Skip to content
Snippets Groups Projects
Commit 9c15feca authored by Daniel Salzman's avatar Daniel Salzman
Browse files

doc/operation: extend XDP description

parent dfdb69bb
Branches
Tags
No related merge requests found
Pipeline #66517 failed with stages
in 3 hours and 5 minutes
......@@ -977,9 +977,13 @@ Pre-requisites
--------------
* Linux kernel 4.18+ (5.x+ is recommended for optimal performance).
* A multiqueue network card, which offers enough Combined channels, with native
XDP support is highly recommended (successfully tested cards are Intel series
500 and 700).
* A multiqueue network card, which offers enough Combined RX/TX channels, with
native XDP support is highly recommended. Successfully tested cards:
* Intel series 700 (driver `i40e`), maximum number of channels per interface is 64.
* Intel series 500 (driver `ixgbe`), maximum number of channels per interface is 64.
The number of CPUs available has to be at most 64!
* If the `knotd` service is not directly executed in the privileged mode, some
additional Linux capabilities have to be set:
......@@ -989,9 +993,9 @@ Pre-requisites
And insert these lines::
[Service]
CapabilityBoundingSet=CAP_NET_RAW CAP_NET_ADMIN CAP_SYS_ADMIN CAP_SYS_RESOURCE
AmbientCapabilities=CAP_NET_RAW CAP_NET_ADMIN CAP_SYS_ADMIN CAP_SYS_RESOURCE
[Service]
CapabilityBoundingSet=CAP_NET_RAW CAP_NET_ADMIN CAP_SYS_ADMIN CAP_SYS_RESOURCE
AmbientCapabilities=CAP_NET_RAW CAP_NET_ADMIN CAP_SYS_ADMIN CAP_SYS_RESOURCE
Optimizations
-------------
......@@ -1016,4 +1020,6 @@ Limitations
* Systems with big-endian byte ordering require special recompilation of the nameserver.
* IPv4 header and UDP checksums are not verified on received DNS messages.
* DNS over XDP traffic is not visible to common system tools (e.g. firewall, tcpdump etc.).
* BPF filter is not automatically unloaded from the network device.
* BPF filter is not automatically unloaded from the network device. Manual filtr unload::
ip link set dev <ETH> xdp off
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment