netdev - Re: [PATCH bpf-next v7 2/8] net: Update an existing TCP congestion control algorithm.

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: <b8b54ef5-8a24-5886-8f4e-8856dbaa9c34@iogearbox.net>
Date:   Fri, 17 Mar 2023 18:23:11 +0100
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Martin KaFai Lau <martin.lau@...ux.dev>
Cc:     netdev@...r.kernel.org, Eric Dumazet <edumazet@...gle.com>,
        Kui-Feng Lee <kuifeng@...a.com>, bpf@...r.kernel.org,
        ast@...nel.org, song@...nel.org, kernel-team@...a.com,
        andrii@...nel.org, sdf@...gle.com
Subject: Re: [PATCH bpf-next v7 2/8] net: Update an existing TCP congestion
 control algorithm.

On 3/17/23 6:18 PM, Martin KaFai Lau wrote:
> On 3/17/23 8:23 AM, Daniel Borkmann wrote:
>>  From the function itself what is not clear whether
>> callers that replace an existing one should do the synchronize_rcu() themselves or if this should
>> be part of tcp_update_congestion_control?
> 
> bpf_struct_ops_map_free (in patch 1) also does synchronize_rcu() for another reason (bpf_setsockopt), so the caller (bpf_struct_ops) is doing it. From looking at tcp_unregister_congestion_control(), make sense that it is more correct to have another synchronize_rcu() also in tcp_update_congestion_control in case there will be other non bpf_struct_ops caller doing update in the future.

Agree, I was looking at 'bpf: Update the struct_ops of a bpf_link', and essentially as-is
it was implicit via map free. +1, tcp_update_congestion_control() would be more obvious and
better for other/non-BPF users.

Thanks,
Daniel