Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 10 May 2023 11:39:54 -0300
From:   Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
To:     Alexander Mikhalitsyn <aleksandr.mikhalitsyn@canonical.com>
Cc:     nhorman@tuxdriver.com, davem@davemloft.net,
        Daniel Borkmann <daniel@iogearbox.net>,
        Christian Brauner <brauner@kernel.org>,
        Stanislav Fomichev <sdf@google.com>,
        Xin Long <lucien.xin@gmail.com>, linux-sctp@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org
Subject: Re: [PATCH net-next] sctp: add bpf_bypass_getsockopt proto callback
Message-ID: <ZFusunmfAaQVmBE2@t14s.localdomain>
References: <20230510131527.1244929-1-aleksandr.mikhalitsyn@canonical.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230510131527.1244929-1-aleksandr.mikhalitsyn@canonical.com>
Precedence: bulk

On Wed, May 10, 2023 at 03:15:27PM +0200, Alexander Mikhalitsyn wrote:
> Add bpf_bypass_getsockopt proto callback and filter out
> SCTP_SOCKOPT_PEELOFF and SCTP_SOCKOPT_PEELOFF_FLAGS socket options
> from running eBPF hook on them.
> 
> These options do fd_install(), and if BPF_CGROUP_RUN_PROG_GETSOCKOPT
> hook returns an error after success of the original handler
> sctp_getsockopt(...), userspace will receive an error from getsockopt
> syscall and will be not aware that fd was successfully installed into fdtable.
> 
> This patch was born as a result of discussion around a new SCM_PIDFD interface:
> https://lore.kernel.org/all/20230413133355.350571-3-aleksandr.mikhalitsyn@canonical.com/

I read some of the emails in there but I don't get why the fd leak is
special here. I mean, I get that it leaks, but masking the error
return like this can lead to several other problems in the application
as well.

For example, SCTP_SOCKOPT_CONNECTX3 will trigger a connect(). If it
failed, and the hook returns success, the user app will at least log a
wrong "connection successful".

If the hook can't be responsible for cleaning up before returning a
different value, then maybe we want to extend the list of sockopts in
here. AFAICT these would be the 3 most critical sockopts.