netdev - Re: [PATCH bpf-next v4 08/15] veth: Support RX XDP metadata

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4bac619d-8767-1364-1924-78c05b1ecf88@redhat.com>
Date:   Wed, 14 Dec 2022 10:48:27 +0100
From:   Jesper Dangaard Brouer <jbrouer@...hat.com>
To:     Stanislav Fomichev <sdf@...gle.com>,
        Jesper Dangaard Brouer <jbrouer@...hat.com>
Cc:     brouer@...hat.com, bpf@...r.kernel.org, ast@...nel.org,
        daniel@...earbox.net, andrii@...nel.org, martin.lau@...ux.dev,
        song@...nel.org, yhs@...com, john.fastabend@...il.com,
        kpsingh@...nel.org, haoluo@...gle.com, jolsa@...nel.org,
        David Ahern <dsahern@...il.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Willem de Bruijn <willemb@...gle.com>,
        Anatoly Burakov <anatoly.burakov@...el.com>,
        Alexander Lobakin <alexandr.lobakin@...el.com>,
        Magnus Karlsson <magnus.karlsson@...il.com>,
        Maryam Tahhan <mtahhan@...hat.com>, xdp-hints@...-project.net,
        netdev@...r.kernel.org
Subject: Re: [PATCH bpf-next v4 08/15] veth: Support RX XDP metadata



On 13/12/2022 21.42, Stanislav Fomichev wrote:
> On Tue, Dec 13, 2022 at 7:55 AM Jesper Dangaard Brouer
> <jbrouer@...hat.com> wrote:
>>
>>
>> On 13/12/2022 03.35, Stanislav Fomichev wrote:
>>> The goal is to enable end-to-end testing of the metadata for AF_XDP.
>>>
>>> Cc: John Fastabend <john.fastabend@...il.com>
>>> Cc: David Ahern <dsahern@...il.com>
>>> Cc: Martin KaFai Lau <martin.lau@...ux.dev>
>>> Cc: Jakub Kicinski <kuba@...nel.org>
>>> Cc: Willem de Bruijn <willemb@...gle.com>
>>> Cc: Jesper Dangaard Brouer <brouer@...hat.com>
>>> Cc: Anatoly Burakov <anatoly.burakov@...el.com>
>>> Cc: Alexander Lobakin <alexandr.lobakin@...el.com>
>>> Cc: Magnus Karlsson <magnus.karlsson@...il.com>
>>> Cc: Maryam Tahhan <mtahhan@...hat.com>
>>> Cc: xdp-hints@...-project.net
>>> Cc: netdev@...r.kernel.org
>>> Signed-off-by: Stanislav Fomichev <sdf@...gle.com>
>>> ---
>>>    drivers/net/veth.c | 24 ++++++++++++++++++++++++
>>>    1 file changed, 24 insertions(+)
>>>
>>> diff --git a/drivers/net/veth.c b/drivers/net/veth.c
>>> index 04ffd8cb2945..d5491e7a2798 100644
>>> --- a/drivers/net/veth.c
>>> +++ b/drivers/net/veth.c
>>> @@ -118,6 +118,7 @@ static struct {
>>>
>>>    struct veth_xdp_buff {
>>>        struct xdp_buff xdp;
>>> +     struct sk_buff *skb;
>>>    };
>>>
>>>    static int veth_get_link_ksettings(struct net_device *dev,
>>> @@ -602,6 +603,7 @@ static struct xdp_frame *veth_xdp_rcv_one(struct veth_rq *rq,
>>>
>>>                xdp_convert_frame_to_buff(frame, xdp);
>>>                xdp->rxq = &rq->xdp_rxq;
>>> +             vxbuf.skb = NULL;
>>>
>>>                act = bpf_prog_run_xdp(xdp_prog, xdp);
>>>
>>> @@ -823,6 +825,7 @@ static struct sk_buff *veth_xdp_rcv_skb(struct veth_rq *rq,
>>>        __skb_push(skb, skb->data - skb_mac_header(skb));
>>>        if (veth_convert_skb_to_xdp_buff(rq, xdp, &skb))
>>>                goto drop;
>>> +     vxbuf.skb = skb;
>>>
>>>        orig_data = xdp->data;
>>>        orig_data_end = xdp->data_end;
>>> @@ -1601,6 +1604,21 @@ static int veth_xdp(struct net_device *dev, struct netdev_bpf *xdp)
>>>        }
>>>    }
>>>
>>> +static int veth_xdp_rx_timestamp(const struct xdp_md *ctx, u64 *timestamp)
>>> +{
>>> +     *timestamp = ktime_get_mono_fast_ns();
>>
>> This should be reading the hardware timestamp in the SKB.
>>
>> Details: This hardware timestamp in the SKB is located in
>> skb_shared_info area, which is also available for xdp_frame (currently
>> used for multi-buffer purposes).  Thus, when adding xdp-hints "store"
>> functionality, it would be natural to store the HW TS in the same place.
>> Making the veth skb/xdp_frame code paths able to share code.
> 
> Does something like the following look acceptable as well?
> 
> *timestamp = skb_hwtstamps(_ctx->skb)->hwtstamp;
> if (!*timestamp)
>          *timestamp = ktime_get_mono_fast_ns(); /* sw fallback */
> 

How can the BPF programmer tell the difference between getting hardware 
or software timestamp?

This will get really confusing when someone implements a tcpdump feature
(like/extend xdpdump) and some packets (e.g. PTP packets) have HW
timestamps and some don't.  The time sequence in the pcap will be strange.

> Because I'd like to be able to test this path in the selftests. As
> long as I get some number from veth_xdp_rx_timestamp, I can test it.
> No amount of SOF_TIMESTAMPING_{SOFTWARE,RX_SOFTWARE,RAW_HARDWARE}
> triggers non-zero hwtstamp for xsk receive path. Any suggestions?
> 

You could implement the "store" operation I mentioned before.
For testing you can store an arbitrary value in the timestamp and check
it later by reading it back.

I can see you have changed the API to send down a pointer. Thus, a
simple flag could implement the writing the provided timestamp.

Regarding flags for reading the timestamp.  Should we be able to specify
what clock type we are asking for?
Have you notice that tcpdump can ask for different types of
timestamps[1]. e.g. for hardware timestamps it is either
'adapter_unsynced' or 'adaptor'. (See semantic in [1])

  # tcpdump -ni eth1 -j adapter_unsynced --time-stamp-precision=nano

[1] https://www.tcpdump.org/manpages/pcap-tstamp.7.txt


Hint list which your NIC 'adaptor' supports via this cmdline:

  # tcpdump -i mlx5p2 --list-time-stamp-types

--Jesper