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>
Limiting allowed channels per device may be required and is commonly
supported on other drivers, so include a pending patch to add support for
the same.
Signed-off-by: Robert Marko <robimarko@gmail.com>
As commit 3ce1e4c3d3 ("d1: define subtarget specifically") added the
'generic' subtarget, without 'BUILD_SUBTARGET' the correspond U-Boot
package will be no longer selected automatically.
Fixes: 3ce1e4c3d3 ("d1: define subtarget specifically")
Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
Update to the latest upstream release to include recent improvements and
bugfixes. Update copyright, fix typo in PKG_NAME, and remove unneeded use
of MAKE_VARS definition in Makefile. Drop 001-cflags.patch and simplify
002-includes.patch after refreshing. Also simplify LTO/DCE build flags.
Link: https://github.com/libbpf/bpftool/releases/tag/v7.4.0
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Update to the latest upstream release to include recent improvements and
bugfixes, and update copyright. Remove MAKE_VARS usage in Makefile and drop
001-cflags.patch which are no longer needed. Also add flags to disable LTO,
mistakenly dropped earlier.
Link: https://github.com/libbpf/libbpf/releases/tag/v1.4.0
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
only available from >= 1.0.15
Comments are useful. Apparently this config parameter was committed when
openwrt used an older version of lldpd which did not yet support it.
Signed-off-by: Paul Donald <newtwen+github@gmail.com>
Starting with Linux 6.3 the .probe call no longer got the id parameter,
see also commit torvalds/linux@03c835f498
("i2c: Switch .probe() to not take an id parameter").
As the parameter is anyway unused by the driver, drop it when
building the GCA230718 LED driver for newer kernels.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
HW specifications:
* Mediatek MT7981A
* 256MB SPI-NAND
* 512MB DRAM
* Uplink: 1 x 10/100/1000Base-T Ethernet, Auto MDIX, RJ-45 with 802.3at
PoE (Built-in GBe PHY)
* LAN: 1 x 10/100/1000Base-T Ethernet, Auto MDIX, RJ-45 (Airoha EN8801SC)
* 1 Tricolor LED
* Reset button
* 12V/2.0A DC input
Installation:
Board comes with OpenWifi/TIP which is OpenWrt based, so sysupgrade can
be used directly over SSH.
Signed-off-by: Robert Marko <robert.marko@sartura.hr>
This module supports the Microchip Technology Inc (SMSC)
EMC2301/EMC2302/EMC2303/EMC2305 fan speed PWM controller chips.
Signed-off-by: Oleg S <remittor@gmail.com>
Historically it's possible to leave the `SUBTARGETS` undefined and
automatically fallback to a "generic" subtarget. This however breaks
various downstream scripts which may have expectations around filenames:
While some targets with an explicit generic subtarget contain `generic`
in the filenames of artifacts, implicit "subtargets" don't.
Right now this breaks the CI[1], possibly also scripts using the ImageBuilders.
This commit removes all code that support implicit handling of
subtargets and instead requires every target to define "SUBTARGETS".
[1]: https://github.com/openwrt/openwrt/actions/runs/8592821105/job/23548273630
Signed-off-by: Paul Spooren <mail@aparcar.org>
e209a4ced1d8 add strdupa macro for compatibility
af1962b9a609 uclient: add helper function for getting ustream-ssl context/ops
488f1d52cfd2 http: add helper function for checking redirect status
b6e5548a3ecc uclient: defer read notifications to uloop timer
352fb3eeb408 http: call ustream_poll if not enough read data is available
e611e6d0ff0b add ucode binding
ddb18d265757 uclient: add function for getting the amount of pending read/write data
980220ad1762 ucode: fix a few ucode binding issues
6c16331e4bf5 ucode: add support for using a prototype for cb, pass it to callbacks
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This is mostly a cosmetic cleanup. The absence of
the return statement was not causing any problems.
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
So, when updating the hash for at91bootstrap it was done via CHECK_ALL=1
so that updated the PKG_MIRROR_HASH for the main v4 version hash, but
at91bootstrap checkout version depends on the subtarget as well.
Choosing to build for sam9x will change the at91bootstrap version to v3
and this hash was not refreshed thus causing the CI to fail.
Fixes: 6918c637b7 ("treewide: package: update missed hashes after switch to ZSTD")
Signed-off-by: Robert Marko <robimarko@gmail.com>
With the switch to ZSTD for git clone packaging, hashes have changed so
fixup remaining package hashes that were missed in the inital update.
Fixes: b3c1c57 ("treewide: update PKG_MIRROR_HASH to zst")
Signed-off-by: Robert Marko <robimarko@gmail.com>
When using zst instead of xz, the hash changes.
This was missed in the initial treewide updated.
Fixes: b3c1c57a35 ("treewide: update PKG_MIRROR_HASH to zst")
Signed-off-by: Robert Marko <robimarko@gmail.com>
When an IBBS interface is configured for IBSS legacy mode, wdev.htmode
is empty. This is empty string results in an empty positional argument
to the "ibbs join" command, for example:
iw dev phy0-ibss0 ibss join crymesh 2412 '' fixed-freq beacon-interval 100
This empty argument is interpreted as an invalid HT mode by 'iw',
causing the entire command to fail and print a "usage" message:
daemon.notice netifd: radio0 (4527): Usage: iw [options] \
dev <devname> ibss join <SSID> <freq in MHz> ...
Although nobody will ever need more than 640K of IBSS, explicitly use
"NOHT" if an HT mode is not given. This fixes the problem.
Fixes: e56c5f7b27 ("hostapd: add ucode support, use ucode for the main ubus object")
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name> [extend to cover more cases]
When using zst instead of xz, the hash changes. This commit fixes the
hash for packages and tools in core.
Signed-off-by: Paul Spooren <mail@aparcar.org>
On Fedora 40 system, some compile error happens when
building iconv-ostream.c. Linking to libiconv-full
fixes this.
Signed-off-by: Yanase Yuki <dev@zpc.st>
hostapd packages were accidentally left out. Clean up this mess by
changing the dependencies to hostapd-common
Signed-off-by: Felix Fietkau <nbd@nbd.name>
With the default BUILD_BOT configuration on a linux 6.6 kernel,
the WNDR4700's kernel no longer fits into the alloted ~3.5MiB,
even with LZMA compression.
Bigger kernels are possible, but there's a problem with Netgear's
"bootcmd":
> if loadn_dniimg 0 0x180000 0x4e0000 && chk_dniimg 0x4e0000; then nand read 0x800000 0x180000 0x20000;bootm 0x500000 - 0x800040;else fw_recovery; fi"
This loads the dni-image starting offset 0x180000 from the NAND
flash (which is the DTB partition) to 0x4e0000 in the RAM. It then
checks whenever the provided image is "valid". If it is then it
reads the DTB again to 0x800000 in the RAM and starts the extraction
and boot process. (If the image wasn't valid then it starts the
automated firmware recovery).
The issues here are that first: the kernel image gets "squeezed"
between 0x500040 and 0x7fffff... And second, the decompressor
only has area 0x0 - 0x500000 for decompression.
Hence the image now requires to update the bootcmd by providing
new values (which have been successfully tested with the original
Netgear WNDR4700 v1.0.0.56 firmware) for the RAM locations and
make full use of the fact that loadn_dniimg loads the DTB as well.
This needs to be done only once. Just connect a serial adapter to
interface with uboot and overwrite (and save) the new bootcmd.
WARNING: The serial port needs a TTL/RS-232 3.3v level converter!
Steps:
0. Power-off the WNDR4700
1. Connect the serial interface (you need to open the WNDR4700)
2. Power-up the WNDR4700
3. Monitor the boot-sequence and hit "Enter"-key when it says:
"Hit any key to stop autoboot" (Be quick, you have a ~2 second window)
4. in the Prompt enter the following commands (copy & paste)
setenv bootcmd "if loadn_dniimg 0 0x180000 0xce0000 && chk_dniimg 0xce0000; then bootm 0xd00000 - 0xce0040;else fw_recovery; fi"
saveenv
run bootcmd
Note: This new bootcmd will also unbrick devices that were bricked
by the bigger 4.19-6.1 kernels.
Note2: This method was tested with a WNDR4700. A big kernel with most
debug features enabled on v6.6.22 measured 4.30 MiB when compressed
with lzma. The uncompressed kernel is 12.34 MiB. This is over the 3 MiB,
the device reserves for the kernel... But it booted! For bigger kernels,
the device needs repartitioning of the the ubi partition due to the
kernel+dtb not fitting into the partition.
Note3: For initramfs development. I would advice to load the initramfs
images to 0x800000 (or higher). i.e.: tftp 800000 wndr4700.bin
Note4: the fw_recovery uboot command to transfer the factory image to
the flash still works.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
The PKG_MIRROR_HASH was wrong (again), likely due to an old set of tools
which did not contain the downgrade of xz.
Ref 2070049 unetd: fix PKG_MIRROR_HASH
Fix 89c594e libubox: update to Git HEAD (2024-03-29)"
Signed-off-by: Paul Spooren <mail@aparcar.org>
Fix compilation warning enum-int-mismatch which results in failure to
build kmod-ltq-vmmc in case CONFIG_KERNEL_WERROR is set.
.../build_dir/target-mips_24kc_musl/linux-lantiq_xrx200/drv_vmmc-1.9.0/src/mps/drv_mps_vmmc_common.c:392:14: error: conflicting types for 'ifx_mps_fastbuf_init' due to enum/integer mismatch; have 'IFX_return_t(void)' [-Werror=enum-int-mismatch]
392 | IFX_return_t ifx_mps_fastbuf_init (IFX_void_t)
| ^~~~~~~~~~~~~~~~~~~~
.../build_dir/target-mips_24kc_musl/linux-lantiq_xrx200/drv_vmmc-1.9.0/src/mps/drv_mps_vmmc_common.c:120:13: note: previous declaration of 'ifx_mps_fastbuf_init' with type 'IFX_int32_t(void)' {aka 'int(void)'}
120 | IFX_int32_t ifx_mps_fastbuf_init (IFX_void_t);
| ^~~~~~~~~~~~~~~~~~~~
.../build_dir/target-mips_24kc_musl/linux-lantiq_xrx200/drv_vmmc-1.9.0/src/mps/drv_mps_vmmc_common.c:420:14: error: conflicting types for 'ifx_mps_fastbuf_close' due to enum/integer mismatch; have 'IFX_return_t(void)' [-Werror=enum-int-mismatch]
420 | IFX_return_t ifx_mps_fastbuf_close (IFX_void_t)
| ^~~~~~~~~~~~~~~~~~~~~
.../build_dir/target-mips_24kc_musl/linux-lantiq_xrx200/drv_vmmc-1.9.0/src/mps/drv_mps_vmmc_common.c:121:13: note: previous declaration of 'ifx_mps_fastbuf_close' with type 'IFX_int32_t(void)' {aka 'int(void)'}
121 | IFX_int32_t ifx_mps_fastbuf_close (IFX_void_t);
| ^~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Refresh patches and bump PKG_RELEASE while at it.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
- move build/ifdef related changes together to the 200 patch range
- reduce adding/removing include statements across patches
- move patches away from the 99x patch range to simplify maintenance
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This adds From:, Date: and Subject: to patches, allowing one to run 'git
am' to import the patches to a hostapd git repository.
From: and Date: fields were taken from the OpenWrt commit where the
patches were first introduced.
Most of the Subject: also followed suit, except for:
- 300-noscan.patch: Took the description from the LuCI web interface
- 350-nl80211_del_beacon_bss.patch: Used the file name
The order of the files in the patch was changed to match what git
format-patch does.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Patch 050-build_fix.patch fixes the abscence of sha384-kdf.o from the
list of needed objetct files when FILS is selected without any other
option that will select the .o file.
While it is a bug waiting to be fixes upstream, it is not needed for
OpenWrt use case, because OWE already selects sha384-kdf.o, and FILS is
selected along with OWE.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
This brings many changes, including fixes for a couple of memory leaks,
and improved interoperability with 802.11r. There are also many changes
related to 802.11be, which is not enabled at this time.
Fixed upstream:
- 022-hostapd-fix-use-of-uninitialized-stack-variables.patch
- 180-driver_nl80211-fix-setting-QoS-map-on-secondary-BSSs.patch
- 993-2023-10-28-ACS-Fix-typo-in-bw_40-frequency-array.patch
Switch PKG_SOURCE_URL to https, since http is not currently working.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Tested-by: Ilya Katsnelson <me@0upti.me>
Tested by: Andrew Sim <andrewsimz@gmail.com>
a2fce001819e CI: add build test run
12bda4bdb197 CI: add CodeQL workflow tests
eb9bcb64185a ustream: prevent recursive calls to the read callback
Signed-off-by: Felix Fietkau <nbd@nbd.name>
It seems that somehow a wrong hash has slipped past PR CI again.
Fixes: 9ef4f7f919 ("qualcommax: ipq60xx: add yuncore fap650 support")
Signed-off-by: Robert Marko <robimarko@gmail.com>
Include the CONFIG_KVM_SMM option in the kvm-x86 package to enable system management mode emulation on KVM.
Co-authored-by: Stefan Hellermann <stefan@the2masters.de>
Signed-off-by: Mieczyslaw Nalewaj <namiltd@yahoo.com>
Fix error: Package kmod-lan743x is missing dependencies for the following libraries:
fixed_phy.ko
Signed-off-by: Mieczyslaw Nalewaj <namiltd@yahoo.com>
Kernel 6.2 folded virqfd (eventd interface for VFIO interrupts)
into the base vfio module, it is no longer a tristate option.
Change suggested by vincejv on GitHub:
https://github.com/openwrt/openwrt/pull/14868#issuecomment-1998260124
Signed-off-by: Mathew McBride <matt@traverse.com.au>
Instead of redefining the version schema in cryptodev, use the one
automatically defined via `kernel.mk`.
Specifically this changes the version from <kernel>+<package> to
<kernel>.<package> and thereby making it compatible with APK.
Signed-off-by: Paul Spooren <mail@aparcar.org>
Our CI on GitHub as well as my local machine generates a different
PKG_MIRROR_HASH from what Felix uploaded the other day.
After receiving Felix file, both have indeed different hashes, however
when unpackaged via `xz -d` both have the same tarball content.
Below the checksums to compare:
a62bef497078c7b825f11fc8358c1a43f5db3e6d4b97812044f7653d60747d5b dl/unetd-2024.03.31~80645766.tar.xz
fbdac59581742bf208c18995b1d69d9848c93bfce487e57ba780d959e0d62fc4 dl/unetd-2024.03.31~80645766_felix.tar.xz
After unpacking:
a7189cae90bc600abf3a3bff3620dc17a9143be8c27d27412de6eb66a1cf1b7d dl/unetd-2024.03.31~80645766.tar
a7189cae90bc600abf3a3bff3620dc17a9143be8c27d27412de6eb66a1cf1b7d dl/unetd-2024.03.31~80645766_felix.tar
The tarball with the wrong hash was accidentally generated without the xz
revert to version 5.4.6
Signed-off-by: Paul Spooren <mail@aparcar.org>
Add patch fixing compilation with kernel 6.6.
class_create now require only the name instead of the module ownership
reference.
Also the kernel enabled checks for enum.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Add patch fixing compilation with kernel 6.6.
class_create now require only the name instead of the module ownership
reference.
Also the kernel enabled checks for enum.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
