On Thu, Jul 25, 2019 at 08:33:36AM -0700, Stanislav Fomichev wrote:
> C flow dissector supports input flags that tell it to customize parsing
> by either stopping early or trying to parse as deep as possible. Pass
> those flags to the BPF flow dissector so it can make the same
> decisions. In the next commits I'll add support for those flags to
> our reference bpf_flow.c
>
> Acked-by: Willem de Bruijn <willemb@xxxxxxxxxx>
> Acked-by: Song Liu <songliubraving@xxxxxx>
> Cc: Song Liu <songliubraving@xxxxxx>
> Cc: Willem de Bruijn <willemb@xxxxxxxxxx>
> Cc: Petar Penkov <ppenkov@xxxxxxxxxx>
> Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> ---
> include/linux/skbuff.h | 2 +-
> include/net/flow_dissector.h | 4 ----
> include/uapi/linux/bpf.h | 5 +++++
> net/bpf/test_run.c | 2 +-
> net/core/flow_dissector.c | 5 +++--
> 5 files changed, 10 insertions(+), 8 deletions(-)
>
> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 718742b1c505..9b7a8038beec 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -1271,7 +1271,7 @@ static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
>
> struct bpf_flow_dissector;
> bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
> - __be16 proto, int nhoff, int hlen);
> + __be16 proto, int nhoff, int hlen, unsigned int flags);
>
> bool __skb_flow_dissect(const struct net *net,
> const struct sk_buff *skb,
> diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
> index 90bd210be060..3e2642587b76 100644
> --- a/include/net/flow_dissector.h
> +++ b/include/net/flow_dissector.h
> @@ -253,10 +253,6 @@ enum flow_dissector_key_id {
> FLOW_DISSECTOR_KEY_MAX,
> };
>
> -#define FLOW_DISSECTOR_F_PARSE_1ST_FRAG BIT(0)
> -#define FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL BIT(1)
> -#define FLOW_DISSECTOR_F_STOP_AT_ENCAP BIT(2)
> -
> struct flow_dissector_key {
> enum flow_dissector_key_id key_id;
> size_t offset; /* offset of struct flow_dissector_key_*
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index fa1c753dcdbc..b4ad19bd6aa8 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -3507,6 +3507,10 @@ enum bpf_task_fd_type {
> BPF_FD_TYPE_URETPROBE, /* filename + offset */
> };
>
> +#define FLOW_DISSECTOR_F_PARSE_1ST_FRAG (1U << 0)
> +#define FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL (1U << 1)
> +#define FLOW_DISSECTOR_F_STOP_AT_ENCAP (1U << 2)
> +
I'm a bit concerned with direct move.
Last time we were in similar situation we've created:
enum {
BPF_TCP_ESTABLISHED = 1,
BPF_TCP_SYN_SENT,
and added:
BUILD_BUG_ON((int)BPF_TCP_ESTABLISHED != (int)TCP_ESTABLISHED);
BUILD_BUG_ON((int)BPF_TCP_SYN_SENT != (int)TCP_SYN_SENT);
It may be overkill here, but feels safer than direct move.
Adding BPF_ prefix also feels necessary to avoid very unlikely
(but still theoretically possible) conflicts.
