summaryrefslogtreecommitdiffstats
path: root/net/core/dev.c
diff options
context:
space:
mode:
authorTom Herbert <therbert@google.com>2011-04-04 22:30:30 -0700
committerDavid S. Miller <davem@davemloft.net>2011-04-04 22:30:30 -0700
commitc6e1a0d12ca7b4f22c58e55a16beacfb7d3d8462 (patch)
tree6955c20538050329d0bdffdf24a787507ae6fdf1 /net/core/dev.c
parent14f98f258f1936e0dba77474bd7eda63f61a9826 (diff)
downloadop-kernel-dev-c6e1a0d12ca7b4f22c58e55a16beacfb7d3d8462.zip
op-kernel-dev-c6e1a0d12ca7b4f22c58e55a16beacfb7d3d8462.tar.gz
net: Allow no-cache copy from user on transmit
This patch uses __copy_from_user_nocache on transmit to bypass data cache for a performance improvement. skb_add_data_nocache and skb_copy_to_page_nocache can be called by sendmsg functions to use this feature, initial support is in tcp_sendmsg. This functionality is configurable per device using ethtool. Presumably, this feature would only be useful when the driver does not touch the data. The feature is turned on by default if a device indicates that it does some form of checksum offload; it is off by default for devices that do no checksum offload or indicate no checksum is necessary. For the former case copy-checksum is probably done anyway, in the latter case the device is likely loopback in which case the no cache copy is probably not beneficial. This patch was tested using 200 instances of netperf TCP_RR with 1400 byte request and one byte reply. Platform is 16 core AMD x86. No-cache copy disabled: 672703 tps, 97.13% utilization 50/90/99% latency:244.31 484.205 1028.41 No-cache copy enabled: 702113 tps, 96.16% utilization, 50/90/99% latency 238.56 467.56 956.955 Using 14000 byte request and response sizes demonstrate the effects more dramatically: No-cache copy disabled: 79571 tps, 34.34 %utlization 50/90/95% latency 1584.46 2319.59 5001.76 No-cache copy enabled: 83856 tps, 34.81% utilization 50/90/95% latency 2508.42 2622.62 2735.88 Note especially the effect on latency tail (95th percentile). This seems to provide a nice performance improvement and is consistent in the tests I ran. Presumably, this would provide the greatest benfits in the presence of an application workload stressing the cache and a lot of transmit data happening. Signed-off-by: Tom Herbert <therbert@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/core/dev.c')
-rw-r--r--net/core/dev.c12
1 files changed, 12 insertions, 0 deletions
diff --git a/net/core/dev.c b/net/core/dev.c
index 02f5637..5d0b4f6 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -5425,6 +5425,14 @@ int register_netdevice(struct net_device *dev)
dev->features &= ~NETIF_F_GSO;
}
+ /* Turn on no cache copy if HW is doing checksum */
+ dev->hw_features |= NETIF_F_NOCACHE_COPY;
+ if ((dev->features & NETIF_F_ALL_CSUM) &&
+ !(dev->features & NETIF_F_NO_CSUM)) {
+ dev->wanted_features |= NETIF_F_NOCACHE_COPY;
+ dev->features |= NETIF_F_NOCACHE_COPY;
+ }
+
/* Enable GRO and NETIF_F_HIGHDMA for vlans by default,
* vlan_dev_init() will do the dev->features check, so these features
* are enabled only if supported by underlying device.
@@ -6182,6 +6190,10 @@ u32 netdev_increment_features(u32 all, u32 one, u32 mask)
}
}
+ /* If device can't no cache copy, don't do for all */
+ if (!(one & NETIF_F_NOCACHE_COPY))
+ all &= ~NETIF_F_NOCACHE_COPY;
+
one |= NETIF_F_ALL_CSUM;
one |= all & NETIF_F_ONE_FOR_ALL;
OpenPOWER on IntegriCloud