The new script uses a different strategy compared to the previous one.
Instead of trying to split flows by hash and spread them to all CPUs,
use RPS to redirect packets to a single core only.
Try to spread NAPI thread and RPS target CPUs across available CPUs
and try to ensure that the NAPI thread is on a different CPU than the
RPS target. This significantly reduces cycles wasted on the scheduler.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Document the existence of this feature. This allows the user to execute a script
at each DHCPv4 event. This is useful, for example, as an ad-hoc way to update a
DDNS entry when (and only when) required.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
The current implementation is significantly lowering lantiq
performace [1][2] by using RPS with non-irq CPUs and XPS
with alternating CPUs.
The previous netifd implementation (by default but could be
configured) simply used all CPUs and this patch essentially
reverts to this behaviour.
The only document suggesting using non-interrupt CPUs is Red
Hat [3] where if the network interrupt rate is extremely high
excluding the CPU that handles network interrupts *may* also
improve performance.
The original packet steering patches [4] advise that optimal
settings for the CPU mask seems to depend on architectures
and cache hierarcy so one size does not fit all. It also
advises that the overhead in processing for a lightly loaded
server can cause performance degradation.
Ideally, proper IRQ balancing is a better option with
the irqbalance daemon or manually.
The kernel does not enable packet steering by default, so
also disable in OpenWRT by default. (Though mvebu with its
hardware scheduling issues [5] might want to enable packet
steering by default.)
Change undocumented "default_ps" parameter to clearer
"packet_steering" parameter. The old parameter was only ever
set in target/linux/mediatek/base-files/etc/uci-defaults/99-net-ps
and matched the default.
[1] https://forum.openwrt.org/t/18-06-4-speed-fix-for-bt-homehub-5a
[2] https://openwrt.ebilan.co.uk/viewtopic.php?f=7&t=1105
[3] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/performance_tuning_guide/network-rps
[4] https://marc.info/?l=linux-netdev&m=125792239522685&w=2
[5] https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=2e1f6f1682d3974d8ea52310e460f1bbe470390fFixes: #1852Fixes: #2573
Signed-off-by: Alan Swanson <reiver@improbability.net>
Change dhcp no/release on shutdown to 'norelease' uci option to match
existing proto dhcpv6 usage.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Acked-by: Hans Dedecker <dedeckeh@gmail.com>
commit eb204d14f75c ("base-files: implement generic service_running")
introduced generic service_running so it's not needed to copy&paste same
3 lines over and over again.
I've removed service_running from netifd/network init script as well,
because it was not working properly, looked quite strange and I didn't
understand the intention:
$ /etc/init.d/network stop
$ service network running && echo "yes" || echo "nope"
( have to wait for 30s )
Command failed: Request timed out
yes
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Commit "generic: ar8216: add mib_poll_interval switch attribute" has added
mib_poll_interval global config option and commit "generic: ar8216: group
MIB counters and use two basic ones only by default" has added mib_type
config option.
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Do not set device runtime property on interfaces in the hotplug handler
and in fixup_interfaces(). This property conflicts with device option
in several proto handlers (mainly QMI and other WWAN/3G protos) and does
not seem to be used anywhere.
Signed-off-by: Ivan Shapovalov <intelfx@intelfx.name>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase]
Remove RPS/XPS support from netifd core, move the logic to a hotplug
script that uses a different policy which provides better performance
and more fairness across flows
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Based on a patch by Alexandru Ardelean.
netifd ubus reload call returns the actual reload error status;
return error status as well in reload_service
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
stop() is overwritten by rc.common, so implement stop_service instead.
While at it, remove the now unnecessary restart() override
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Move the pattern setting from netifd's service script to
/etc/sysctl.conf. Put the timestamp component '%t' just after
executable name '%e' for more natural order from output of ls command.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
SVN-Revision: 46867
When 'wifi down' is called by /etc/init.d/network, it is run from
stop_service( ). This function is in turn invoked from stop( ).
stop( ) messes up the order by first procd_kill-ing the network
settings, then calling wifi to down the wifi networking
interfaces. By redefining stop( ) instead, the proper order is
restored.
Signed-off-by: Michel Stam <m.stam@fugro.nl>
SVN-Revision: 42745
somebody started to set a function returncode in the validation
stuff and everybody copies it, e.g.
myfunction()
{
fire_command
return $?
}
a function automatically returns with the last returncode,
so we can safely remove the command 'return $?'. reference:
http://tldp.org/LDP/abs/html/exit-status.html
"The last command executed in the function or script determines the exit status."
Signed-off-by: Bastian Bittorf <bittorf@bluebottle.com>
SVN-Revision: 42278
