, 2022 at 11:11 AM David Vernet <void@manifault.com> wrote:
>
> > What kind of bpf prog will be able to pass 'struct nf_conn___init *' into these bpf_ct_* ?
> > We've introduced / vs nf_conf specifically to express the relationship
> > between allocated nf_conn and other nf_conn-s via different types.
> > Why is this not enough?
>
> Kumar should have more context here (he originally suggested this in
> [0]),
Quoting:
"
Unfortunately a side effect of this change is that now since
PTR_TO_BTF_ID without ref_obj_id is considered trusted, the bpf_ct_*
functions would begin working with tp_btf args.
"
I couldn't find any tracepoint that has nf_conn___init as an argument.
The whole point of that new type was to return it to bpf prog,
so the verifier type matches it when it's passed into bpf_ct_*
in turn.
So I don't see a need for a new OWNED flag still.
If nf_conn___init is passed into tracepoint it's a bug and
we gotta fix it.
> but AFAICT you're correct that this should be sufficient. I added
> a negative test case that correctly fails if a BPF program tries to call
> these helpers with a struct nf_conn* rather than a struct
> nf_conn__init*.
>
> [0]: https://lore.kernel.org/all/CAP01T77PTK+bD2mBrxJShKNPhEypT2+nSHcr3=uuJbrghv_wFg@mail.gmail.com/
>
> > I prefer to keep only one flag KF_TRUSTED_ARGS that kfunc-s need to use
> > and eventually make all kfuncs KF_TRUSTED_ARGS by default and remove that flag.
>
> Yes, KF_TRUSTED_ARGS really should be the default. As Kumar describes in
> [1], we'll have to figure out how to avoid trace progs with unsafe args
> from calling these kfuncs. Maybe the right thing to do is allow-listing
> rather than deny-listing, as you pointed out.
>
> [1]: https://lore.kernel.org/bpf/CAP01T77goGbF3GVithEuJ7yMQR9PxHNA9GXFODq_nfA66G=F9g@mail.gmail.com/
That is still the plan. more or less.
> > Separately...
> > I think there was a plan to strengthen PTR_TO_BTF_ID and apply PTR_UNTRUSTED.
>
> That would be nice if we could do it. I assume that the issue is we're
> breaking backwards compat if we do, so I'd be curious to hear what the
> plan was if you're aware. The only plan that I've seen so far is what
> Kumar spelled out above in [1] above.
Right. Backward compat with existing usage of PTR_TO_BTF_ID
is the main issue.
>
> > This PTR_WALKED looks like new thing.
> > If we really need it PTR_TO_BTF_ID should be allowlisted instead of denylisted
> > as PTR_WALKED is doing.
> > I mean we can introduce PTR_TRUSTED and add this flag to return value
> > of bpf_get_current_task_btf() and arguments of tracepoints.
> > As soon as any ptr walking is done we can clear PTR_TRUSTED to keep
> > backward compat behavior of PTR_TO_BTF_ID.
> > PTR_WALKED is sort-of doing the same, but not conservative enough.
> > Too many things produce PTR_TO_BTF_ID. Auditing it all is challenging.
>
> I very much prefer the idea of allowlisting instead of denylisting,
> though I wish we'd taken that approach from the start rather than going
> with PTR_UNTRUSTED. It feels wrong to have both PTR_UNTRUSTED and
> PTR_TRUSTED as type modifiers, as the absence of PTR_UNTRUSTED should
> (and currently does) imply PTR_TRUSTED.
I kind agree, but we gotta have both because of backward compat.
We cannot change PTR_TO_BTF_ID as a whole right now.
Note PTR_TO_BTF_ID appears in kfuncs too.
I'm proposing to use PTR_TO_BTF_ID | PTR_TRUSTED
only in tracepoint args and as return value from
certain helpers like bpf_get_current_task_btf().
afaik it's all safe. There is no uaf here.
uaf is for kfunc. Especially fexit.
Those will stay PTR_TO_BTF_ID. Without PTR_TRUSTED.
>
> If we're going to go with an allowlist approach, then I think we should
> just get rid of PTR_UNTRUSTED altogether. Is that what you're
> suggesting? Otherwise, if we don't get rid of PTR_UNTRUSTED, then
> PTR_WALKED seems like a more natural type modifier addition.
Eventually either PTR_TRUSTED or PTR_UNTRUSTED will be removed.
> > I might have missed earlier discussions on this patch set. Apologies if so.
>
> Just FYI, the main initial thread where this was all discussed was [2].
>
> [2]: https://lore.kernel.org/all/CAP01T76OR3J_P8YMq4ZgKHBpuZyA0zgsPy+tq9htbX=j6AVyOg@mail.gmail.com/
next prev parent reply other threads:[~2022-11-01 20:22 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-20 22:24 [PATCH bpf-next v6 0/3] Support storing struct task_struct objects as kptrs David Vernet
2022-10-20 22:24 ` [PATCH bpf-next v6 1/3] bpf: Allow trusted pointers to be passed to KF_TRUSTED_ARGS kfuncs David Vernet
2022-11-01 0:02 ` Alexei Starovoitov
2022-11-01 18:11 ` David Vernet
2022-11-01 20:22 ` Alexei Starovoitov [this message]
2022-11-01 21:36 ` David Vernet
2022-11-01 22:31 ` Kumar Kartikeya Dwivedi
2022-11-02 0:32 ` Alexei Starovoitov
2022-11-02 1:00 ` Kumar Kartikeya Dwivedi
2022-11-02 2:34 ` Alexei Starovoitov
2022-11-02 20:48 ` Kumar Kartikeya Dwivedi
2022-11-03 0:51 ` Alexei Starovoitov
2022-10-20 22:24 ` [PATCH bpf-next v6 2/3] bpf: Add kfuncs for storing struct task_struct * as a kptr David Vernet
2022-10-20 22:24 ` [PATCH bpf-next v6 3/3] bpf/selftests: Add selftests for new task kfuncs David Vernet
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAADnVQ+KZcFZdC=W_qZ3kam9yAjORtpN-9+Ptg_Whj-gRxCZNQ@mail.gmail.com' \
--to=alexei.starovoitov@gmail.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=haoluo@google.com \
--cc=john.fastabend@gmail.com \
--cc=jolsa@kernel.org \
--cc=kernel-team@fb.com \
--cc=kpsingh@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=martin.lau@linux.dev \
--cc=memxor@gmail.com \
--cc=sdf@google.com \
--cc=song@kernel.org \
--cc=tj@kernel.org \
--cc=void@manifault.com \
--cc=yhs@fb.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).