| Message ID | 1427368406-28084-1-git-send-email-jonathan.davies@citrix.com |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On Thu, 2015-03-26 at 11:13 +0000, Jonathan Davies wrote: > xen-netfront limits transmitted skbs to be at most 44 segments in size. However, > GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 > bytes each. This slight reduction in the size of packets means a slight loss in > efficiency. > > Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to > XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, > where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. > > The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s > 6c09fa09d) in determining when to split an skb into two is > sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. > > So the maximum permitted size of an skb is calculated to be > (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. > > Intuitively, this looks like the wrong formula -- we don't need two TCP headers. > Instead, there is no need to deviate from the default gso_max_size of 65536 as > this already accommodates the size of the header. > > Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments > of 1448 bytes each), as observed via tcpdump. This patch makes netfront send > skbs of up to 65160 bytes (45 segments of 1448 bytes each). > > Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") > Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> > --- > drivers/net/xen-netfront.c | 2 -- > 1 file changed, 2 deletions(-) > > diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c > index e9b960f..fb6e978 100644 > --- a/drivers/net/xen-netfront.c > +++ b/drivers/net/xen-netfront.c > @@ -1279,8 +1279,6 @@ static struct net_device *xennet_create_dev(struct xenbus_device *dev) > netdev->ethtool_ops = &xennet_ethtool_ops; > SET_NETDEV_DEV(netdev, &dev->dev); > > - netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER); > - > np->netdev = netdev; > > netif_carrier_off(netdev); Hmm, this partially reverts commit 9ecd1a75d977e2e8c48139c7d3efed183f898d94 Why xennet_change_mtu() is not changed by your patch ? -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello. On 3/26/2015 2:13 PM, Jonathan Davies wrote: > xen-netfront limits transmitted skbs to be at most 44 segments in size. However, > GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 > bytes each. This slight reduction in the size of packets means a slight loss in > efficiency. > Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to c/s == commit? Please also specify that commit's summary line in parens. > XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, > where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. > The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s > 6c09fa09d) in determining when to split an skb into two is Likewise. > sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. > So the maximum permitted size of an skb is calculated to be > (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. > Intuitively, this looks like the wrong formula -- we don't need two TCP headers. > Instead, there is no need to deviate from the default gso_max_size of 65536 as > this already accommodates the size of the header. > Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments > of 1448 bytes each), as observed via tcpdump. This patch makes netfront send > skbs of up to 65160 bytes (45 segments of 1448 bytes each). > Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") Ah, here's the summary for the first mentioned commit... > Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> WBR, Sergei -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 26/03/15 12:05, Eric Dumazet wrote: > On Thu, 2015-03-26 at 11:13 +0000, Jonathan Davies wrote: >> xen-netfront limits transmitted skbs to be at most 44 segments in size. However, >> GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 >> bytes each. This slight reduction in the size of packets means a slight loss in >> efficiency. >> >> Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to >> XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, >> where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. >> >> The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s >> 6c09fa09d) in determining when to split an skb into two is >> sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. >> >> So the maximum permitted size of an skb is calculated to be >> (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. >> >> Intuitively, this looks like the wrong formula -- we don't need two TCP headers. >> Instead, there is no need to deviate from the default gso_max_size of 65536 as >> this already accommodates the size of the header. >> >> Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments >> of 1448 bytes each), as observed via tcpdump. This patch makes netfront send >> skbs of up to 65160 bytes (45 segments of 1448 bytes each). >> >> Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") >> Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> >> --- >> drivers/net/xen-netfront.c | 2 -- >> 1 file changed, 2 deletions(-) >> >> diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c >> index e9b960f..fb6e978 100644 >> --- a/drivers/net/xen-netfront.c >> +++ b/drivers/net/xen-netfront.c >> @@ -1279,8 +1279,6 @@ static struct net_device *xennet_create_dev(struct xenbus_device *dev) >> netdev->ethtool_ops = &xennet_ethtool_ops; >> SET_NETDEV_DEV(netdev, &dev->dev); >> >> - netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER); >> - >> np->netdev = netdev; >> >> netif_carrier_off(netdev); > > Hmm, this partially reverts commit > 9ecd1a75d977e2e8c48139c7d3efed183f898d94 > > > > Why xennet_change_mtu() is not changed by your patch ? I think you are right: the mtu calculation suffers from the same problem. I believe the value of mtu relates to the size of the whole packet, including the header (which is why the value of dev->mtu is normally 1500 rather than 1448). Wei, as the author of commit 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header"), do you agree that the max mtu formula should not need to subtract MAX_TCP_HEADER? Regards, Jonathan -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Mar 26, 2015 at 03:08:58PM +0000, Jonathan Davies wrote: > > On 26/03/15 12:05, Eric Dumazet wrote: > >On Thu, 2015-03-26 at 11:13 +0000, Jonathan Davies wrote: > >>xen-netfront limits transmitted skbs to be at most 44 segments in size. However, > >>GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 > >>bytes each. This slight reduction in the size of packets means a slight loss in > >>efficiency. > >> > >>Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to > >> XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, > >>where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. > >> > >>The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s > >>6c09fa09d) in determining when to split an skb into two is > >> sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. > >> > >>So the maximum permitted size of an skb is calculated to be > >> (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. > >> > >>Intuitively, this looks like the wrong formula -- we don't need two TCP headers. > >>Instead, there is no need to deviate from the default gso_max_size of 65536 as > >>this already accommodates the size of the header. > >> > >>Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments > >>of 1448 bytes each), as observed via tcpdump. This patch makes netfront send > >>skbs of up to 65160 bytes (45 segments of 1448 bytes each). > >> > >>Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") > >>Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> > >>--- > >> drivers/net/xen-netfront.c | 2 -- > >> 1 file changed, 2 deletions(-) > >> > >>diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c > >>index e9b960f..fb6e978 100644 > >>--- a/drivers/net/xen-netfront.c > >>+++ b/drivers/net/xen-netfront.c > >>@@ -1279,8 +1279,6 @@ static struct net_device *xennet_create_dev(struct xenbus_device *dev) > >> netdev->ethtool_ops = &xennet_ethtool_ops; > >> SET_NETDEV_DEV(netdev, &dev->dev); > >> > >>- netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER); > >>- > >> np->netdev = netdev; > >> > >> netif_carrier_off(netdev); > > > >Hmm, this partially reverts commit > >9ecd1a75d977e2e8c48139c7d3efed183f898d94 > > > > > > > >Why xennet_change_mtu() is not changed by your patch ? > > I think you are right: the mtu calculation suffers from the same problem. I > believe the value of mtu relates to the size of the whole packet, including > the header (which is why the value of dev->mtu is normally 1500 rather than > 1448). > > Wei, as the author of commit 9ecd1a75d977 ("xen-netfront: reduce > gso_max_size to account for max TCP header"), do you agree that the max mtu > formula should not need to subtract MAX_TCP_HEADER? > IIRC at the time I wrote that patch I needed to subtract MAX_TCP_HEADER otherwise netfront would generate oversized packets then get marked as malicious by backend. I think your reasoning is straightforward. Probably other core driver changes have somehow mitigated the issues I saw. Presuming you tested this change and saw no problems, I'm of course happy with making netfront more efficient. :-) Wei. > Regards, > Jonathan -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 30/03/15 14:46, Wei Liu wrote: > On Thu, Mar 26, 2015 at 03:08:58PM +0000, Jonathan Davies wrote: >> >> On 26/03/15 12:05, Eric Dumazet wrote: >>> On Thu, 2015-03-26 at 11:13 +0000, Jonathan Davies wrote: >>>> xen-netfront limits transmitted skbs to be at most 44 segments in size. However, >>>> GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 >>>> bytes each. This slight reduction in the size of packets means a slight loss in >>>> efficiency. >>>> >>>> Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to >>>> XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, >>>> where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. >>>> >>>> The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s >>>> 6c09fa09d) in determining when to split an skb into two is >>>> sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. >>>> >>>> So the maximum permitted size of an skb is calculated to be >>>> (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. >>>> >>>> Intuitively, this looks like the wrong formula -- we don't need two TCP headers. >>>> Instead, there is no need to deviate from the default gso_max_size of 65536 as >>>> this already accommodates the size of the header. >>>> >>>> Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments >>>> of 1448 bytes each), as observed via tcpdump. This patch makes netfront send >>>> skbs of up to 65160 bytes (45 segments of 1448 bytes each). >>>> >>>> Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") >>>> Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> >>>> --- >>>> drivers/net/xen-netfront.c | 2 -- >>>> 1 file changed, 2 deletions(-) >>>> >>>> diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c >>>> index e9b960f..fb6e978 100644 >>>> --- a/drivers/net/xen-netfront.c >>>> +++ b/drivers/net/xen-netfront.c >>>> @@ -1279,8 +1279,6 @@ static struct net_device *xennet_create_dev(struct xenbus_device *dev) >>>> netdev->ethtool_ops = &xennet_ethtool_ops; >>>> SET_NETDEV_DEV(netdev, &dev->dev); >>>> >>>> - netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER); >>>> - >>>> np->netdev = netdev; >>>> >>>> netif_carrier_off(netdev); >>> >>> Hmm, this partially reverts commit >>> 9ecd1a75d977e2e8c48139c7d3efed183f898d94 >>> >>> >>> >>> Why xennet_change_mtu() is not changed by your patch ? >> >> I think you are right: the mtu calculation suffers from the same problem. I >> believe the value of mtu relates to the size of the whole packet, including >> the header (which is why the value of dev->mtu is normally 1500 rather than >> 1448). >> >> Wei, as the author of commit 9ecd1a75d977 ("xen-netfront: reduce >> gso_max_size to account for max TCP header"), do you agree that the max mtu >> formula should not need to subtract MAX_TCP_HEADER? >> > > IIRC at the time I wrote that patch I needed to subtract MAX_TCP_HEADER > otherwise netfront would generate oversized packets then get marked as > malicious by backend. > > I think your reasoning is straightforward. Probably other core driver > changes have somehow mitigated the issues I saw. > > Presuming you tested this change and saw no problems, I'm of course > happy with making netfront more efficient. :-) Okay, thanks for confirming. I'll post a v2 including the change to xennet_change_mtu. Jonathan -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c index e9b960f..fb6e978 100644 --- a/drivers/net/xen-netfront.c +++ b/drivers/net/xen-netfront.c @@ -1279,8 +1279,6 @@ static struct net_device *xennet_create_dev(struct xenbus_device *dev) netdev->ethtool_ops = &xennet_ethtool_ops; SET_NETDEV_DEV(netdev, &dev->dev); - netif_set_gso_max_size(netdev, XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER); - np->netdev = netdev; netif_carrier_off(netdev);
xen-netfront limits transmitted skbs to be at most 44 segments in size. However, GSO permits up to 65536 bytes, which means a maximum of 45 segments of 1448 bytes each. This slight reduction in the size of packets means a slight loss in efficiency. Since c/s 9ecd1a75d, xen-netfront sets gso_max_size to XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER, where XEN_NETIF_MAX_TX_SIZE is 65535 bytes. The calculation used by tcp_tso_autosize (and also tcp_xmit_size_goal since c/s 6c09fa09d) in determining when to split an skb into two is sk->sk_gso_max_size - 1 - MAX_TCP_HEADER. So the maximum permitted size of an skb is calculated to be (XEN_NETIF_MAX_TX_SIZE - MAX_TCP_HEADER) - 1 - MAX_TCP_HEADER. Intuitively, this looks like the wrong formula -- we don't need two TCP headers. Instead, there is no need to deviate from the default gso_max_size of 65536 as this already accommodates the size of the header. Currently, the largest skb transmitted by netfront is 63712 bytes (44 segments of 1448 bytes each), as observed via tcpdump. This patch makes netfront send skbs of up to 65160 bytes (45 segments of 1448 bytes each). Fixes: 9ecd1a75d977 ("xen-netfront: reduce gso_max_size to account for max TCP header") Signed-off-by: Jonathan Davies <jonathan.davies@citrix.com> --- drivers/net/xen-netfront.c | 2 -- 1 file changed, 2 deletions(-)