Re: [PATCH] virtio_net: introduce TX timeout watchdog

From: Jason Wang
Date: Sun Sep 26 2021 - 00:36:31 EST


On Fri, Sep 17, 2021 at 4:45 PM tonylu_linux <tonylu@xxxxxxxxxxxxxxxxx> wrote:
>
> From: Tony Lu <tony.ly@xxxxxxxxxxxxxxxxx>
>
> This implements ndo_tx_timeout handler and put this into stats. When
> there is something wrong to send out packets, we could notice tx timeout
> events and total timeout counter.
>
> We have suffered send timeout issues due to the backends hung. With this,
> we can find the details, and collect the counters by monitor systems.
>
> Signed-off-by: Tony Lu <tony.ly@xxxxxxxxxxxxxxxxx>

Note that we support non tx interrupt mode (which could be turned on
via ethtool).

I wonder if this can work well in that case.

Or maybe it's the time to remove the non tx interrupt mode completely.
Want to do that?

Thanks

> ---
> drivers/net/virtio_net.c | 22 +++++++++++++++++++++-
> 1 file changed, 21 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 271d38c1d9f8..90fed0fdc40f 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -80,6 +80,7 @@ struct virtnet_sq_stats {
> u64 xdp_tx;
> u64 xdp_tx_drops;
> u64 kicks;
> + u64 tx_timeouts;
> };
>
> struct virtnet_rq_stats {
> @@ -103,6 +104,7 @@ static const struct virtnet_stat_desc virtnet_sq_stats_desc[] = {
> { "xdp_tx", VIRTNET_SQ_STAT(xdp_tx) },
> { "xdp_tx_drops", VIRTNET_SQ_STAT(xdp_tx_drops) },
> { "kicks", VIRTNET_SQ_STAT(kicks) },
> + { "tx_timeouts", VIRTNET_SQ_STAT(tx_timeouts) },
> };
>
> static const struct virtnet_stat_desc virtnet_rq_stats_desc[] = {
> @@ -1856,7 +1858,7 @@ static void virtnet_stats(struct net_device *dev,
> int i;
>
> for (i = 0; i < vi->max_queue_pairs; i++) {
> - u64 tpackets, tbytes, rpackets, rbytes, rdrops;
> + u64 tpackets, tbytes, terrors, rpackets, rbytes, rdrops;
> struct receive_queue *rq = &vi->rq[i];
> struct send_queue *sq = &vi->sq[i];
>
> @@ -1864,6 +1866,7 @@ static void virtnet_stats(struct net_device *dev,
> start = u64_stats_fetch_begin_irq(&sq->stats.syncp);
> tpackets = sq->stats.packets;
> tbytes = sq->stats.bytes;
> + terrors = sq->stats.tx_timeouts;
> } while (u64_stats_fetch_retry_irq(&sq->stats.syncp, start));
>
> do {
> @@ -1878,6 +1881,7 @@ static void virtnet_stats(struct net_device *dev,
> tot->rx_bytes += rbytes;
> tot->tx_bytes += tbytes;
> tot->rx_dropped += rdrops;
> + tot->tx_errors += terrors;
> }
>
> tot->tx_dropped = dev->stats.tx_dropped;
> @@ -2659,6 +2663,21 @@ static int virtnet_set_features(struct net_device *dev,
> return 0;
> }
>
> +static void virtnet_tx_timeout(struct net_device *dev, unsigned int txqueue)
> +{
> + struct virtnet_info *priv = netdev_priv(dev);
> + struct send_queue *sq = &priv->sq[txqueue];
> + struct netdev_queue *txq = netdev_get_tx_queue(dev, txqueue);
> +
> + u64_stats_update_begin(&sq->stats.syncp);
> + sq->stats.tx_timeouts++;
> + u64_stats_update_end(&sq->stats.syncp);
> +
> + netdev_err(dev, "TX timeout on queue: %u, sq: %s, vq: 0x%x, name: %s, %u usecs ago\n",
> + txqueue, sq->name, sq->vq->index, sq->vq->name,
> + jiffies_to_usecs(jiffies - txq->trans_start));
> +}
> +
> static const struct net_device_ops virtnet_netdev = {
> .ndo_open = virtnet_open,
> .ndo_stop = virtnet_close,
> @@ -2674,6 +2693,7 @@ static const struct net_device_ops virtnet_netdev = {
> .ndo_features_check = passthru_features_check,
> .ndo_get_phys_port_name = virtnet_get_phys_port_name,
> .ndo_set_features = virtnet_set_features,
> + .ndo_tx_timeout = virtnet_tx_timeout,
> };
>
> static void virtnet_config_changed_work(struct work_struct *work)
> --
> 2.19.1.6.gb485710b
>