Re: [PATCH 1/1] add selftest for virtio-net

[Jason Wang]

On 08/27/2014 09:45 AM, Hengjinxiao wrote:
> Selftest is an important part of network driver, this patch adds selftest for
> virtio-net, including loopback test, negotiate test and reset test. Loopback 
> test checks whether virtio-net can send and receive packets normally. Negotiate test
> executes feature negotiation between virtio-net driver in Guest OS and virtio-net 
> device in Host OS. Reset test resets virtio-net.

Thanks for the patch. Feature negotiation part brings some complicity
and need more through. And this could be extended for CVE regression in
the future. And you probably also need to send a patch of virtio spec to
implement the loop back mode.

See comments inline.
>
> Signed-off-by: Hengjinxiao <hjxiaohust@xxxxxxxxx>
>
> ---
>  drivers/net/virtio_net.c        | 233 +++++++++++++++++++++++++++++++++++++++-
>  include/uapi/linux/virtio_net.h |   9 ++
>  2 files changed, 241 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 59caa06..f83f6e4 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -28,6 +28,7 @@
>  #include <linux/cpu.h>
>  #include <linux/average.h>
>  #include <net/busy_poll.h>
> +#include <linux/pci.h>
>  
>  static int napi_weight = NAPI_POLL_WEIGHT;
>  module_param(napi_weight, int, 0444);
> @@ -51,6 +52,23 @@ module_param(gso, bool, 0444);
>  #define MERGEABLE_BUFFER_ALIGN max(L1_CACHE_BYTES, 256)
>  
>  #define VIRTNET_DRIVER_VERSION "1.0.0"
> +#define __VIRTNET_TESTING  0
> +

Why need this marco?
> +enum {
> +	VIRTNET_LOOPBACK_TEST,
> +	VIRTNET_FEATURE_NEG_TEST,
> +	VIRTNET_RESET_TEST,
> +};
> +
> +static const struct {
> +	const char string[ETH_GSTRING_LEN];
> +} virtnet_gstrings_test[] = {
> +	[VIRTNET_LOOPBACK_TEST]	= { "loopback test   (offline)" },
> +	[VIRTNET_FEATURE_NEG_TEST]	= { "negotiate test  (offline)" },
> +	[VIRTNET_RESET_TEST]	= { "reset test     (offline)" },
> +};
> +
> +#define VIRTNET_NUM_TEST	ARRAY_SIZE(virtnet_gstrings_test)
>  
>  struct virtnet_stats {
>  	struct u64_stats_sync tx_syncp;
> @@ -104,6 +122,8 @@ struct virtnet_info {
>  	struct send_queue *sq;
>  	struct receive_queue *rq;
>  	unsigned int status;
> +	unsigned long flags;
> +	atomic_t lb_count;
>  
>  	/* Max # of queue pairs supported by the device */
>  	u16 max_queue_pairs;
> @@ -436,6 +456,19 @@ err_buf:
>  	return NULL;
>  }
>  
> +void virtnet_check_lb_frame(struct virtnet_info *vi,
> +				   struct sk_buff *skb)
> +{
> +	unsigned int frame_size = skb->len;
> +
> +	if (*(skb->data + 3) == 0xFF) {
> +		if ((*(skb->data + frame_size / 2 + 10) == 0xBE) &&
> +		   (*(skb->data + frame_size / 2 + 12) == 0xAF)) {
> +			atomic_dec(&vi->lb_count);
> +		}
> +	}
> +}
> +
>  static void receive_buf(struct receive_queue *rq, void *buf, unsigned int len)
>  {
>  	struct virtnet_info *vi = rq->vq->vdev->priv;
> @@ -485,7 +518,12 @@ static void receive_buf(struct receive_queue *rq, void *buf, unsigned int len)
>  	} else if (hdr->hdr.flags & VIRTIO_NET_HDR_F_DATA_VALID) {
>  		skb->ip_summed = CHECKSUM_UNNECESSARY;
>  	}
> -
> +	/* loopback self test for ethtool */
> +	if (test_bit(__VIRTNET_TESTING, &vi->flags)) {
> +		virtnet_check_lb_frame(vi, skb);
> +		dev_kfree_skb_any(skb);
> +		return;
> +	}

Not sure it's a good choice for adding such in fast path. We may need a
test specific rx interrupt handler (and disable NAPI) for this.
>  	skb->protocol = eth_type_trans(skb, dev);
>  	pr_debug("Receiving skb proto 0x%04x len %i type %i\n",
>  		 ntohs(skb->protocol), skb->len, skb->pkt_type);
> @@ -813,6 +851,9 @@ static int virtnet_open(struct net_device *dev)
>  {
>  	struct virtnet_info *vi = netdev_priv(dev);
>  	int i;
> +	/* disallow open during test */
> +	if (test_bit(__VIRTNET_TESTING, &vi->flags))
> +		return -EBUSY;
>  
>  	for (i = 0; i < vi->max_queue_pairs; i++) {
>  		if (i < vi->curr_queue_pairs)
> @@ -1363,12 +1404,158 @@ static void virtnet_get_channels(struct net_device *dev,
>  	channels->other_count = 0;
>  }
>  
> +static int virtnet_reset(struct virtnet_info *vi);
> +
> +static void virtnet_create_lb_frame(struct sk_buff *skb,
> +					unsigned int frame_size)
> +{
> +	memset(skb->data, 0xFF, frame_size);
> +	frame_size &= ~1;
> +	memset(&skb->data[frame_size / 2], 0xAA, frame_size / 2 - 1);
> +	memset(&skb->data[frame_size / 2 + 10], 0xBE, 1);
> +	memset(&skb->data[frame_size / 2 + 12], 0xAF, 1);
> +}
> +
> +static int virtnet_start_loopback(struct virtnet_info *vi)
> +{
> +	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_LOOPBACK,
> +				  VIRTIO_NET_CTRL_LOOPBACK_SET, NULL, NULL)) {
> +		dev_warn(&vi->dev->dev, "Failed to set loopback.\n");
> +		return -EINVAL;
> +	}
> +	return 0;

You may need to test the feature bit of loop back first and report the
card does not support loop back in some way.
> +}
> +
> +static int virtnet_run_loopback_test(struct virtnet_info *vi)
> +{
> +	int i;
> +	netdev_tx_t rc;
> +	struct sk_buff *skb;
> +	unsigned int size = GOOD_COPY_LEN;
> +
> +	for (i = 0; i < 100; i++) {
> +		skb = netdev_alloc_skb(vi->dev, size);
> +		if (!skb)
> +			return -ENOMEM;
> +
> +		skb->queue_mapping = 0;
> +		skb_put(skb, size);
> +		virtnet_create_lb_frame(skb, size);
> +		rc = start_xmit(skb, vi->dev);

virtio_net does not use tx interrupt to free old xmit skbs. It poll tx
completion only during xmit_skb(). So at least the last skb is leaked
since it was not freed. A possible solution is using tx interrupt here.
> +		if (rc != NETDEV_TX_OK)
> +			return -EPIPE;

It looks to me that start_xmit() never return other value than NETDEV_TX_OK.
> +		atomic_inc(&vi->lb_count);
> +	}
> +	/* Give queue time to settle before testing results. */
> +	msleep(20);

Need to make sure this value is also ok for qemu. ixgbe use 200 to 64
packets.
> +	return atomic_read(&vi->lb_count) ? -EIO : 0;
> +}
> +
> +static int virtnet_stop_loopback(struct virtnet_info *vi)
> +{
> +	if (!virtnet_send_command(vi, VIRTIO_NET_CTRL_LOOPBACK,
> +				  VIRTIO_NET_CTRL_LOOPBACK_UNSET, NULL, NULL)) {
> +		dev_warn(&vi->dev->dev, "Failed to unset loopback.\n");
> +		return -EINVAL;
> +	}
> +	return 0;
> +}
> +
> +static int virtnet_loopback_test(struct virtnet_info *vi, u64 *data)
> +{
> +	*data = virtnet_start_loopback(vi);
> +	if (*data)
> +		goto out;
> +	*data = virtnet_run_loopback_test(vi);
> +	if (*data)

Do we need to stop loopback here?
> +		goto out;
> +	*data = virtnet_stop_loopback(vi);
> +out:
> +	return *data;
> +}
> +
> +static void virtnet_feature_neg_test(struct virtnet_info *vi)
> +{
> +	struct virtio_device *dev = vi->vdev;
> +	struct virtio_driver *drv = drv_to_virtio(dev->dev.driver);
> +	int i;
> +	u32 device_features;
> +
> +	/* Figure out what features the device supports. */
> +	device_features = dev->config->get_features(dev);
> +
> +	/* Features supported by both device and driver into dev->features. */
> +	memset(dev->features, 0, sizeof(dev->features));
> +	for (i = 0; i < drv->feature_table_size; i++) {
> +		unsigned int f = drv->feature_table[i];
> +
> +		BUG_ON(f >= 32);
> +		if (device_features & (1 << f))
> +			set_bit(f, dev->features);
> +	}
> +
> +	/* Transport features always preserved to pass to finalize_features. */
> +	for (i = VIRTIO_TRANSPORT_F_START; i < VIRTIO_TRANSPORT_F_END; i++)
> +		if (device_features & (1 << i))
> +			set_bit(i, dev->features);
> +
> +	dev->config->finalize_features(dev);
> +}

A problem of the function is it may be called during DRIVER_OK, not sure
this is ok since spec suggest to do the feature negotiation after DRIVER
bit but before DRIVER_OK bit. And this function duplicates some of the
code from virtio core, may consider a method to share between them.

Another issue is the test never fail which needs more thought.
> +
> +static int virtnet_get_sset_count(struct net_device *netdev, int sset)
> +{
> +	switch (sset) {
> +	case ETH_SS_TEST:
> +		return VIRTNET_NUM_TEST;
> +	default:
> +		return -EOPNOTSUPP;
> +	}
> +}
> +
> +static void virtnet_get_strings(struct net_device *dev, u32 stringset, u8 *buf)
> +{
> +	switch (stringset) {
> +	case ETH_SS_TEST:
> +		memcpy(buf, &virtnet_gstrings_test,
> +			sizeof(virtnet_gstrings_test));
> +		break;
> +	default:
> +		break;
> +	}
> +}
> +
> +static void virtnet_self_test(struct net_device *netdev,
> +			    struct ethtool_test *eth_test, u64 *data)
> +{
> +	struct virtnet_info *vi = netdev_priv(netdev);
> +	bool if_running = netif_running(netdev);
> +
> +	set_bit(__VIRTNET_TESTING, &vi->flags);
> +	memset(data, 0, sizeof(u64) * VIRTNET_NUM_TEST);
> +
> +	if (eth_test->flags == ETH_TEST_FL_OFFLINE) {
> +		if (!if_running) {
> +			dev_warn(&vi->dev->dev, "Failed to execute self test.\n");
> +			eth_test->flags |= ETH_TEST_FL_FAILED;
> +			return;
> +		}
> +		if (virtnet_loopback_test(vi, &data[VIRTNET_LOOPBACK_TEST]))
> +			eth_test->flags |= ETH_TEST_FL_FAILED;
> +		virtnet_feature_neg_test(vi);
> +		virtnet_reset(vi);
> +	}
> +	clear_bit(__VIRTNET_TESTING, &vi->flags);
> +}
> +
>  static const struct ethtool_ops virtnet_ethtool_ops = {
>  	.get_drvinfo = virtnet_get_drvinfo,
>  	.get_link = ethtool_op_get_link,
>  	.get_ringparam = virtnet_get_ringparam,
>  	.set_channels = virtnet_set_channels,
>  	.get_channels = virtnet_get_channels,
> +	.self_test = virtnet_self_test,
> +	.get_strings		= virtnet_get_strings,
> +	.get_sset_count = virtnet_get_sset_count,
>  };
>  
>  #define MIN_MTU 68
> @@ -1957,6 +2144,50 @@ static int virtnet_restore(struct virtio_device *vdev)
>  }
>  #endif
>  
> +static int virtnet_reset(struct virtnet_info *vi)

If this is needed, better split this into anther patch.
> +{
> +	struct virtio_device *vdev = vi->vdev;
> +	int err, i;
> +	u8 status;
> +
> +	mutex_lock(&vi->config_lock);
> +	vi->config_enable = false;
> +	mutex_unlock(&vi->config_lock);
> +
> +	cancel_delayed_work_sync(&vi->refill);
> +
> +	if (netif_running(vi->dev))
> +		for (i = 0; i < vi->max_queue_pairs; i++) {
> +			napi_disable(&vi->rq[i].napi);
> +			netif_napi_del(&vi->rq[i].napi);
> +		}
> +
> +	remove_vq_common(vi);
> +	flush_work(&vi->config_work);
> +
> +	virtnet_feature_neg_test(vi);

Is this used for feature negotiation? If yes, need a better name and can
we reuse virtio core function to do this?
> +	err = init_vqs(vi);
> +	if (err)
> +		return err;
> +	if (netif_running(vi->dev)) {
> +		for (i = 0; i < vi->curr_queue_pairs; i++)
> +			if (!try_fill_recv(&vi->rq[i], GFP_KERNEL))
> +				schedule_delayed_work(&vi->refill, 0);
> +
> +		for (i = 0; i < vi->max_queue_pairs; i++)
> +			virtnet_napi_enable(&vi->rq[i]);
> +	}
> +
> +	mutex_lock(&vi->config_lock);
> +	vi->config_enable = true;
> +	mutex_unlock(&vi->config_lock);
> +
> +	virtnet_set_queues(vi, vi->curr_queue_pairs);
> +	status = vdev->config->get_status(vdev);
> +	vdev->config->set_status(vdev, status | VIRTIO_CONFIG_S_DRIVER_OK);
> +	return 0;
> +}
> +
>  static struct virtio_device_id id_table[] = {
>  	{ VIRTIO_ID_NET, VIRTIO_DEV_ANY_ID },
>  	{ 0 },
> diff --git a/include/uapi/linux/virtio_net.h b/include/uapi/linux/virtio_net.h
> index 172a7f0..1f31f90 100644
> --- a/include/uapi/linux/virtio_net.h
> +++ b/include/uapi/linux/virtio_net.h
> @@ -201,4 +201,13 @@ struct virtio_net_ctrl_mq {
>   #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN        1
>   #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX        0x8000
>  
> + /*
> +  * Control Loopback(5 is used by VIRTIO_NET_CTRL_GUEST_OFFLOADS in latest qemu)
> +  *
> +  * The command VIRTIO_NET_CTRL_LOOPBACK_SET is used to require the device come
> +  * into loopback state.
> +  */
> +#define VIRTIO_NET_CTRL_LOOPBACK   6
> + #define VIRTIO_NET_CTRL_LOOPBACK_SET        0
> + #define VIRTIO_NET_CTRL_LOOPBACK_UNSET        1
>  #endif /* _LINUX_VIRTIO_NET_H */

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/