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>
Global attributes enable_mirror_tx/enable_mirror_rx depend on runtime
value of another global attribute mirror_source_port which just resides
in the memory
The same functionality can be achieved by directly setting port
attribute of the same names. E.g. the following two groups of commands
achieve the same thing
swconfig dev switch0 set mirror_source_port 3
swconfig dev switch0 set enable_mirror_tx 1
swconfig dev switch0 set mirror_source_port 4
swconfig dev switch0 set enable_mirror_tx 1
swconfig dev switch0 port 3 set enable_mirror_tx 1
swconfig dev switch0 port 4 set enable_mirror_tx 1
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Image building process was missing "asus-trx" step which resulted in raw
TRX files (without ASUS footer with device id).
Fixes: 0b9de8daa7 ("bcm53xx: add profiles for all other (SoftMAC) devices")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
getopt is the only command where /usr/local/bin is specified explicitly.
All other commands are assumed to exist in the PATH in one form or
another. Remove this exception and require gnugetopt/getopt to be in
the user's PATH.
In the case of macos Homebrew, getopt is 'keg only' hence not linked
into /usr/local/bin whilst other commands are linked and likely found by
virtue of /usr/local/bin being in PATH.
Since 2019 Homebrew is very reluctant to install links that have
potential to override default OS behaviour, eg: following instructions
on our current 'how to build on macos' wiki page:
$ brew ln gnu-getopt --force
Warning: Refusing to link macOS-provided software: gnu-getopt
If you need to have gnu-getopt first in your PATH run:
echo 'export PATH="/usr/local/opt/gnu-getopt/bin:$PATH"' >> ~/.zshrc
A better option for macos is to link getopt as 'gnugetopt' in
/usr/local/bin, thus the build system will find 'gnugetopt' but other
applications looking for just 'getopt' will find the original macos
binary.
Ultimately it makes sense that 'GNU' dependencies are placed in
/usr/local/bin and /usr/local/bin is included in the user's PATH.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Fixes following build error on mpc85xx/p2020:
Package kmod-iio-st_accel is missing dependencies for the following libraries:
regmap-core.ko
Fixes: 2d8f4c4fbd ("kernel: iio: add st-accel driver modules")
Signed-off-by: Petr Štetiar <ynezz@true.cz>
When changing the LED names for the Archer C7 to represent the correct
color, a migration for existing UCI entries was not created.
Add a migration to keep existing LED configurations working.
Fixes commit c79c001b59 ("ar71xx: Archer C7 v1 LED names and RFKILL
fixes")
Signed-off-by: David Bauer <mail@david-bauer.net>
This commit fills the void for current OpenWrt installations which will
be still on old bootloader version even after "x86: add bootloader
upgrade on sysupgrade", since it performs bootloader upgrade only on
sysupgrade. To keep all OpenWrt deploynents on the same GRUB version,
add preinit hook, which will perform upgrade of the bootloader on first
boot after sysupgrade.
It's temporary solution and should be deleted, when the first release
including this hook will no longer be supported by OpenWrt team.
We can assume that all installations should be on current bootolader
version and from there sysupgrade routine will be sufficient.
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
Currently bootloader always stays on the same version as when first
written to boot medium (not true if partition layout changed, which will
trigger sysupgrade process to write full disk image). That creates
inconveniences as it always stays with same features or/and bugs. Users
wishing to add support to additional modules or new version, would need
to write the whole image, potentially destroying previous system
configuration. To fix these, this commit adds additional routine to
sysupgrade which upgrades unconditionally the bootloader to the latest
state provided by OpenWrt.
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
There can be some leftovers from other image recipes, if the same
directory names are used and multiply image types are selected.
Therefore remove directories used in the recipe, before contents for the
image are prepared.
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
Adds kernel modules for various STMicroelectronics accelerometers.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
[added missing commit description]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Add one node for the accel/gyro i2c device and another for the separate
magnetometer device in the lsm9ds1.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
[commit subject/description tweaks, kernel version in patch filename]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Backport kernel module from 5.5 for FXOS8700CQ, which is a small,
low-power, 3-axis linear accelerometer and 3-axis magnetometer combined
into a single package. The device features a selectable I2C or
point-to-point SPI serial interface with 14-bit accelerometer and 16-bit
magnetometer ADC resolution along with smart-embedded functions.
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
[added commit description]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Code was attempting to determine the size of the file
before it was actually known and allocating insufficient
memory space. Images above a certain size caused a
segmentation fault. Moving the calloc() ensured ensured
that large images didn't result in a buffer overflow on
memcpy().
Signed-off-by: Michael T Farnworth <michael@turf.org>
[fixed name in From to match one in SoB]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
pkgconf is a newer, actively maintained implementation of pkg-config that
supports more aspects of the pkg-config file specification and provides a
library interface that applications can use to incorporate intelligent
handling of pkg-config files into themselves (such as build file
generators, IDEs, and compilers). Through its pkg-config compatibility
interface (activated when it is run as "pkg-config"), it also can
completely replace the original implementation.
It is also lighterweight and does not require glib2, as pkg-config does.
On other distros, pkgconf is symlinked to pkg-config. For simplicity here,
it is renamed to pkg-config.real, as in the original package.
Initial results have been positive. As before, pkgconf works as long as
the pkg-config files point to the proper paths.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
[backported upstream fix for Meson]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
The BUILD_VARIANT might differ from UBOOT_CONFIG, so point to a file we
are actually changing. Being here let's call 'Build/Configure/U-Boot'
definition, instead of definig the same command. This'll be more future
proof, if U-Boot configuration procedure will change.
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
If device recipe has specified DEVICE_DTS variable, the dtb is built
anyway by OpenWrt buildroot image rules. Drop the patch and adjust the
location of compiled dtb.
Cc: Scott Roberts <ttocsr@gmail.com>
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
Tar has ability to change current dir, so use that instead additional
command invocation. Also being here, change tar arguments to make final
archive reproducible.
Cc: Scott Roberts <ttocsr@gmail.com>
Signed-off-by: Tomasz Maciej Nowak <tomek_n@o2.pl>
