Date:   Tue, 7 Mar 2023 23:53:17 +0000
From:   Bobby Eshleman <bobbyeshleman@gmail.com>
To:     Stefano Garzarella <sgarzare@redhat.com>
Cc:     Arseniy Krasnov <avkrasnov@sberdevices.ru>,
        Bobby Eshleman <bobby.eshleman@bytedance.com>,
        kvm@vger.kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        virtualization@lists.linux-foundation.org, oxffffaa@gmail.com,
        Eric Dumazet <edumazet@google.com>,
        Stefan Hajnoczi <stefanha@redhat.com>, kernel@sberdevices.ru,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        "David S. Miller" <davem@davemloft.net>
Subject: Re: [RFC PATCH v2 2/4] virtio/vsock: remove all data from sk_buff
Message-ID: <ZAfObcosX5PlS4Lf@bullseye>
References: <a7ab414b-5e41-c7b6-250b-e8401f335859@sberdevices.ru>
 <dfadea17-a91e-105f-c213-a73f9731c8bd@sberdevices.ru>
 <20230306120857.6flftb3fftmsceyl@sgarzare-redhat>
 <b18e3b13-3386-e9ee-c817-59588e6d5fb6@sberdevices.ru>
 <20230306155121.7xwxzgxtle7qjbnc@sgarzare-redhat>
 <9b882d45-3d9d-c44d-a172-f23fff54962b@sberdevices.ru>
 <20230306161852.4s7qf4qm3fnwjck7@sgarzare-redhat>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230306161852.4s7qf4qm3fnwjck7@sgarzare-redhat>
Precedence: bulk

On Mon, Mar 06, 2023 at 05:18:52PM +0100, Stefano Garzarella wrote:
> On Mon, Mar 06, 2023 at 07:00:10PM +0300, Arseniy Krasnov wrote:
> > 
> > 
> > On 06.03.2023 18:51, Stefano Garzarella wrote:
> > > On Mon, Mar 06, 2023 at 06:31:22PM +0300, Arseniy Krasnov wrote:
> > > > 
> > > > 
> > > > On 06.03.2023 15:08, Stefano Garzarella wrote:
> > > > > On Sun, Mar 05, 2023 at 11:07:37PM +0300, Arseniy Krasnov wrote:
> > > > > > In case of SOCK_SEQPACKET all sk_buffs are used once - after read some
> > > > > > data from it, it will be removed, so user will never read rest of the
> > > > > > data. Thus we need to update credit parameters of the socket like whole
> > > > > > sk_buff is read - so call 'skb_pull()' for the whole buffer.
> > > > > > 
> > > > > > Fixes: 71dc9ec9ac7d ("virtio/vsock: replace virtio_vsock_pkt with sk_buff")
> > > > > > Signed-off-by: Arseniy Krasnov <AVKrasnov@sberdevices.ru>
> > > > > > ---
> > > > > > net/vmw_vsock/virtio_transport_common.c | 2 +-
> > > > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > 
> > > > > Maybe we could avoid this patch if we directly use pkt_len as I
> > > > > suggested in the previous patch.
> > > > Hm, may be we can avoid calling 'skb_pull()' here if 'virtio_transport_dec_rx_pkt()'
> > > > will use integer argument?
> > > 
> > > Yep, exactly!
> > > 
> > > > Just call 'virtio_transport_dec_rx_pkt(skb->len)'. skb
> > > 
> > > It depends on how we call virtio_transport_inc_rx_pkt(). If we use
> > > hdr->len there I would use the same to avoid confusion. Plus that's the
> > > value the other peer sent us, so definitely the right value to increase
> > > fwd_cnt with. But if skb->len always reflects it, then that's fine.
> > i've checked 'virtio_transport_rx_work()', it calls 'virtio_vsock_skb_rx_put()' which
> > sets 'skb->len'. Value is used from header, so seems 'skb->len' == 'hdr->len' in this
> > case.
> 
> Thank you for checking it.
> 
> However, I still think it is better to use `hdr->len` (we have to assign it
> to `pkt_len` anyway, as in the proposal I sent for patch 1), otherwise we
> have to go every time to check if skb_* functions touch skb->len.
> 
> E.g. skb_pull() decrease skb->len, so I'm not sure we can call
> virtio_transport_dec_rx_pkt(skb->len) if we don't remove `skb_pull(skb,
> bytes_to_copy);` inside the loop.
> 

I think it does make reasoning about the bytes accounting easier if it
is based off of the non-mutating hdr->len.

Especially if vsock does ever support tunneling (e.g., through
virtio-net) or some future feature that makes the skb->len more dynamic.

Best,
Bobby