Re: [PATCH bpf-next V3 1/6] bpf: Remove MTU check in __bpf_skb_max_len

From: "Maciej Żenczykowski" <maze@google.com>
To: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>,
	bpf <bpf@vger.kernel.org>, Linux NetDev <netdev@vger.kernel.org>,
	Daniel Borkmann <borkmann@iogearbox.net>,
	Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	Lorenz Bauer <lmb@cloudflare.com>,
	Shaun Crampton <shaun@tigera.io>,
	Lorenzo Bianconi <lorenzo@kernel.org>,
	Marek Majkowski <marek@cloudflare.com>,
	John Fastabend <john.fastabend@gmail.com>,
	Jakub Kicinski <kuba@kernel.org>,
	Eyal Birger <eyal.birger@gmail.com>,
	willemdebruijn.kernel@gmail.com
Subject: Re: [PATCH bpf-next V3 1/6] bpf: Remove MTU check in __bpf_skb_max_len
Date: Fri, 9 Oct 2020 11:26:40 -0700	[thread overview]
Message-ID: <CANP3RGd7dqog_v7qPd+FJ8iCqgPS=1trzQ778yeQ=9caGfxMGQ@mail.gmail.com> (raw)
In-Reply-To: <20b1e1dc-7ce7-dc42-54cd-5c4040ccdb30@iogearbox.net>

> > Multiple BPF-helpers that can manipulate/increase the size of the SKB uses
> > __bpf_skb_max_len() as the max-length. This function limit size against
> > the current net_device MTU (skb->dev->mtu).
> >
> > When a BPF-prog grow the packet size, then it should not be limited to the
> > MTU. The MTU is a transmit limitation, and software receiving this packet
> > should be allowed to increase the size. Further more, current MTU check in
> > __bpf_skb_max_len uses the MTU from ingress/current net_device, which in
> > case of redirects uses the wrong net_device.
> >
> > Keep a sanity max limit of IP6_MAX_MTU (under CONFIG_IPV6) which is 64KiB
> > plus 40 bytes IPv6 header size. If compiled without IPv6 use IP_MAX_MTU.
> >
> > V3: replace __bpf_skb_max_len() with define and use IPv6 max MTU size.
> >
> > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
> > ---
> >   net/core/filter.c |   16 ++++++++--------
> >   1 file changed, 8 insertions(+), 8 deletions(-)
> >
> > diff --git a/net/core/filter.c b/net/core/filter.c
> > index 05df73780dd3..ddc1f9ba89d1 100644
> > --- a/net/core/filter.c
> > +++ b/net/core/filter.c
> > @@ -3474,11 +3474,11 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff,
> >       return 0;
> >   }
> >
> > -static u32 __bpf_skb_max_len(const struct sk_buff *skb)
> > -{
> > -     return skb->dev ? skb->dev->mtu + skb->dev->hard_header_len :
> > -                       SKB_MAX_ALLOC;
> > -}
> > +#ifdef IP6_MAX_MTU /* Depend on CONFIG_IPV6 */
> > +#define BPF_SKB_MAX_LEN IP6_MAX_MTU
> > +#else
> > +#define BPF_SKB_MAX_LEN IP_MAX_MTU
> > +#endif
>
> Shouldn't that check on skb->protocol? The way I understand it is that a number of devices
> including virtual ones use ETH_MAX_MTU as their dev->max_mtu, so the mtu must be in the range
> of dev->min_mtu(=ETH_MIN_MTU), dev->max_mtu(=ETH_MAX_MTU). __dev_set_mtu() then sets the user
> value to dev->mtu in the core if within this range. That means in your case skb->dev->hard_header_len
> for example is left out, meaning if we go for some constant, that would need to be higher.

I think in the past skb->protocol was not guaranteed to be correct -
could be zero...
(with [misconfigured] raw sockets - maybe that's fixed now?)