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
References: <20201103153132.2717326-1-kpsingh@chromium.org>
 <20201103153132.2717326-8-kpsingh@chromium.org> <20201103184714.iukuqfw2byls3s4k@ast-mbp.dhcp.thefacebook.com>
 <CACYkzJ6A5GrQhBhv7GC8aeeLpoc7bnN=6Rn2UoM1P90odLZZ=g@mail.gmail.com> <CACYkzJ6D=vwaEhgaB2vevOo0186m=yfxeKBQ8eWWck8xjtczNA@mail.gmail.com>
In-Reply-To: <CACYkzJ6D=vwaEhgaB2vevOo0186m=yfxeKBQ8eWWck8xjtczNA@mail.gmail.com>
From:   Alexei Starovoitov <alexei.starovoitov@gmail.com>
Date:   Tue, 3 Nov 2020 17:27:22 -0800
Message-ID: <CAADnVQ+DBHXkf8SFwnTKmSKi7mdAx56dWbpp5++Cc02CQjz+Ng@mail.gmail.com>
Subject: Re: [PATCH bpf-next v2 7/8] bpf: Add tests for task_local_storage
To:     KP Singh <kpsingh@chromium.org>
Cc:     open list <linux-kernel@vger.kernel.org>,
        bpf <bpf@vger.kernel.org>, Alexei Starovoitov <ast@kernel.org>,
        Daniel Borkmann <daniel@iogearbox.net>,
        Martin KaFai Lau <kafai@fb.com>,
        Song Liu <songliubraving@fb.com>, Paul Turner <pjt@google.com>,
        Jann Horn <jannh@google.com>, Hao Luo <haoluo@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Nov 3, 2020 at 4:05 PM KP Singh <kpsingh@chromium.org> wrote:
>
> On Tue, Nov 3, 2020 at 7:59 PM KP Singh <kpsingh@chromium.org> wrote:
> >
> > On Tue, Nov 3, 2020 at 7:47 PM Alexei Starovoitov
> > <alexei.starovoitov@gmail.com> wrote:
> > >
> > > On Tue, Nov 03, 2020 at 04:31:31PM +0100, KP Singh wrote:
> > > > +
> > > > +struct storage {
> > > > +     void *inode;
> > > > +     unsigned int value;
> > > > +     /* Lock ensures that spin locked versions of local stoage operations
> > > > +      * also work, most operations in this tests are still single threaded
> > > > +      */
> > > > +     struct bpf_spin_lock lock;
> > > > +};
> > >
> > > I think it's a good idea to test spin_lock in local_storage,
> > > but it seems the test is not doing it fully.
> > > It's only adding it to the storage, but the program is not accessing it.
> >
> > I added it here just to check if the offset calculations (map->spin_lock_off)
> > are correctly happening for these new maps.
> >
> > As mentioned in the updates, I do intend to generalize
> > tools/testing/selftests/bpf/map_tests/sk_storage_map.c which already has
> >  the threading logic to exercise bpf_spin_lock in storage maps.
> >
>
> Actually, after I added simple bpf_spin_{lock, unlock} to the test programs, I
> ended up realizing that we have not exposed spin locks to LSM programs
> for now, this is because they inherit the tracing helpers.
>
> I saw the docs mention that these are not exposed to tracing programs due to
> insufficient preemption checks. Do you think it would be okay to allow them
> for LSM programs?

hmm. Isn't it allowed already?
The verifier does:
        if ((is_tracing_prog_type(prog_type) ||
             prog_type == BPF_PROG_TYPE_SOCKET_FILTER) &&
            map_value_has_spin_lock(map)) {
                verbose(env, "tracing progs cannot use bpf_spin_lock yet\n");
                return -EINVAL;
        }

BPF_PROG_TYPE_LSM is not in this list.