Message-ID: <aa07f74e-ad9e-bc94-9ace-257db0aa5b4d@redhat.com>
Date:   Mon, 23 Jan 2023 10:25:44 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.6.0
Subject: Re: [PATCH 0/4] virtio_net: vdpa: update MAC address when it is
 generated by virtio-net
Content-Language: en-US
To:     "Michael S. Tsirkin" <mst@redhat.com>
Cc:     linux-kernel@vger.kernel.org, Parav Pandit <parav@nvidia.com>,
        virtualization@lists.linux-foundation.org, netdev@vger.kernel.org,
        Eli Cohen <elic@nvidia.com>, Jason Wang <jasowang@redhat.com>,
        Gautam Dawar <gautam.dawar@xilinx.com>,
        Cindy Lu <lulu@redhat.com>,
        "David S. Miller" <davem@davemloft.net>,
        =?UTF-8?Q?Eugenio_P=c3=a9rez?= <eperezma@redhat.com>
References: <20230122100526.2302556-1-lvivier@redhat.com>
 <20230122052211-mutt-send-email-mst@kernel.org>
From:   Laurent Vivier <lvivier@redhat.com>
In-Reply-To: <20230122052211-mutt-send-email-mst@kernel.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 1/22/23 11:23, Michael S. Tsirkin wrote:
> On Sun, Jan 22, 2023 at 11:05:22AM +0100, Laurent Vivier wrote:
>> When the MAC address is not provided by the vdpa device virtio_net
>> driver assigns a random one without notifying the device.
>> The consequence, in the case of mlx5_vdpa, is the internal routing
>> tables of the device are not updated and this can block the
>> communication between two namespaces.
>>
>> To fix this problem, use virtnet_send_command(VIRTIO_NET_CTRL_MAC)
>> to set the address from virtnet_probe() when the MAC address is
>> randomly assigned from virtio_net.
>>
>> While I was testing this change I found 3 other bugs in vdpa_sim_net:
>>
>> - vdpa_sim_net sets the VIRTIO_NET_F_MAC even if no MAC address is
>>    provided. So virtio_net doesn't generate a random MAC address and
>>    the MAC address appears to be 00:00:00:00:00:00
>>
>> - vdpa_sim_net never processes the command and virtnet_send_command()
>>    hangs in an infinite loop. To avoid a kernel crash add a timeout
>>    in the loop.
>>
>> - To allow vdpa_sim_net to process the command, replace the cpu_relax()
>>    in the loop by a schedule(). vdpa_sim_net uses a workqueue to process
>>    the queue, and if we don't allow the kernel to schedule, the queue
>>    is not processed and the loop is infinite.
> 
> I'd split these things out as opposed to a series unless there's
> a dependency I missed.

We needed to fix virtio_net before fixing vdpa_sim_net otherwise the 
virtnet_send_command() hangs when we define the vdpa device with "vdpa dev" but without a 
MAC address.

> All this reminds me of
> https://lore.kernel.org/r/20221226074908.8154-5-jasowang%40redhat.com
> 
> how is this patch different/better?
> Pls also CC people involved in that original discussion.

I was not aware of the Jason's series.

It seems to address better the problem, except it triggers the ASSERT_RTNL() in 
virtnet_send_command() when it is called from virtnet_probe().

I will remove patches 2 and 4 from my series.
PATCH 3 can be sent on independently too.

Thanks,
Laurent