Re: [PATCH net-next v4] net: Re-use and set mono_delivery_time bit for userspace tstamp packets

From: Daniel Borkmann <daniel@iogearbox.net>
To: Martin KaFai Lau <martin.lau@linux.dev>,
	"Abhishek Chauhan (ABC)" <quic_abchauha@quicinc.com>,
	Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Cc: kernel@quicinc.com, "David S. Miller" <davem@davemloft.net>,
	Eric Dumazet <edumazet@google.com>,
	Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	Andrew Halaney <ahalaney@redhat.com>,
	Martin KaFai Lau <martin.lau@kernel.org>,
	bpf <bpf@vger.kernel.org>, Alexei Starovoitov <ast@kernel.org>,
	Andrii Nakryiko <andrii@kernel.org>
Subject: Re: [PATCH net-next v4] net: Re-use and set mono_delivery_time bit for userspace tstamp packets
Date: Wed, 13 Mar 2024 22:41:44 +0100	[thread overview]
Message-ID: <1e419a00-41cf-c760-7912-68b34b74f8d1@iogearbox.net> (raw)
In-Reply-To: <0a0f948a-ca0e-4241-8a21-e7c7fc2ea471@linux.dev>

On 3/13/24 10:19 PM, Martin KaFai Lau wrote:
> On 3/13/24 1:59 PM, Abhishek Chauhan (ABC) wrote:
>>>> I think all this is crying for another bit in skb to mean user_delivery_time
>>>> (skb->tstamp depends on skb->sk->sk_clockid) while mono_delivery_time is the
>>>> mono time either set by kernel-tcp or bpf.
>>>
>>> It does sound like the approach with least side effects.
>>>
>> Martin and Willem , while we are discussing to use seperate bit per skb to check if its
>> SO_TXTIME from userspace or (rcv) timestamp . I came across two flags used in skbuff
>> called in filter framework
>> @tc_at_ingress: used within tc_classify to distinguish in/egress
>> @from_ingress: packet was redirected from the ingress path
>>
>> Since i believe the above testcase is based on redirecting from ingress to egress part
>> why cant we use these already existing flags ? Please advice
>>
>> I am not completely sure if we can use these flags to solve the bpf problem or not.
> 
> I don't see how they are related. It can tell where a skb is at. It cannot tell what is in skb->tstamp.

+1, if we go that route as per discussion, I'd also prefer clearly encoding delivery
time clock base in skb to avoid ambiguity.