[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87czqbq6ic.fsf@toke.dk>
Date: Wed, 18 Aug 2021 14:31:07 +0200
From: Toke Høiland-Jørgensen <toke@...hat.com>
To: Lorenzo Bianconi <lorenzo@...nel.org>, bpf@...r.kernel.org,
netdev@...r.kernel.org
Cc: lorenzo.bianconi@...hat.com, davem@...emloft.net, kuba@...nel.org,
ast@...nel.org, daniel@...earbox.net, shayagr@...zon.com,
john.fastabend@...il.com, dsahern@...nel.org, brouer@...hat.com,
echaudro@...hat.com, jasowang@...hat.com,
alexander.duyck@...il.com, saeed@...nel.org,
maciej.fijalkowski@...el.com, magnus.karlsson@...el.com,
tirthendu.sarkar@...el.com
Subject: Re: [PATCH v11 bpf-next 17/18] net: xdp: introduce
bpf_xdp_adjust_data helper
Lorenzo Bianconi <lorenzo@...nel.org> writes:
> For XDP frames split over multiple buffers, the xdp_md->data and
> xdp_md->data_end pointers will point to the start and end of the first
> fragment only. bpf_xdp_adjust_data can be used to access subsequent
> fragments by moving the data pointers. To use, an XDP program can call
> this helper with the byte offset of the packet payload that
> it wants to access; the helper will move xdp_md->data and xdp_md ->data_end
> so they point to the requested payload offset and to the end of the
> fragment containing this byte offset, and return the byte offset of the
> start of the fragment.
> To move back to the beginning of the packet, simply call the
> helper with an offset of '0'.
> Note also that the helpers that modify the packet boundaries
> (bpf_xdp_adjust_head(), bpf_xdp_adjust_tail() and
> bpf_xdp_adjust_meta()) will fail if the pointers have been
> moved; it is the responsibility of the BPF program to move them
> back before using these helpers.
>
> Suggested-by: John Fastabend <john.fastabend@...il.com>
> Signed-off-by: Lorenzo Bianconi <lorenzo@...nel.org>
> ---
> include/net/xdp.h | 8 +++++
> include/uapi/linux/bpf.h | 32 ++++++++++++++++++
> net/bpf/test_run.c | 8 +++++
> net/core/filter.c | 61 +++++++++++++++++++++++++++++++++-
> tools/include/uapi/linux/bpf.h | 32 ++++++++++++++++++
> 5 files changed, 140 insertions(+), 1 deletion(-)
>
> diff --git a/include/net/xdp.h b/include/net/xdp.h
> index cdaecf8d4d61..ce4764c7cd40 100644
> --- a/include/net/xdp.h
> +++ b/include/net/xdp.h
> @@ -82,6 +82,11 @@ struct xdp_buff {
> struct xdp_txq_info *txq;
> u32 frame_sz; /* frame size to deduce data_hard_end/reserved tailroom*/
> u16 flags; /* supported values defined in xdp_flags */
> + /* xdp multi-buff metadata used for frags iteration */
> + struct {
> + u16 headroom; /* frame headroom: data - data_hard_start */
> + u16 headlen; /* first buffer length: data_end - data */
> + } mb;
> };
>
> static __always_inline bool xdp_buff_is_mb(struct xdp_buff *xdp)
> @@ -127,6 +132,9 @@ xdp_prepare_buff(struct xdp_buff *xdp, unsigned char *hard_start,
> xdp->data = data;
> xdp->data_end = data + data_len;
> xdp->data_meta = meta_valid ? data : data + 1;
> + /* mb metadata for frags iteration */
> + xdp->mb.headroom = headroom;
> + xdp->mb.headlen = data_len;
> }
>
> /* Reserve memory area at end-of data area.
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index ddbf9ccc2f74..c20a8b7c5c7c 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -4853,6 +4853,37 @@ union bpf_attr {
> * Get the total size of a given xdp buff (linear and paged area)
> * Return
> * The total size of a given xdp buffer.
> + *
> + * long bpf_xdp_adjust_data(struct xdp_buff *xdp_md, u32 offset)
> + * Description
> + * For XDP frames split over multiple buffers, the
> + * *xdp_md*\ **->data** and*xdp_md *\ **->data_end** pointers
> + * will point to the start and end of the first fragment only.
> + * This helper can be used to access subsequent fragments by
> + * moving the data pointers. To use, an XDP program can call
> + * this helper with the byte offset of the packet payload that
> + * it wants to access; the helper will move *xdp_md*\ **->data**
> + * and *xdp_md *\ **->data_end** so they point to the requested
> + * payload offset and to the end of the fragment containing this
> + * byte offset, and return the byte offset of the start of the
> + * fragment.
This comment is wrong now :)
-Toke
Powered by blists - more mailing lists