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;
MIME-Version: 1.0
References: <20221222060427.21626-1-jasowang@redhat.com> <20221222060427.21626-5-jasowang@redhat.com>
 <CAJs=3_D6sug80Bb9tnAw5T0_NaL_b=u8ZMcwZtd-dy+AH_yqzQ@mail.gmail.com>
 <CACGkMEv4YxuqrSx_HW2uWgXXSMOFCzTJCCD_EVhMwegsL8SoCg@mail.gmail.com>
 <CAJs=3_Akv1zoKy_HARjnqMdNsy_n34TzzGA6a25xrkF2rCnqwg@mail.gmail.com> <CACGkMEvtgr=pDpcZeE4+ssh+PiL0k2B2+3kzdDmEvxxe=2mtGA@mail.gmail.com>
In-Reply-To: <CACGkMEvtgr=pDpcZeE4+ssh+PiL0k2B2+3kzdDmEvxxe=2mtGA@mail.gmail.com>
From:   Alvaro Karsz <alvaro.karsz@solid-run.com>
Date:   Fri, 23 Dec 2022 09:38:37 +0200
Message-ID: <CAJs=3_BqDqEoXGiU9zCNfGNqEjd1eijqkE_8_mb3nC=GwQoxHA@mail.gmail.com>
Subject: Re: [RFC PATCH 4/4] virtio-net: sleep instead of busy waiting for cvq command
To:     Jason Wang <jasowang@redhat.com>
Cc:     mst@redhat.com, davem@davemloft.net, edumazet@google.com,
        kuba@kernel.org, pabeni@redhat.com,
        virtualization@lists.linux-foundation.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, maxime.coquelin@redhat.com,
        eperezma@redhat.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

> This needs to be proposed to the virtio spec first. And actually we
> need more than this:
>
> 1) we still need a way to deal with the device without this feature
> 2) driver can't depend solely on what is advertised by the device (e.g
> device can choose to advertise a very long timeout)

I think that I wasn't clear, sorry.
I'm not talking about a new virtio feature, I'm talking about a situation when:
* virtio_net issues a control command.
* the device gets the command, but somehow, completes the command
after timeout.
* virtio_net assumes that the command failed (timeout), and issues a
different control command.
* virtio_net will then call virtqueue_wait_for_used, and will
immediately get the previous response (If I'm not wrong).

So, this is not a new feature that I'm proposing, just a situation
that may occur due to cvq timeouts.

Anyhow, your solution calling BAD_RING if we reach a timeout should
prevent this situation.

Thanks