netdev - perhaps inet_csk_reqsk_queue_is_full should also allow zero backlog

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-next>] [day] [month] [year] [list]

Message-ID: <CAFmV8Nc758FDNK3FNSLQui4RmE3-TQr7d2tM_tOM6bC=OfEDwQ@mail.gmail.com>
Date: Tue, 31 Dec 2024 16:24:31 +0800
From: Zhongqiu Duan <dzq.aishenghu0@...il.com>
To: netdev@...r.kernel.org
Cc: davem@...emloft.net, kuba@...nel.org, pabeni@...hat.com, 
	edumazet@...gle.com
Subject: perhaps inet_csk_reqsk_queue_is_full should also allow zero backlog

Hi all,

We use a proprietary library in our product, it passes hardcoded zero
as the backlog of listen().
It works fine when syncookies is enabled, but when we disable syncookies
by business requirement, no connection can be made.

After some investigation, the problem is focused on the
inet_csk_reqsk_queue_is_full().

static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
{
        return inet_csk_reqsk_queue_len(sk) >=
READ_ONCE(sk->sk_max_ack_backlog);
}

I noticed that the stories happened to sk_acceptq_is_full() about this
in the past, like
the commit c609e6a (Revert "net: correct sk_acceptq_is_full()").

Perhaps we can also avoid the problem by using ">" in the decision
condition like
`inet_csk_reqsk_queue_len(sk) > READ_ONCE(sk->sk_max_ack_backlog)`.

Best regards,
Zhongqiu