Re: [PATCH v2] bpf: Replace bpf_lpm_trie_key 0-length array with flexible array

[Daniel Borkmann]

On 2/17/24 4:03 AM, Kees Cook wrote:
> On Fri, Feb 16, 2024 at 06:27:08PM -0600, Gustavo A. R. Silva wrote:
> > On 2/16/24 17:55, Kees Cook wrote:
> > > Replace deprecated 0-length array in struct bpf_lpm_trie_key with
> > > flexible array. Found with GCC 13:
> > >
> > > ../kernel/bpf/lpm_trie.c:207:51: warning: array subscript i is outside array bounds of 'const __u8[0]' {aka 'const unsigned char[]'} [-Warray-bounds=]
> > >     207 |                                        *(__be16 *)&key->data[i]);
> > >         |                                                   ^~~~~~~~~~~~~
> > > ../include/uapi/linux/swab.h:102:54: note: in definition of macro '__swab16'
> > >     102 | #define __swab16(x) (__u16)__builtin_bswap16((__u16)(x))
> > >         |                                                      ^
> > > ../include/linux/byteorder/generic.h:97:21: note: in expansion of macro '__be16_to_cpu'
> > >      97 | #define be16_to_cpu __be16_to_cpu
> > >         |                     ^~~~~~~~~~~~~
> > > ../kernel/bpf/lpm_trie.c:206:28: note: in expansion of macro 'be16_to_cpu'
> > >     206 |                 u16 diff = be16_to_cpu(*(__be16 *)&node->data[i]
> > > ^
> > >         |                            ^~~~~~~~~~~
> > > In file included from ../include/linux/bpf.h:7:
> > > ../include/uapi/linux/bpf.h:82:17: note: while referencing 'data'
> > >      82 |         __u8    data[0];        /* Arbitrary size */
> > >         |                 ^~~~
> > >
> > > And found at run-time under CONFIG_FORTIFY_SOURCE:
> > >
> > >     UBSAN: array-index-out-of-bounds in kernel/bpf/lpm_trie.c:218:49
> > >     index 0 is out of range for type '__u8 [*]'
> > >
> > > This includes fixing the selftest which was incorrectly using a
> > > variable length struct as a header, identified earlier[1]. Avoid this
> > > by just explicitly including the prefixlen member instead of struct
> > > bpf_lpm_trie_key.
> > >
> > > Note that it is not possible to simply remove the "data" member, as it
> > > is referenced by userspace
> > >
> > > cilium:
> > >           struct egress_gw_policy_key in_key = {
> > >                   .lpm_key = { 32 + 24, {} },
> > >                   .saddr   = CLIENT_IP,
> > >                   .daddr   = EXTERNAL_SVC_IP & 0Xffffff,
> > >           };
> > >
> > > systemd:
> > > 	ipv6_map_fd = bpf_map_new(
> > > 			BPF_MAP_TYPE_LPM_TRIE,
> > > 			offsetof(struct bpf_lpm_trie_key, data) + sizeof(uint32_t)*4,
> > > 			sizeof(uint64_t),
> > > 			...
> > >
> > > The only risk to UAPI would be if sizeof() were used directly on the
> > > data member, which it does not seem to be. It is only used as a static
> > > initializer destination and to find its location via offsetof().
> > >
> > > Link: https://lore.kernel.org/all/202206281009.4332AA33@keescook/ [1]
> > > Reported-by: Mark Rutland <mark.rutland@xxxxxxx>
> > > Closes: https://paste.debian.net/hidden/ca500597/
> >
> > mmh... this URL expires: 2024-05-15
>
> Yup, but that's why I included the run-time splat above too. :)

I don't quite follow, this basically undoes 3024d95a4c52 ("bpf: Partially revert
flexible-array member replacement") again with the small change that this 'fixes'
up the BPF selftest to not embed struct bpf_lpm_trie_key.

Outside of BPF selftests though aren't we readding the same error that we fixed
earlier for BPF programs in the wild which embed struct bpf_lpm_trie_key into their
key structure?

Thanks,
Daniel