[dpdk-dev] [PATCH v1] add mtu set in virtio

[dpdk-dev] [PATCH v1] add mtu set in virtio

[souvikdey33]

This functionality is required mostly in the cloud infrastructure.
For example, if we use gre or vxlan network between compute and
controller, then we should not use 1500 mtu in the guest as with
encapsulation the sixe of the packet will be more and will get 
dropped in the infrastructure. So, in that case we should honor
the mtu size sent by the dhcp server and configure the same on 
the virtual interfaces in the guest. This will also keep a 
consistent mtu through out the infrastructure.

souvikdey33 (1):
  Signed-off-by: Souvik Dey <sodey@sonusnet.com>

 drivers/net/virtio/virtio_ethdev.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

-- 
>From 2e4b391fe90ba5e617611e341a7d260dd3dd9144 Mon Sep 17 00:00:00 2001
From: souvikdey33 <sodey@sonusnet.com>
Date: Fri, 26 Aug 2016 20:46:21 -0400
Subject: [PATCH v1] Signed-off-by: Souvik Dey <sodey@sonusnet.com>

Fixes: 1fb8e8896ca8 ("Signed-off-by: Souvik Dey <sodey@sonusnet.com>")

Virtio interfaces should also support setting of mtu, as in case of cloud
it is expected to have the consistent mtu across the infrastructure that
the dhcp server sends and not hardcoded to 1500(default).
---
 drivers/net/virtio/virtio_ethdev.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
index 07d6449..da16ad4 100644
--- a/drivers/net/virtio/virtio_ethdev.c
+++ b/drivers/net/virtio/virtio_ethdev.c
@@ -92,6 +92,7 @@ static void virtio_mac_addr_add(struct rte_eth_dev *dev,
 static void virtio_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index);
 static void virtio_mac_addr_set(struct rte_eth_dev *dev,
 				struct ether_addr *mac_addr);
+static int  virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int virtio_dev_queue_stats_mapping_set(
 	__rte_unused struct rte_eth_dev *eth_dev,
@@ -652,6 +653,16 @@ virtio_dev_allmulticast_disable(struct rte_eth_dev *dev)
 		PMD_INIT_LOG(ERR, "Failed to disable allmulticast");
 }
 
+static int 
+virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu) 
+{
+   struct virtio_hw *hw = dev->data->dev_private;
+   if (unlikely(mtu < (uint32_t)hw->vtnet_hdr_size + ETHER_HDR_LEN)) {
+       return -1;
+   }
+   return 0;
+}
+
 /*
  * dev_ops for virtio, bare necessities for basic operation
  */
@@ -664,6 +675,7 @@ static const struct eth_dev_ops virtio_eth_dev_ops = {
 	.promiscuous_disable     = virtio_dev_promiscuous_disable,
 	.allmulticast_enable     = virtio_dev_allmulticast_enable,
 	.allmulticast_disable    = virtio_dev_allmulticast_disable,
+	.mtu_set                 = virtio_mtu_set,
 
 	.dev_infos_get           = virtio_dev_info_get,
 	.stats_get               = virtio_dev_stats_get,
-- 
2.9.3.windows.1

Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

[Stephen Hemminger]

On Fri, 26 Aug 2016 20:54:28 -0400
souvikdey33 <sodey@sonusnet.com> wrote:

> This functionality is required mostly in the cloud infrastructure.
> For example, if we use gre or vxlan network between compute and
> controller, then we should not use 1500 mtu in the guest as with
> encapsulation the sixe of the packet will be more and will get 
> dropped in the infrastructure. So, in that case we should honor
> the mtu size sent by the dhcp server and configure the same on 
> the virtual interfaces in the guest. This will also keep a 
> consistent mtu through out the infrastructure.
> 
> souvikdey33 (1):
>   Signed-off-by: Souvik Dey <sodey@sonusnet.com>
> 
>  drivers/net/virtio/virtio_ethdev.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 

Thanks for the patch, it is a good step forward but it looks
like more code is needed to do this safely. At a minimum,
need to check that MTU is not greater than VIRTIO_MAX_RX_PKTLEN.
And error return should be negative errno not -1.

Something like:
   if (mtu < VIRTIO_MIN_MTU || mtu > VIRTIO_MAX_RX_PKTLEN)
	return -EINVAL;

Looking at Linux driver, it allows MTU of up to 64K, yet DPDK
only allows 9728.  That should probably be fixed.

Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

[Dey, Souvik]

Hi ,
	Currently as you have mentioned, I have changed the code to:
static int 
 virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu) 
 {
    struct virtio_hw *hw = dev->data->dev_private;
-   if (unlikely(mtu < (uint32_t)hw->vtnet_hdr_size + ETHER_HDR_LEN)) {
-       return -1;
+   if (unlikely(mtu < VIRTIO_MIN_RX_BUFSIZE || mtu > VIRTIO_MAX_RX_PKTLEN)) {
+	   PMD_INIT_LOG(ERR,"Mtu should be between 64 and 9728."
+       return -EINVAL;
    }
    return 0;
 }

Yes, we should support till 64K as the kernel does , but I need to go through the changes and test it properly before submitting it for review. Moreover I was thinking with the changes in the mtu, we should also support multi-segment buffers in kni. What do you suggest ?

--
Regards,
Souvik

-----Original Message-----
From: Stephen Hemminger [mailto:stephen@networkplumber.org] 
Sent: Saturday, August 27, 2016 8:16 PM
To: Dey, Souvik <sodey@sonusnet.com>
Cc: huawei.xie@intel.com; yuanhan.liu@linux.intel.com; dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

On Fri, 26 Aug 2016 20:54:28 -0400
souvikdey33 <sodey@sonusnet.com> wrote:

> This functionality is required mostly in the cloud infrastructure.
> For example, if we use gre or vxlan network between compute and 
> controller, then we should not use 1500 mtu in the guest as with 
> encapsulation the sixe of the packet will be more and will get dropped 
> in the infrastructure. So, in that case we should honor the mtu size 
> sent by the dhcp server and configure the same on the virtual 
> interfaces in the guest. This will also keep a consistent mtu through 
> out the infrastructure.
> 
> souvikdey33 (1):
>   Signed-off-by: Souvik Dey <sodey@sonusnet.com>
> 
>  drivers/net/virtio/virtio_ethdev.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 

Thanks for the patch, it is a good step forward but it looks like more code is needed to do this safely. At a minimum, need to check that MTU is not greater than VIRTIO_MAX_RX_PKTLEN.
And error return should be negative errno not -1.

Something like:
   if (mtu < VIRTIO_MIN_MTU || mtu > VIRTIO_MAX_RX_PKTLEN)
	return -EINVAL;

Looking at Linux driver, it allows MTU of up to 64K, yet DPDK only allows 9728.  That should probably be fixed.

Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

[Stephen Hemminger]

On Sun, 28 Aug 2016 22:43:54 +0000
"Dey, Souvik" <sodey@sonusnet.com> wrote:

> Hi ,
> 	Currently as you have mentioned, I have changed the code to:
> static int 
>  virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu) 
>  {
>     struct virtio_hw *hw = dev->data->dev_private;
> -   if (unlikely(mtu < (uint32_t)hw->vtnet_hdr_size + ETHER_HDR_LEN)) {
> -       return -1;
> +   if (unlikely(mtu < VIRTIO_MIN_RX_BUFSIZE || mtu > VIRTIO_MAX_RX_PKTLEN)) {
> +	   PMD_INIT_LOG(ERR,"Mtu should be between 64 and 9728."
> +       return -EINVAL;
>     }
>     return 0;
>  }
> 
> Yes, we should support till 64K as the kernel does , but I need to go through the changes and test it properly before submitting it for review. Moreover I was thinking with the changes in the mtu, we should also support multi-segment buffers in kni. What do you suggest ?

This looks good, but you really don't need likely/unlikely in this code.
It is not at all performance critical.

Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

[Dey, Souvik]

Thanks , I will remove the unlikely from the conditionally statement.
Below is the complete patch :

>From 2e4b391fe90ba5e617611e341a7d260dd3dd9144 Mon Sep 17 00:00:00 2001
From: souvikdey33 <sodey@sonusnet.com>
Date: Fri, 26 Aug 2016 20:46:21 -0400
Subject: [PATCH v2 2/3] Signed-off-by: Souvik Dey <sodey@sonusnet.com>

Fixes: 1fb8e8896ca8 ("Signed-off-by: Souvik Dey <sodey@sonusnet.com>")
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>

Virtio interfaces should also support setting of mtu, as in case of cloud
it is expected to have the consistent mtu across the infrastructure that
the dhcp server sends and not hardcoded to 1500(default).
---
 drivers/net/virtio/virtio_ethdev.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
index 07d6449..da16ad4 100644
--- a/drivers/net/virtio/virtio_ethdev.c
+++ b/drivers/net/virtio/virtio_ethdev.c
@@ -92,6 +92,7 @@ static void virtio_mac_addr_add(struct rte_eth_dev *dev,
 static void virtio_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index);
 static void virtio_mac_addr_set(struct rte_eth_dev *dev,
 				struct ether_addr *mac_addr);
+static int  virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int virtio_dev_queue_stats_mapping_set(
 	__rte_unused struct rte_eth_dev *eth_dev,
@@ -652,6 +653,16 @@ virtio_dev_allmulticast_disable(struct rte_eth_dev *dev)
 		PMD_INIT_LOG(ERR, "Failed to disable allmulticast");
 }
 
+static int 
+virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu) 
+{
+   struct virtio_hw *hw = dev->data->dev_private;
+   if (mtu < VIRTIO_MIN_RX_BUFSIZE || mtu > VIRTIO_MAX_RX_PKTLEN) {
+	   PMD_INIT_LOG(ERR,"Mtu should be between 64 and 9728."
+          return -EINVAL;
+   }
+   return 0;
+}
+
 /*
  * dev_ops for virtio, bare necessities for basic operation
  */
@@ -664,6 +675,7 @@ static const struct eth_dev_ops virtio_eth_dev_ops = {
 	.promiscuous_disable     = virtio_dev_promiscuous_disable,
 	.allmulticast_enable     = virtio_dev_allmulticast_enable,
 	.allmulticast_disable    = virtio_dev_allmulticast_disable,
+	.mtu_set                 = virtio_mtu_set,
 
 	.dev_infos_get           = virtio_dev_info_get,
 	.stats_get               = virtio_dev_stats_get,
-- 
2.9.3.windows.1

As I am submitting to DPDK for the first time can you please guide with the next steps.

--
Regards,
Souvik

-----Original Message-----
From: Stephen Hemminger [mailto:stephen@networkplumber.org] 
Sent: Monday, August 29, 2016 3:33 PM
To: Dey, Souvik <sodey@sonusnet.com>
Cc: huawei.xie@intel.com; yuanhan.liu@linux.intel.com; dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH v1] add mtu set in virtio

On Sun, 28 Aug 2016 22:43:54 +0000
"Dey, Souvik" <sodey@sonusnet.com> wrote:

> Hi ,
> 	Currently as you have mentioned, I have changed the code to:
> static int
>  virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)  {
>     struct virtio_hw *hw = dev->data->dev_private;
> -   if (unlikely(mtu < (uint32_t)hw->vtnet_hdr_size + ETHER_HDR_LEN)) {
> -       return -1;
> +   if (unlikely(mtu < VIRTIO_MIN_RX_BUFSIZE || mtu > VIRTIO_MAX_RX_PKTLEN)) {
> +	   PMD_INIT_LOG(ERR,"Mtu should be between 64 and 9728."
> +       return -EINVAL;
>     }
>     return 0;
>  }
> 
> Yes, we should support till 64K as the kernel does , but I need to go through the changes and test it properly before submitting it for review. Moreover I was thinking with the changes in the mtu, we should also support multi-segment buffers in kni. What do you suggest ?

This looks good, but you really don't need likely/unlikely in this code.
It is not at all performance critical.