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;
Feedback-ID: i6a694271:Fastmail
Date:   Thu, 18 Aug 2022 13:31:27 -0600
From:   Daniel Xu <dxu@dxuuu.xyz>
To:     Kumar Kartikeya Dwivedi <memxor@gmail.com>
Cc:     bpf@vger.kernel.org, ast@kernel.org, daniel@iogearbox.net,
        andrii@kernel.org, pablo@netfilter.org, fw@strlen.de,
        toke@kernel.org, netfilter-devel@vger.kernel.org,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH bpf-next v2 3/4] bpf: Add support for writing to
 nf_conn:mark
Message-ID: <20220818193127.ykwoinzpoyxqwevg@kashmir.localdomain>
References: <cover.1660761470.git.dxu@dxuuu.xyz>
 <edbca42217a73161903a50ba07ec63c5fa5fde00.1660761470.git.dxu@dxuuu.xyz>
 <CAP01T75P5uM9EyX38bcoF4L2cbQ8orNVNhZsdoMXRThX5fd6JQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAP01T75P5uM9EyX38bcoF4L2cbQ8orNVNhZsdoMXRThX5fd6JQ@mail.gmail.com>
Precedence: bulk

On Wed, Aug 17, 2022 at 09:48:31PM +0200, Kumar Kartikeya Dwivedi wrote:
> On Wed, 17 Aug 2022 at 20:43, Daniel Xu <dxu@dxuuu.xyz> wrote:
[...]
> > diff --git a/include/net/netfilter/nf_conntrack_bpf.h b/include/net/netfilter/nf_conntrack_bpf.h
> > index a473b56842c5..0f584c2bd475 100644
> > --- a/include/net/netfilter/nf_conntrack_bpf.h
> > +++ b/include/net/netfilter/nf_conntrack_bpf.h
> > @@ -3,6 +3,7 @@
> >  #ifndef _NF_CONNTRACK_BPF_H
> >  #define _NF_CONNTRACK_BPF_H
> >
> > +#include <linux/bpf.h>
> >  #include <linux/btf.h>
> >  #include <linux/kconfig.h>
> >
> > @@ -10,6 +11,12 @@
> >      (IS_MODULE(CONFIG_NF_CONNTRACK) && IS_ENABLED(CONFIG_DEBUG_INFO_BTF_MODULES))
> >
> >  extern int register_nf_conntrack_bpf(void);
> > +extern int nf_conntrack_btf_struct_access(struct bpf_verifier_log *log,
> > +                                         const struct btf *btf,
> > +                                         const struct btf_type *t, int off,
> > +                                         int size, enum bpf_access_type atype,
> > +                                         u32 *next_btf_id,
> > +                                         enum bpf_type_flag *flag);
> >
> >  #else
> >
> > @@ -18,6 +25,17 @@ static inline int register_nf_conntrack_bpf(void)
> >         return 0;
> >  }
> >
> > +static inline int
> > +nf_conntrack_btf_struct_access(struct bpf_verifier_log *log,
> > +                              const struct btf *btf,
> > +                              const struct btf_type *t, int off,
> > +                              int size, enum bpf_access_type atype,
> > +                              u32 *next_btf_id,
> > +                              enum bpf_type_flag *flag)
> > +{
> > +       return -EACCES;
> > +}
> > +
> 
> We should make it work when nf_conntrack is a kernel module as well,
> not just when it is compiled in. The rest of the stuff already works
> when it is a module. For that, you can have a global function pointer
> for this callback, protected by a mutex. register/unregister sets
> it/unsets it. Each time you call it requires mutex to be held during
> the call.
> 
> Later when we have more modules that supply btf_struct_access callback
> for their module types we can generalize it, for now it should be ok
> to hardcode it for nf_conn.

Ok, will look into that.

> 
> >  #endif
> >
> >  #endif /* _NF_CONNTRACK_BPF_H */
[...]
> >
> > +/* Check writes into `struct nf_conn` */
> > +int nf_conntrack_btf_struct_access(struct bpf_verifier_log *log,
> > +                                  const struct btf *btf,
> > +                                  const struct btf_type *t, int off,
> > +                                  int size, enum bpf_access_type atype,
> > +                                  u32 *next_btf_id,
> > +                                  enum bpf_type_flag *flag)
> > +{
> > +       const struct btf_type *nct = READ_ONCE(nf_conn_type);
> > +       s32 type_id;
> > +       size_t end;
> > +
> > +       if (!nct) {
> > +               type_id = btf_find_by_name_kind(btf, "nf_conn", BTF_KIND_STRUCT);
> > +               if (type_id < 0)
> > +                       return -EINVAL;
> > +
> > +               nct = btf_type_by_id(btf, type_id);
> > +               WRITE_ONCE(nf_conn_type, nct);
> 
> Instead of this, why not just use BTF_ID_LIST_SINGLE to get the
> type_id and then match 't' to the result of btf_type_by_id?
> btf_type_by_id is not expensive.

Ah yeah, good idea. Will fix.

[...]

Thanks,
Daniel