Re: [PATCH net-next v6 09/19] net: psp: update the TCP MSS to reflect PSP packet overhead

[Daniel Zahka <daniel.zahka@gmail.com>]

On 8/14/25 12:38 PM, Paolo Abeni wrote:
> On 8/14/25 4:50 PM, Daniel Zahka wrote:
>> On 8/14/25 9:58 AM, Paolo Abeni wrote:
>>> On 8/12/25 2:29 AM, Daniel Zahka wrote:
>>>> @@ -236,6 +237,10 @@ int psp_sock_assoc_set_tx(struct sock *sk, struct psp_dev *psd,
>>>>    	tcp_write_collapse_fence(sk);
>>>>    	pas->upgrade_seq = tcp_sk(sk)->rcv_nxt;
>>>>    
>>>> +	icsk = inet_csk(sk);
>>>> +	icsk->icsk_ext_hdr_len += psp_sk_overhead(sk);
>>> I'm likely lost, but AFAICS the user-space can successfully call
>>> multiple times psp_sock_assoc_set_tx() on the same socket, increasing
>>> icsk->icsk_ext_hdr_len in an unbounded way.
>> If it were possible to execute the code you have highlighted more than
>> once per socket, that would be a bug. This should not be possible
>> because of the preceding checks in the function i.e.
>>
>>       if (pas->tx.spi) {
>>           NL_SET_ERR_MSG(extack, "Tx key already set");
>>           err = -EBUSY;
>>           goto exit_unlock;
>>       }
> AFAICS the nl code ensures the SPI attribute must be present, but also
> allows a 0 value, so the above check could be eluded.

I think you are right that we need some extra validation for the SPI. 
The PSP spec states that 0 is a reserved value for SPI, and our 
implementation does indeed depend on that here.