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:   Fri, 4 Aug 2023 16:28:25 +0200
From:   Stefano Garzarella <sgarzare@redhat.com>
To:     Arseniy Krasnov <oxffffaa@gmail.com>
Cc:     Arseniy Krasnov <AVKrasnov@sberdevices.ru>,
        Stefan Hajnoczi <stefanha@redhat.com>,
        "David S. Miller" <davem@davemloft.net>,
        Eric Dumazet <edumazet@google.com>,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        "Michael S. Tsirkin" <mst@redhat.com>,
        Jason Wang <jasowang@redhat.com>,
        Bobby Eshleman <bobby.eshleman@bytedance.com>,
        kvm@vger.kernel.org, virtualization@lists.linux-foundation.org,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        kernel@sberdevices.ru
Subject: Re: [RFC PATCH v1 1/2] vsock: send SIGPIPE on write to shutdowned
 socket
Message-ID: <bzkwqp26joyzgvqyoypyv43wv7t3b6rzs3v5hkch45yggmrzp6@25byvzqwiztb>
References: <20230801141727.481156-1-AVKrasnov@sberdevices.ru>
 <20230801141727.481156-2-AVKrasnov@sberdevices.ru>
 <qgn26mgfotc7qxzp6ad7ezkdex6aqniv32c5tvehxh4hljsnvs@x7wvyvptizxx>
 <44fef482-579a-fed6-6e8c-d400546285fc@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <44fef482-579a-fed6-6e8c-d400546285fc@gmail.com>
Precedence: bulk

On Fri, Aug 04, 2023 at 03:46:47PM +0300, Arseniy Krasnov wrote:
>Hi Stefano,
>
>On 02.08.2023 10:46, Stefano Garzarella wrote:
>> On Tue, Aug 01, 2023 at 05:17:26PM +0300, Arseniy Krasnov wrote:
>>> POSIX requires to send SIGPIPE on write to SOCK_STREAM socket which was
>>> shutdowned with SHUT_WR flag or its peer was shutdowned with SHUT_RD
>>> flag. Also we must not send SIGPIPE if MSG_NOSIGNAL flag is set.
>>>
>>> Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru>
>>> ---
>>> net/vmw_vsock/af_vsock.c | 3 +++
>>> 1 file changed, 3 insertions(+)
>>>
>>> diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
>>> index 020cf17ab7e4..013b65241b65 100644
>>> --- a/net/vmw_vsock/af_vsock.c
>>> +++ b/net/vmw_vsock/af_vsock.c
>>> @@ -1921,6 +1921,9 @@ static int vsock_connectible_sendmsg(struct socket *sock, struct msghdr *msg,
>>> ??????????? err = total_written;
>>> ????}
>>> out:
>>> +??? if (sk->sk_type == SOCK_STREAM)
>>> +??????? err = sk_stream_error(sk, msg->msg_flags, err);
>>
>> Do you know why we don't need this for SOCK_SEQPACKET and SOCK_DGRAM?
>
>Yes, here is my explanation:
>
>This function checks that input error is SIGPIPE, and if so it sends SIGPIPE to the 'current' thread
>(except case when MSG_NOSIGNAL flag is set). This behaviour is described in POSIX:
>
>Page 367 (description of defines from sys/socket.h):
>MSG_NOSIGNAL: No SIGPIPE generated when an attempt to send is made on a stream-
>oriented socket that is no longer connected.
>
>Page 497 (description of SOCK_STREAM):
>A SIGPIPE signal is raised if a thread sends on a broken stream (one that is
>no longer connected).

Okay, but I think we should do also for SEQPACKET:

https://pubs.opengroup.org/onlinepubs/009696699/functions/xsh_chap02_10.html

In 2.10.6 Socket Types:

"The SOCK_SEQPACKET socket type is similar to the SOCK_STREAM type, and
is also connection-oriented. The only difference between these types is
that record boundaries ..."

Then in  2.10.14 Signals:

"The SIGPIPE signal shall be sent to a thread that attempts to send data
on a socket that is no longer able to send. In addition, the send
operation fails with the error [EPIPE]."

It's honestly not super clear, but I assume the problem is similar with
seqpacket since it's connection-oriented, or did I miss something?

For example in sctp_sendmsg() IIUC we raise a SIGPIPE regardless of
whether the socket is STREAM or SEQPACKET.

>
>Page 1802 (description of 'send()' call):
>MSG_NOSIGNAL
>
>Requests not to send the SIGPIPE signal if an attempt to
>send is made on a stream-oriented socket that is no
>longer connected. The [EPIPE] error shall still be
>returned
>
>And the same for 'sendto()' and 'sendmsg()'
>
>Link to the POSIX document:
>https://www.open-std.org/jtc1/sc22/open/n4217.pdf
>
>TCP (I think we must rely on it), KCM, SMC sockets (all of them are stream) work in the same
>way by calling this function. AF_UNIX also works in the same way, but it implements SIGPIPE handling
>without this function.

I'm okay calling this function.

>
>The only thing that confused me a little bit, that sockets above returns EPIPE when
>we have only SEND_SHUTDOWN set, but for AF_VSOCK EPIPE is returned for RCV_SHUTDOWN
>also, but I think it is related to this patchset.

Do you mean that it is NOT related to this patchset?

Thanks,
Stefano