Florian Eckert [Fri, 30 Nov 2018 13:03:50 +0000 (14:03 +0100)]
openvpn: add generic hotplug mechanism
Pass a default --up and --down executable to each started OpenVPN instance
which triggers /etc/hotplug.d/openvpn/ scripts whenever an instance
goes up or down.
User-configured up and down scripts are invoked by the default shipped
01-user hotplug handler to ensure that existing setups continue to work
as before.
As a consequence of this change, the up, down and script_security OpenVPN
options are removed from the option file, since we're always passing them
via the command line, they do not need to get included into the generated
configuration.
CONFIG_HZ_100 is not set to "y" in generic config anymore since 3a761c90afc0 ("treewide: Don't diverge from upstream default HZ
settings on 4.19"), so we don't need to disable it anymore either.
This was the only remaining case in the entire tree.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Petr Štetiar [Wed, 10 Jun 2020 14:01:06 +0000 (16:01 +0200)]
mvebu: fix bootloader kernel commandline mangling
Currently I'm unable to boot initramfs image with `console=ttyS0,115200`
kernel commandline as the kernel commandline mangling resets kernel
commandline if there is no `root=` option provided, efectively clearing
whatever I pass to the kernel, making the `root=` option mandatory.
So if the kernel commandline mangling is not appropriate just leave the
kernel commandline as it is.
bcm63xx: use the generic nand upgrade on Sercomm devices
The Sercomm devices don't need the jffs2 cleanmarkers for the kernel
partition. With the stock firmware they are only added when the
bootloader partition is upgraded.
Furthermore, on the Sercomm H500-s, the specific cfe_jffs2_upgrade_tar
function causes OOB corruption. As a result of this, combined with the
nature of the mtdsplit driver, the upgrade procedure ends in a bricked
device.
Use the generic nand_do_upgrade provided by the OpenWrt common functions.
Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
mediatek: fix IPv4-only corner case and commit in 99-net-ps
The uci config section network.globals set up in /bin/config_generate
will only be created if /proc/sys/net/ipv6 exists.
Correspondingly, lacking IPv6 support, the command
uci set network.globals.packet_steering=1
will fail with "uci: Invalid argument" as the network.globals config
has not been set up.
Fix that by adding the setup there as well.
While at it, limit the uci commit to the network config file.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Specifications:
- SoC: Atheros AR9344
- RAM: 64MB
- Storage: 8 MB SPI NOR
- Wireless: 2.4GHz N based built into SoC
- Ethernet: 1x 10/100 Mbps with 24V POE IN, 1x 10/100 Mbps
Installation:
Flash factory image through stock firmware WEB UI
or through TFTP
To get to TFTP recovery just hold reset button while powering on for
around 4-5 seconds and release.
Rename factory image to recovery.bin
Stock TFTP server IP:192.168.0.100
Stock device TFTP adress:192.168.0.254
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Christopher Hill [Tue, 26 May 2020 01:05:00 +0000 (21:05 -0400)]
ath79: add Mikrotik rb4xx series drivers
This adds 3 Mikrotik rb4xx series drivers as follows:
rb4xx-cpld: This is in the mfd subsystem, and is the parent CPLD device
that interfaces between the SoC SPI bus and its two children below.
rb4xx-gpio: This is the GPIO expander.
rb4xx-nand: This is the NAND driver.
The history of this code comes in three phases.
1. The first is a May 2015 attempt to push the equivalient ar71xx rb4xx
drivers upstream. See https://lore.kernel.org/patchwork/patch/940880/.
The Pirelli A226M-FWB has a wrong linux partition offset, caused
by a copy-paste error. As of result of this, OpenWrt is currently
broken in this unit.
Fix it.
While at it, also use generic node names and fix the addresses
there as well.
Fixes: a27d59bb4274 ("brcm63xx: switch to new partition layout
specification")
Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
[also fix/update node names, extend commit message] Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
ramips: fix IPv4-only corner case in 01_enable_packet_steering
The uci config section network.globals set up in /bin/config_generate
will only be created if /proc/sys/net/ipv6 exists.
Correspondingly, lacking IPv6 support, the command
uci set network.globals.packet_steering=1
will fail with "uci: Invalid argument" as the network.globals config
has not been set up.
Fix that by adding the setup there as well.
Fixes: dfd62e575c6c ("ramips: enable packet steering by default on mt7621") Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
To prevent "wrong" musl packages which have a new version number
but the package still contains an old version, because
PKG_SOURCE_VERSION was unchanged.
Ref: musl ml https://www.openwall.com/lists/musl/2020/05/22/4
Reviewed-by: Petr Štetiar <ynezz@true.cz> Signed-off-by: Alexander Couzens <lynxis@fe80.eu>
hostapd: hostapd_set_psk_file: fix defaut value for mac
The sender domain has a DMARC Reject/Quarantine policy which disallows
sending mailing list messages using the original "From" header.
To mitigate this problem, the original message has been wrapped
automatically by the mailing list software.
Bringing up of station vlan fails if the optional mac entry isn't set.
The default mac "00:00:00:00:00:00", which should match all stations,
is mistakenly set to the non used variable "isolate". This results in
a wrong formatted .psk file which has to be "vlan_id mac key".
fixes: 5aa2ddd0: hostapd: add support for wifi-station and wifi-vlan sections
Signed-off-by: Johann Neuhauser <johann@it-neuhauser.de>
David Bauer [Sat, 13 Jun 2020 14:57:53 +0000 (16:57 +0200)]
ipq40xx: use zImage for EX6100v2 and EX6150v2
The NETGEAR EX61500v2 and EX6150v2 U-Boot does not support booting LZMA
compressed images. Currently, they are using GZIP compressed kernels,
which results in ledd flash being available to the root and overlay
filesystems.
Using a zImage results in a smaller kernel and therefore increases
available space for rootfs and overlayfs.
In Firmware Upgrade page, upgrade your openwrt-ipq40xx-generic-engenius_emr3500-squashfs-factory.bin directly.
From OpenWrt firmware to EnGenius firmware:
1. Setup a TFTP server on your computer and configure static IP to 192.168.99.8
Put the EnGenius firmware in the TFTP server directory on your computer.
2. Power up EMR3500. Press 4 and then press any key to enter u-boot.
3. Download EnGenius firmware
(IPQ40xx) # tftpboot 0x84000000 openwrt-ipq40xx-emr3500-nor-fw-s.img
4. Flash the firmware
(IPQ40xx) # imgaddr=0x84000000 && source 0x84000000:script
5. Reboot
(IPQ40xx) # reset
Signed-off-by: Yen-Ting-Shen <frank.shen@senao.com>
[squashed update patch, updated to 5.4, dropped BOARD_NAME,
migrated to SOC] Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Sven Eckelmann [Tue, 9 Jun 2020 13:35:04 +0000 (15:35 +0200)]
ipq40xx: essedma: Disable TCP segmentation offload for IPv6
It was noticed that the the whole MAC can hang when transferring data from
one ar40xx port (WAN ports) to the CPU and from the CPU back to another
ar40xx port (LAN ports). The CPU was doing only NATing in that process.
Usually, the problem first starts with a simple data corruption:
$ wget https://cdimage.debian.org/debian-cd/current/amd64/iso-cd/debian-10.4.0-amd64-netinst.iso -O /dev/null
...
Connecting to saimei.ftp.acc.umu.se (saimei.ftp.acc.umu.se)|2001:6b0:19::138|:443... connected.
...
Read error at byte 48807936/352321536 (Decryption has failed.). Retrying.
But after a short while, the whole MAC will stop to react. No traffic can
be transported anymore from the CPU port from/to the AR40xx PHY/switch and
the MAC has to be resetted.
Signed-off-by: Sven Eckelmann <sven@narfation.org> Acked-by: John Crispin <john@phrozen.org>
With several subtargets, the image/Makefile becomes crowded after a
while. Many targets have moved their device definitions to $subtarget.mk
files to have them more organized, let's do this here as well.
Cc: Rafał Miłecki <rafal@milecki.pl> Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
J. Scott Heppler [Fri, 12 Jun 2020 16:04:15 +0000 (18:04 +0200)]
ramips: fix port display for TRENDnet TEW-810DR
This updates the display port order for the TEW-810DR to be in line
with the DIR-810L. Both share the same board and pictures on the
vendors' pages indicate the same external numbering scheme as well.
Signed-off-by: J. Scott Heppler <shep971@centurylink.net>
[replace commit message] Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Installation via stock firmware:
* Install the factory image via the stock firmware web interface
Installation via bootloader Emergency Web Server:
* Connect your PC to the LAN1(PoE) port
* Configure your PC with IP address 192.168.0.90
* Open a serial console to the Console port (115200,8n1)
* Press "q" within 2s when "press 'q' to stop autoboot" appears
* Open http://192.168.0.50 in a browser
* Upload either the factory or the sysupgrade image
* Once you see "write image into flash...OK,dest addr=0x9f070000" you
can power-cycle the device. Ignore "checksum bad" messages.
Setting the MAC addresses for the ethernet interfaces via
/etc/board.d/02_network adds the following snippets to
/etc/config/network:
config device 'lan_eth0_1_dev'
option name 'eth0.1'
option macaddr 'xx:xx:xx:xx:xx:xx'
config device 'wan_eth1_2_dev'
option name 'eth1.2'
option macaddr 'xx:xx:xx:xx:xx:xx'
This would result in the proper MAC addresses being set for the VLAN
subinterfaces, but the parent interfaces would still have a random MAC
address. Using untagged VLANs could solve this, but would still leave
those extra snippets in /etc/config/network, and then the device VLAN
setup would differ from the one used in ar71xx. Therefore, the MAC
addresses of the ethernet interfaces are being set via preinit instead.
The bdcfg partition contains 4 MAC address labels:
- lanmac
- wanmac
- wlanmac
- wlanmac_a
The first 3 all contain the same MAC address, which is also the one on
the label.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be> Reviewed-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
IPKG_INSTROOT is only set under image builder and we won't be running
this script at build time either, so remove the reference before it gets
cargo-culted into other scripts.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk> Acked-by: Hans Dedecker <dedeckeh@gmail.com>
But the driver in U-Boot of RT-N56U GPL tar blocks using RGMII_33V (9)
mode and it seems to be unsupported on RTL8367B, so drop it from
switch-case in rtl8367b_extif_set_mode.
Paul Spooren [Tue, 9 Jun 2020 04:15:05 +0000 (18:15 -1000)]
imagebuilder: Remove json_info_files/ before build
The folder `json_info_files` contains multiple JSON files which describe
created firmware images. The folder is not removed between builds as the
ImageBuilder does not use `image.mk`.
Not removing the JSON files result in a merged `profiles.json` file
containing entries for outdated or non-existing images.
This commit adds the `json_info_files/` cleanup step to the ImageBuilder
Makefile.
Paul Spooren [Wed, 10 Jun 2020 02:24:29 +0000 (22:24 -0400)]
imagebuilder: pass IB=1 on checking requirements
The patch 4a1a58a3 build, imagebuilder: Do not require libncurses-dev
was supposed to remove libncurses as a requirement for the ImageBuilder.
However as the IB=1 is only exported during building, not for checking
requirements, it did never actually work.
Strangely, the owner of a TRENDnet TEW-810DR reports that the
initial port order is correct, while both devices share the
same board and look similar from the outside. Since I cannot
investigate this without having any of the devices, this does
only touch the DIR-810L for now.
While at it, also merge in the case for zbtlink,zbt-we2026, as
the display port specified for WAN there won't have any effect
anyway.
Reported-by: Roger Pueyo Centelles <roger.pueyo@guifi.net> Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
ramips: fix WAN LED for D-Link DIR-810L/TRENDnet TEW-810DR
The WAN LED on DIR-810L was actually blinking on LAN1 port
activity. This has already been improved for the TEW-810DR, where
the GPIO has been set up explicitly rather than having it controlled
by the switch.
This patch also applies this setup to the DIR-810L.
In addition, the trigger in 01_leds is set up with
ucidef_set_led_switch for both devices now, so state changes should
be displayed correctly as well.
Reported-by: Roger Pueyo Centelles <roger.pueyo@guifi.net> Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de> Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net> [DIR-810L] Tested-by: J. Scott Heppler <shep971@centurylink.net> [TEW-810DR]
With several subtargets, the image/Makefile becomes crowded after a
while. Many targets have moved their device definitions to $subtarget.mk
files to have them more organized, let's do this here as well.
While at it, also move subtarget-specific build recipes.
Cc: Christian Lamparter <chunkeey@gmail.com> Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de> Acked-by: Christian Lamparter <chunkeey@gmail.com>
Renaud Lepage [Sun, 10 May 2020 22:10:20 +0000 (15:10 -0700)]
ath79: add support for the Netgear WNDRMAC v1
The Netgear WNDRMAC v1 is a hardware variant of the Netgear WNDR3700 v2
Specifications
==============
* SoC: Atheros AR7161
* RAM: 64mb
* Flash on board: 16mb
* WiFi: Atheros AR9220 (a/n), Atheros AR9223 (b/g/n)
* Ethernet: RealTek RTL8366SR (1xWAN, 4xLAN, Gigabit)
* Power: 12 VDC, 2.5 A
* Full specs on [openwrt.org](https://openwrt.org/toh/hwdata/netgear/netgear_wndrmac_v1)
Flash Instructions
==================
It is possible to use the OEM Upgrade page to install the `factory`
variant of the firmware.
After the initial upgrade, you will need to telnet into the router
(default IP 192.168.1.1) to install anything. You may install LuCI
this way. At this point, you will have a web interface to configure
OpenWRT on the WNDRMAC v1.
Please use the `sysupgrade` variant for subsequent flashes.
Recovery Instructions
=====================
A TFTP-based recovery flash is possible if the need arises. Please refer
to the WNDR3700 page on openwrt.org for details.
Flash Instructions
==================
It is possible to use the OEM Upgrade page to install the `factory`
variant of the firmware.
After the initial upgrade, you will need to telnet into the router
(default IP 192.168.1.1) to install anything. You may install LuCI
this way. At this point, you will have a web interface to configure
OpenWRT on the WNDRMAC v2.
Please use the `sysupgrade` variant for subsequent flashes.
Recovery Instructions
=====================
A TFTP-based recovery flash is possible if the need arises. Please refer
to the WNDR3800 page on openwrt.org for details.
ath79: wndr3700 series: fix wifi range & throughput
This patch adds ar71xx's GPIO setup for the 2.4GHz and 5GHz antennae
demultiplexer:
| 158 /* 2.4 GHz uses the first fixed antenna group (1, 0, 1, 0) */
| 159 ap9x_pci_setup_wmac_gpio(0, (0xf << 6), (0xa << 6));
| 160
| 161 /* 5 GHz uses the second fixed antenna group (0, 1, 1, 0) */
| 162 ap9x_pci_setup_wmac_gpio(1, (0xf << 6), (0x6 << 6));
This should restore the range and throughput of the 2.4GHz radio
on all the derived wndr3700 variants and versions with the AR7161 SoC.
A special case is the 5GHz radio. The original wndr3700(v1) will
benefit from this change. However the wndr3700v2 and later revisions
were unaffected by the missing bits, as there is no demultiplexer
present in the later designs.
This patch uses gpio-hogs within the device-tree for all
wndr3700/wndr3800/wndrmac variants.
Notes:
Based on the PCB pictures, the WNDR3700(v1) really had eight
independent antennae. Four antennae for each radio and all of
those were printed on the circut board.
The WNDR3700v2 and later have just six antennae. Four of those
are printed on the circuit board and serve the 2.4GHz radio.
Whereas the remaining two are special 5GHz Rayspan Patch Antennae
which are directly connected to the 5GHz radio.
Hannu Nyman dug pretty deep and unearthed a treasure of information
regarding the history of how these values came to be in the OpenWrt
archives: <https://dev.archive.openwrt.org/ticket/6533.html>.
Mark Mentovai came across the fixed antenna group when he was looking
into the driver:
Fixes: FS#3088 Reported-by: Luca Bensi Reported-by: Maciej Mazur Reported-by: Hannu Nyman <hannu.nyman@iki.fi> Debugged-by: Hannu Nyman <hannu.nyman@iki.fi> Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
This patch updates the ca-certificates and ca-bundle package.
This version changed the files directory again, to work/, so
PKG_BUILD_DIR was brought back.
A list of changes from Debian's change-log entry for 20200601 [0]:
* mozilla/{certdata.txt,nssckbi.h}:
Update Mozilla certificate authority bundle to version 2.40. Closes: #956411, #955038
* mozilla/blacklist.txt
Add distrusted Symantec CA list to blacklist for explicit removal. Closes: #911289
Blacklist expired root certificate, "AddTrust External Root" Closes: #961907
The following certificate authorities were added (+):
+ "Certigna Root CA"
+ "emSign ECC Root CA - C3"
+ "emSign ECC Root CA - G3"
+ "emSign Root CA - C1"
+ "emSign Root CA - G1"
+ "Entrust Root Certification Authority - G4"
+ "GTS Root R1"
+ "GTS Root R2"
+ "GTS Root R3"
+ "GTS Root R4"
+ "Hongkong Post Root CA 3"
+ "UCA Extended Validation Root"
+ "UCA Global G2 Root"
The following certificate authorities were removed (-):
- "AddTrust External Root"
- "Certinomis - Root CA"
- "Certplus Class 2 Primary CA"
- "Deutsche Telekom Root CA 2"
- "GeoTrust Global CA"
- "GeoTrust Primary Certification Authority"
- "GeoTrust Primary Certification Authority - G2"
- "GeoTrust Primary Certification Authority - G3"
- "GeoTrust Universal CA"
- "thawte Primary Root CA"
- "thawte Primary Root CA - G2"
- "thawte Primary Root CA - G3"
- "VeriSign Class 3 Public Primary Certification Authority - G4"
- "VeriSign Class 3 Public Primary Certification Authority - G5"
- "VeriSign Universal Root Certification Authority"
Bump to latest Git and refresh all patches in order to get fix for "UPnP
SUBSCRIBE misbehavior in hostapd WPS AP" (CVE-2020-12695).
General security vulnerability in the way the callback URLs in the UPnP
SUBSCRIBE command are used were reported (VU#339275, CVE-2020-12695).
Some of the described issues may be applicable to the use of UPnP in WPS
AP mode functionality for supporting external registrars.
Ref: https://w1.fi/security/2020-1/ Signed-off-by: Petr Štetiar <ynezz@true.cz>
Perry Melange [Sun, 26 Apr 2020 08:51:33 +0000 (10:51 +0200)]
ramips: erx and erx-sfp: fix missing WAN interface
This partially reverts commit 5acd1ed0be0d ("ramips: mt7621: fix
Ubiquiti ER-X ports names and MAC addresses"), this change was discussed
in https://github.com/openwrt/openwrt/pull/2901#discussion_r407238452
With commit 5acd1ed0be0d ("ramips: mt7621: fix Ubiquiti ER-X ports names
and MAC addresses"), all the ports were put into the LAN bridge, with
the argument that the OEM firmware does not have a WAN port enabled. In
the default OEM setup, all of the ports except eth0 are dead and eth0 is
set to a static IP address without providing DHCP services when
connected. It is only after the wizard has been run that eth0 becomes
the WAN port and all the rest of the ports belong to LAN with DHCP
enabled.
Having all of the ports set to the LAN bridge does not mirror the default
OEM setup. To accomplish that, then only eth0 would be in the LAN bridge.
But this is not the expected behaviour of OpenWrt.
Therefore this proposal to set eth0 to WAN and eth1-N to LAN provides
the expected behaviour expected from OpenWrt, maintains the current
documentation as up-to-date, and does not require the user to manually
detach eth0 from the LAN bridge, create the WAN(6) interface(s), and set
eth0 to the WAN(6) interface(s).
Fixes: 5acd1ed0be0d ("ramips: mt7621: fix Ubiquiti ER-X ports names and MAC addresses") Signed-off-by: Perry Melange <isprotejesvalkata@gmail.com>
[commit subject and description tweaks] Signed-off-by: Petr Štetiar <ynezz@true.cz>
Joseph C. Lehner [Fri, 22 Mar 2019 14:19:54 +0000 (15:19 +0100)]
mkchkimg: use higher version code
This patch changes the version code of the image header
from `1.1.99_0.0.0.0` to `99.99.99_99.99.99.99`. This
is neccessary on some devices where the stock firmware
checks the version field, possibly preventing third-party
firmware from being installed.
Reviewed-by: Thibaut VARÈNE <hacks@slashdirt.org> Signed-off-by: Joseph C. Lehner <joseph.c.lehner@gmail.com>
Stijn Tintel [Mon, 8 Jun 2020 03:03:05 +0000 (06:03 +0300)]
hostapd: silence rm
When bringing up wifi the first time after boot, these warnings appear:
netifd: radio0 (1370): rm: can't remove '/var/run/hostapd-wlan0.psk': No such file or directory
netifd: radio0 (1370): rm: can't remove '/var/run/hostapd-wlan0.vlan': No such file or directory
Silence them by adding the "-f" option to rm.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be> Acked-by: John Crispin <john@phrozen.org>
Flash instruction, web UI:
1. Reset to defaults using the reset button if the admin password is
unknown
2. Login into the web UI as admin.
Address: http://192.168.0.1
User: admin
Password: VF-ESVodafone-H-500-s or l033i-h500s
3. Go to Settings -> Firmware Update, and select the Openwrt factory
firmware
4. Update the firmware.
5. Wait until it finish, the device will reboot with Openwrt installed
on the alternative image partitions keeping the stock firmware in
the former.
Notes:
- The patch also adds support for the lowi version. Only the factory
firmware is different.
- The integrated Wifi in the Broadcom Soc isn't still supported.
- The Quantenna 802.11ac wifi works ok, but needs to be configured with
the Quantenna client application. It can't be configured with Luci
nor any iw command since it's a separated subsystem linked via
ethernet.
- The BCM53134S external switch is managed via MDIO which isn't
supported in this target. Therefore it will behave as a dumb switch.
Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
bcm63xx: image: don't add the CFE to the sercomm factory
There is no need to include the CFE bootloader in the Sercomm factory
images.
There might be a case when this could be useful:
- We are running the stock firmware on the first Sercomm image
- The second partition storing the botloader was erased (unlikely)
Even in this case flashing an image without a bootlader is harmless.
Don't include the bootloader in the factory image creation and rid of the
risk of flashing factory images with an untested bootloader partition.
Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
The Raspberry Pi firmware recently switched to building from the common
firmware branch. This introduces changes in the core clock handling,
causing various issues.
E.g. enable_uart=1 no longer fixes the core clock frequency to 250MHz.
When the disable-bt DT overlay is not loaded, the core clock frequency
is increased to 400MHz. As a result, the UART baud rate is no longer
correct, and this causes garbled serial console, or communication
problems with HATs that use the UART.
As a workaround, the core clock could be fixed to 250MHz by adding
'core_freq=250' in /boot/config.txt, but as there appear to be other
issues than just the UART being broken, the safer bet is to revert the
firmware for now.
As reported in https://github.com/openwrt/packages/issues/12072, the
imagebuilder fails due to a dependency resolution error when the userspace
packages are built using a target that has a different kernel version than
that which is being run. To resolve this, add a virtual kernel package with
the conditional dependency currently used in sqm-scripts. The idea is to
move the sqm-scripts dependency to this virtual package, which hopefully
should be consistent with the actual kernel module being built.
kernel: sch_cake: use skb hash improve wireguard compatibility
While the other fq-based qdiscs take advantage of skb->hash and doesn't
recompute it if it is already set, sch_cake does not.
This was a deliberate choice because sch_cake hashes various parts of the
packet header to support its advanced flow isolation modes. However,
foregoing the use of skb->hash entirely loses a few important benefits:
- When skb->hash is set by hardware, a few CPU cycles can be saved by not
hashing again in software.
- Tunnel encapsulations will generally preserve the value of skb->hash from
before the encapsulation, which allows flow-based qdiscs to distinguish
between flows even though the outer packet header no longer has flow
information.
It turns out that we can preserve these desirable properties in many cases,
while still supporting the advanced flow isolation properties of sch_cake.
This patch does so by reusing the skb->hash value as the flow_hash part of
the hashing procedure in cake_hash() only in the following conditions:
- If the skb->hash is marked as covering the flow headers (skb->l4_hash is
set)
AND
- NAT header rewriting is either disabled, or did not change any values
used for hashing. The latter is important to match local-origin packets
such as those of a tunnel endpoint.
The immediate motivation for fixing this was the recent patch to WireGuard
to preserve the skb->hash on encapsulation. As such, this is also what I
tested against; with this patch, added latency under load for competing
flows drops from ~8 ms to sub-1ms on an RRUL test over a WireGuard tunnel
going through a virtual link shaped to 1Gbps using sch_cake. This matches
the results we saw with a similar setup using sch_fq_codel when testing the
WireGuard patch.
Fixes: 046f6fd5daef ("sched: Add Common Applications Kept Enhanced (cake) qdisc") Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
DENG Qingfang [Thu, 4 Jun 2020 04:30:19 +0000 (12:30 +0800)]
mvebu: remove ClearFog Pro SUPPORTED_DEVICES
A direct upgrade from previous swconfig version with
incompatible settings to DSA will break the internet.
Remove SUPPORTED_DEVICES so users cannot upgrade directly.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
[rebase after Linksys rename, adjust title] Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Paul Spooren [Fri, 10 Apr 2020 10:14:21 +0000 (00:14 -1000)]
mvebu: rename Linksys devices based on their common names
The Linksys devices in mvebu target feature a mixed naming,
where parts are based on the official product name (device
node, image; e.g. WRT3200ACM) and parts are based on the
internal code name (DTS file name, compatible, LED labels;
e.g. rango). This inconsistent naming has been perceived
as quite confusing.
A recent attempt by Paul Spooren to harmonize this naming
in kernel has been declined there. However, for us it still
makes sense to apply at least a part of these changes
locally.
Primarily, this patch changes the compatible in DTS and thus
the board name used in various scripts to have them in line
with the device, model and image names. Due to the recent
switch from swconfig to DSA, this allows us to drop
SUPPORTED_DEVICES and thus prevent seamless upgrade between
these incompatible setups.
However, this does not include the LED label rename from
Paul's initial patch: I don't think it's worth keeping the
enormous diff locally for this case, as we can implement
this much easier in 01_leds if we have to live with the
inconsistency anyway.
Signed-off-by: Paul Spooren <mail@aparcar.org>
[rebase, extend to all devices, drop DT LED changes] Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
DENG Qingfang [Sat, 25 Apr 2020 12:08:00 +0000 (20:08 +0800)]
generic: mt7530: fix roaming from DSA user ports
When a client moves from a DSA user port to a software port in a bridge,
it cannot reach any other clients that connected to the DSA user ports.
That is because SA learning on the CPU port is disabled, so the switch
ignores the client's frames from the CPU port and still thinks it is at
the user port.
Fix it by enabling SA learning on the CPU port.
To prevent the switch from learning from flooding frames from the CPU
port, set skb->offload_fwd_mark to 1 for unicast and broadcast frames,
and let the switch flood them instead of trapping to the CPU port.
Multicast frames still need to be trapped to the CPU port for snooping,
so set the SA_DIS bit of the MTK tag to 1 when transmitting those frames
to disable SA learning.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
DENG Qingfang [Mon, 20 Apr 2020 13:37:17 +0000 (21:37 +0800)]
generic: fix DSA VLAN filtering
Currently enabling VLAN filtering blocks all traffic in the bridge
immediately. That is because DSA ignores all VLAN setup when VLAN
filtering is disabled, and when it is enabled, there is no VLAN entry
in the VLAN table, causing all traffic to be blocked.
Add patches to allow VLAN setup even if VLAN filtering is disabled.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
DENG Qingfang [Thu, 14 May 2020 03:00:58 +0000 (11:00 +0800)]
generic: mt7530: set CPU port to fallback mode
Currently, setting a bridge's self PVID to other value and deleting
the default VID 1 renders untagged ports of that VLAN unable to talk to
the CPU port:
bridge vlan add dev br0 vid 2 pvid untagged self
bridge vlan del dev br0 vid 1 self
bridge vlan add dev sw0p0 vid 2 pvid untagged
bridge vlan del dev sw0p0 vid 1
# br0 cannot send untagged frames out of sw0p0 anymore
That is because the CPU port is set to security mode and its PVID is
still 1, and untagged frames are dropped due to VLAN member violation.
Set the CPU port to fallback mode so untagged frames can pass through.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
projects / thirdparty / openwrt.git / log
