5 years agotarget/sdk: Switch to xz compression instead of bz2
Daniel Engberg [Tue, 11 Oct 2016 11:51:44 +0000 (13:51 +0200)]
target/sdk: Switch to xz compression instead of bz2

Switch to xz compression instead of using bz2.
This makes a considerable difference in size, ar71xx SDK from 60M to 35M.

Signed-off-by: Daniel Engberg <>
5 years agoapm821xx: add back end-of-UBI marker for the WNDR4700 and MR24
Christian Lamparter [Thu, 3 Nov 2016 20:47:06 +0000 (21:47 +0100)]
apm821xx: add back end-of-UBI marker for the WNDR4700 and MR24

The patch "build: drop UBI EOF marker from images by default"
(commit d27bce8d28eb129af0abd9c80a7756301b7d588a) removed the
end-of-UBI marker. Without this marker, the boards will panic
during boot:

[    1.683458] ubi0: attaching mtd4
[    1.696181] ubi0 warning: scan_peb: valid VID header but corrupted EC header at PEB 31
[    1.704367] ubi0 error: scan_peb: bad image sequence number 549886691 in PEB 32, expected 184585623
[    1.713377] Erase counter header dump:
[    1.717110]  magic          0x55424923
[    1.720843]  version        1
[    1.723797]  ec             0
[    1.726752]  vid_hdr_offset 512
[    1.729880]  data_offset    2048
[    1.733094]  image_seq      549886691
[    1.736740]  hdr_crc        0x92ba8130
[    1.740472] erase counter header hexdump:
[    1.744493] ubi0 error: ubi_attach_mtd_dev: failed to attach mtd4, error -22
[    1.751528] UBI error: cannot attach mtd4
[    1.755373] hctosys: unable to open rtc device (rtc0)
[    1.761130] VFS: Cannot open root device "(null)" or unknown-block(0,0): error -6
[    1.768604] Please append a correct "root=" boot option; here are the available partitions:
[    1.856992] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)
[    1.866519] Rebooting in 1 seconds..Auto calibration ---

Signed-off-by: Christian Lamparter <>
5 years agomt76: update to the latest version, adds a tx queue configuration fix
Felix Fietkau [Thu, 3 Nov 2016 11:37:00 +0000 (12:37 +0100)]
mt76: update to the latest version, adds a tx queue configuration fix

Signed-off-by: Felix Fietkau <>
5 years agomac80211: fix A-MSDU tx aggregation (FS#174)
Felix Fietkau [Fri, 4 Nov 2016 09:32:05 +0000 (10:32 +0100)]
mac80211: fix A-MSDU tx aggregation (FS#174)

Signed-off-by: Felix Fietkau <>
5 years agomac80211: minor cleanup
Felix Fietkau [Fri, 4 Nov 2016 09:31:42 +0000 (10:31 +0100)]
mac80211: minor cleanup

Signed-off-by: Felix Fietkau <>
5 years agomac80211: fix a tx A-MPDU aggregation issue
Felix Fietkau [Fri, 4 Nov 2016 09:31:27 +0000 (10:31 +0100)]
mac80211: fix a tx A-MPDU aggregation issue

Signed-off-by: Felix Fietkau <>
5 years agolantiq: use external pci clock on ARV7506PW11
Mathias Kresin [Wed, 2 Nov 2016 17:14:11 +0000 (18:14 +0100)]
lantiq: use external pci clock on ARV7506PW11

Fixes a ralink wireless driver related kernel crash as soon as a lot of
data is tranfered via wireless.

Signed-off-by: Mathias Kresin <>
5 years agolantiq: cleanup dts files
Mathias Kresin [Wed, 2 Nov 2016 18:49:04 +0000 (19:49 +0100)]
lantiq: cleanup dts files

- remove not existing properties
- remove properties having the same values as the included dtsi
- remove nodes which are disabled in the included dtsi and not enabled
  in dts
- replace the deprecated pinctrl-* compatible strings
- use the same labels for nodes as the included dtsi
- move common used vr9 pci properties to vr9.dtsi
- remove the unused stp node from HomeHub 2B devcie tree source file
- fix spaces vs. tabs and remove superfluous linebreaks

Signed-off-by: Mathias Kresin <>
5 years agomac80211: replace the previous fix with a revert of the faulty upstream commit
Felix Fietkau [Thu, 3 Nov 2016 11:13:32 +0000 (12:13 +0100)]
mac80211: replace the previous fix with a revert of the faulty upstream commit

This avoids potential breakage in rt2x00usb

Signed-off-by: Felix Fietkau <>
5 years agoelfutils: bump to 0.167
Luiz Angelo Daros de Luca [Fri, 14 Oct 2016 00:02:47 +0000 (21:02 -0300)]
elfutils: bump to 0.167

Signed-off-by: Luiz Angelo Daros de Luca <>
5 years agomwl8k: remove synchronous device init hack
Felix Fietkau [Tue, 18 Oct 2016 09:33:09 +0000 (11:33 +0200)]
mwl8k: remove synchronous device init hack

Signed-off-by: Felix Fietkau <>
5 years agomac80211: remove ath10k delayed initialization hack
Mathias Kresin [Mon, 17 Oct 2016 13:59:15 +0000 (15:59 +0200)]
mac80211: remove ath10k delayed initialization hack

LEDE no longer requires all PHYs to be initialized to
create the configuration files during bootup. This patch
removes the now obsolete ath10k patch.

Signed-off-by: Mathias Kresin <>
Signed-off-by: Christian Lamparter <>
5 years agobase-files, mac80211, broadcom-wl: wifi detection and configuration
Christian Lamparter [Mon, 17 Oct 2016 13:59:14 +0000 (15:59 +0200)]
base-files, mac80211, broadcom-wl: wifi detection and configuration

Currently, the wifi detection script is executed as part of
the (early) boot process. Pluggable wifi USB devices, which
are inserted at a later time are not automatically
detected and therefore they don't show up in LuCI.

A user has to deal with wifi detection manually, or restart
the router.

However, the current "sleep 1" window - which the boot
process waits for wifi devices to "settle down" - is too
short to detect wifi devices for some routers anyway.

For example, this can happen with USB WLAN devices on the
WNDR4700. This is because the usb controller needs to load
its firmware from UBI and initialize, before it can operate.

The issue can be seen on a BT HomeHub 5A as well as soon as
the caldata are on an ubi volume. This is because the ath9k
card has to be initialized by owl-loader first. Which has to
wait for the firmware extraction script to retrieve the pci
initialization values inside the caldata.

This patch moves the wifi configuration to hotplug scripts.
For mac80211, the wifi configuration will now automatically
run any time a "ieee80211" device is added. Likewise
broadcom-wl's script checks for new "net" devices which
have the "wl$NUMBER" moniker.

Issues with spawning multiple interface configuration - in
case the detection script is run concurrently - have been
resolved by using a named section for the initial
configuration. Concurrent configuration scripts will now
simply overwrite the same existing configuration.

A workaround which preserves the "sleep 1" window for just
the first boot has been added. This allows the existing
brcm47xx boot and mvebu uci-default scripts to correctly
setup the initial mac addresses and regulatory domain.

And finally, the patch renames the "wifi detect" into
"wifi config". As the script no longer produces any output
that has to be redirected or appended to the configuration

Thanks to Martin Blumenstingl for helping with the implementation
and testing of the patch.

Acked-by: Jo-Philipp Wich <>
Signed-off-by: Mathias Kresin <>
Signed-off-by: Christian Lamparter <>
Signed-off-by: Felix Fietkau <>
5 years agobase-files, mac80211, broadcom-wl: use uci to populate wireless config
Christian Lamparter [Mon, 17 Oct 2016 13:59:13 +0000 (15:59 +0200)]
base-files, mac80211, broadcom-wl: use uci to populate wireless config

Previously, wifi detect simply dumped its generated wireless
configuration to STDOUT. A second step was needed to append
the configuration to /etc/config/wireless (or create it, if
it didn't exist).

With this patch, The wifi detection script will now use uci
to update the wireless configuration directly.

This patch also makes the initially created wifi-iface a
named section ('default_radio$X' for mac80211 and
'default_wl$X' for broadcom). With this change, uci will
not print the cfgHASH to STDOUT (which would now corrupt
the wireless configuration). It will also prevent adding
duplicated wifi interface configurations, if the wifi
configuration is run concurrently.

Signed-off-by: Christian Lamparter <>
5 years agobase-files: generate /etc/config/wireless, if it doesn't exist
Mathias Kresin [Mon, 17 Oct 2016 13:59:12 +0000 (15:59 +0200)]
base-files: generate /etc/config/wireless, if it doesn't exist

This patch adds a check in "wifi detect" to test if the wireless
configuration file does exist. If it doesn't exist, an empty
/etc/config/wireless file will be created.

This is necessary because uci doesn't create new files,
instead the tool just exits with "uci: Entry not found".

Signed-off-by: Mathias Kresin <>
Signed-off-by: Christian Lamparter <>
5 years agomac80211: fix regdomain change issues with CONFIG_ATH_USER_REGD
Felix Fietkau [Mon, 24 Oct 2016 06:58:46 +0000 (08:58 +0200)]
mac80211: fix regdomain change issues with CONFIG_ATH_USER_REGD

Signed-off-by: Felix Fietkau <>
5 years agomac80211: fix a minor issue in the header padding patch
Felix Fietkau [Wed, 2 Nov 2016 10:43:12 +0000 (11:43 +0100)]
mac80211: fix a minor issue in the header padding patch

Signed-off-by: Felix Fietkau <>
5 years agomac80211: fix AP powersave issues introduced in the last wireless-testing update...
Felix Fietkau [Thu, 3 Nov 2016 10:01:45 +0000 (11:01 +0100)]
mac80211: fix AP powersave issues introduced in the last wireless-testing update (FS#241)

Signed-off-by: Felix Fietkau <>
5 years agoscripts/feeds: use 10 chars for feed name column width
Rafał Miłecki [Wed, 2 Nov 2016 14:02:53 +0000 (15:02 +0100)]
scripts/feeds: use 10 chars for feed name column width

It's always hard to find a reasonable width that will make everyone
happy. This one at least makes "telephony" (one of default feeds) name
fit the column and hopefully isn't too big.

Signed-off-by: Rafał Miłecki <>
5 years agoipq806x: update DT in accordance to new drivers And add some more DT nodes
Pavel Kubelun [Wed, 2 Nov 2016 19:42:55 +0000 (22:42 +0300)]
ipq806x: update DT in accordance to new drivers And add some more DT nodes

Signed-off-by: Pavel Kubelun <>
5 years agoipq806x: switch to generic cpufreq driver cpufreq-dt
Pavel Kubelun [Wed, 2 Nov 2016 19:37:28 +0000 (22:37 +0300)]
ipq806x: switch to generic cpufreq driver cpufreq-dt

This fixes ondemand frequency scaling and moves ipq806x onto upstream driver

Also switching to ondemand frequency scaling as it is fixed now

Signed-off-by: Pavel Kubelun <>
5 years agoipq806x: add opp patches for voltage scaling
Pavel Kubelun [Wed, 2 Nov 2016 19:32:36 +0000 (22:32 +0300)]
ipq806x: add opp patches for voltage scaling

Preparing for cpufreq driver switch to generic cpufreq-dt

Signed-off-by: Pavel Kubelun <>
5 years agoipq806x: update clk-qcom patches
Pavel Kubelun [Wed, 2 Nov 2016 19:08:45 +0000 (22:08 +0300)]
ipq806x: update clk-qcom patches

Preparing for cpufreq driver switch to generic cpufreq-dt

Signed-off-by: Pavel Kubelun <>
5 years agoipq806x: fix leds, sata port for Netgear R7800 and minor fixes - renamed leds in...
Pavel Kubelun [Sat, 1 Oct 2016 15:32:58 +0000 (11:32 -0400)]
ipq806x: fix leds, sata port for Netgear R7800 and minor fixes - renamed leds in correct color accordance - blink power led with white during boot and with amber when flashing firmware - fixed usb leds - enabled unused wps and rfkill leds as wlan leds. Now rfkill led is for 2.4GHz and wps - 5GHz WIFI - removed unneeded bootargs - removed unneeded pci pins from R7800 DT (driver already handles it in proper way) and add tx offsetting - nand ecc step size - fixed sata ports

Signed-off-by: Pavel Kubelun <>
5 years agoramips: Use MT7621 I2C driver for MT7628/MT7688
Sven Schwermer [Wed, 2 Nov 2016 10:44:23 +0000 (11:44 +0100)]
ramips: Use MT7621 I2C driver for MT7628/MT7688

The i2c-ramips driver does not work for the MT7628/MT7688 SoCs, the
i2c-mt7621 driver does.

Signed-off-by: Sven Schwermer <>
5 years agouboot-envtools: move to Boot Loaders submenu
Alberto Bursi [Wed, 2 Nov 2016 19:33:01 +0000 (20:33 +0100)]
uboot-envtools: move to Boot Loaders submenu

Boot Loaders submenu is where other bootloader tools (fconfig and grub2-editenv) are.

Signed-off-by: Alberto Bursi <>
5 years agoipq806x: TP-Link Archer C2600: convert old usbdev trigger to new usbport
Cezary Jackiewicz [Wed, 2 Nov 2016 17:11:33 +0000 (18:11 +0100)]
ipq806x: TP-Link Archer C2600: convert old usbdev trigger to new usbport

Signed-off-by: Cezary Jackiewicz <>
5 years agoramips: add support for MikroTik hEX v3 (RB750Gr3)
Andrew Yong [Tue, 18 Oct 2016 20:05:32 +0000 (04:05 +0800)]
ramips: add support for MikroTik hEX v3 (RB750Gr3)

The MikroTik hEX v3 (RB750Gr3) is a MT7621AT board which is similar to most MT7621 reference designs, it can be easily supported by this patch; however, the stock RouterBOOT bootloader has to be replaced by a MT7621 SDK U-Boot such as - U-Boot configured for the RB750Gr3 (16MiB SPI flash, 256MiB DDR3 RAM at 1200MHz).

RouterBOOT, the stock bootloader, does not initialize the UART and boots silently, making it preferable to replace it with a MT7621 SDK U-Boot with UART (57600 8N1) that supports HTTP, TFTP or serial upload of sysupgrade firmware and U-Boot.

Furthermore, RouterOS, the stock firmware, is contained in a proprietary modification of SquashFS without GPL sources; UART is also disabled in stock firmware.

The combination of LEDE firmware generated by this PR and MT7621 SDK U-Boot expects the printed MAC address to reside at offset `0xe000` of the factory partition (absolute offset is `0x4e000`); this is similar to the factory MAC address offset for several other MT7621 devices.

A 16MiB flash dump suitable for use with flashrom will be provided if/once this patch is accepted and binaries are built by LEDE buildbot. Alternatively, writing the U-Boot to the SPI flash starting at 0x0 offset and booting the board with serial console attached will allow TFTP, HTTP or serial upload of sysupgrade firmware.

Signed-off-by: Andrew Yong <>
5 years agodnsmasq: support log-dhcp option
Karl Palsson [Mon, 31 Oct 2016 14:00:06 +0000 (14:00 +0000)]
dnsmasq: support log-dhcp option

Helpful when trying to resolve issues with quirky dhcp client devices.

Signed-off-by: Karl Palsson <>
5 years agokernel: update kernel 4.4 to version 4.4.30
Stijn Segers [Tue, 1 Nov 2016 11:21:15 +0000 (12:21 +0100)]
kernel: update kernel 4.4 to version 4.4.30

This patch bumps the 4.4 kernel from .28 to .30 and refreshes the patches.
Compile-tested on ar71xx, x86/64, ramips/mt7621, brcm47xx and kirkwood.

Run-tested on ar71xx & ramips/mt7621, brcm47xx and kirkwood (last two confirmed
by P. Wassi).

Signed-off-by: Stijn Segers <>
5 years agoramips: add usb packages into DIR-860L B1 profile
Stijn Segers [Mon, 31 Oct 2016 21:23:32 +0000 (22:23 +0100)]
ramips: add usb packages into DIR-860L B1 profile

This patch adds the kmod-usb3 and kmod-usb-ledtrig-usbport packages to the
DIR-860L B1 profile. The DIR-860L B1 has a USB 3 port.

Signed-off-by: Stijn Segers <>
5 years agoiproute2: rename ip to ip-tiny and let both ip-tiny and ip-full provide "ip"
Jo-Philipp Wich [Thu, 13 Oct 2016 01:07:21 +0000 (03:07 +0200)]
iproute2: rename ip to ip-tiny and let both ip-tiny and ip-full provide "ip"

Rename the "ip" package declaration to "ip-tiny" and let both "ip-tiny" and
"ip-full" provide the virtual "ip" package. This allows users to freely choose
the "ip" command variant while other packages can continue to depend on "ip"
without needing to enforce a specific variant.

Note that this commit does not add busybox as "ip" provider due to
the following reasons:

 - The builtin Busybox ip applet cannot be added or removed at runtime
 - Both "ip-tiny" and "ip-full" are able to install without file clashes even
   if the busybox applet is enabled
 - The system is preferring full "ip-tiny" and "ip-full" at runtime, even
   if Busybox ip is still present.

Signed-off-by: Jo-Philipp Wich <>
5 years agouboot-sunxi: fix default config for OLIMEX A13 SOM (FS#239)
Jo-Philipp Wich [Mon, 31 Oct 2016 19:19:27 +0000 (20:19 +0100)]
uboot-sunxi: fix default config for OLIMEX A13 SOM (FS#239)

The current uboot default config for the A13 SOM erroneously enables support
for the AXP209 power regulator IC which is not present on the board.

This superfluous support module sets an incorrect initial clock frequency and
confuses the kernel, ultimately leading to a boot failure later on.

Properly disable the PMIC support and enable the EHCI support by translating
the deprecated SYS_EXTRA_OPTIONS values into proper SUNXI_NO_PMIC and
USB_EHCI_HCD symbols respectively.

Also rename 002-add-olimex-a13-som.diff to 002-add-olimex-a13-som.patch and
refresh the remaining patches of the series while we're at it.

Reported-by: Mario Fischer <>
Signed-off-by: Jo-Philipp Wich <>
5 years agoscripts/ honour DEFAULT_VARIANT
Jo-Philipp Wich [Wed, 2 Nov 2016 00:45:06 +0000 (01:45 +0100)]
scripts/ honour DEFAULT_VARIANT

So far, always considered the first provider of a virtual
package to be the default variant which might deviate from what buildroot
considers to be the default.

Change the Kconfig dependency / select code generation for virtual package
providers to consider the DEFAULT_VARIANT to be the primary provider and only
fall back to the first provider if no default variant was explicitely tagged.

Signed-off-by: Jo-Philipp Wich <>
5 years agobase-files: uci-defaults: support requesting untagged switch port configuration
Jo-Philipp Wich [Sat, 15 Oct 2016 23:13:55 +0000 (01:13 +0200)]
base-files: uci-defaults: support requesting untagged switch port configuration

Introduce support for a new [0-9]u@netdev syntax to ucidev_add_switch() to let
board.d network files request default network switch configurations which do
not use tagged CPU ports.

This commit itself has no effect on generated configurations at the moment
since we still emit untagged configurations by default but it allows boards to
opt-out from default tagged configs in case we start emitting tagged settings
by default.

Signed-off-by: Jo-Philipp Wich <>
5 years agoinclude: properly update .install stamp files
Jo-Philipp Wich [Thu, 13 Oct 2016 00:54:34 +0000 (02:54 +0200)]
include: properly update .install stamp files

Right now the $(PKG_INSTALL_STAMP) files are only written if a package is
selected as <*> but never deleted or emptied if the corresponding package
is getting deselected.

For ordinary packages this usually is no problem as the package/install
recipe performs its own check for enabled packages when assembling the
list of install stamp files to consider, but this logic might fail under
certain circumstances for packages providing multiple build variants.

In case of a multi-variant package, the buildroot first checks if any
of the variants is enabled, then resolves all variants of the common
source package and finally processes the corresponding .install stamp
files of all variants, relying on the assumption that only the selected
.install stamp file exists.

When an initially selected variant is getting deselected or changed from
<*> to <m> and another variant is marked as <*> instead, the .install
stamp file of the deselected variant remains unchanged and a second
.install stamp file for the newly selected variant is getting created,
causing the package/install recipe to pick up two .install stamps with
conflicting variants, leading to opkg file clashes.

This issue happens for example if package "ip" is set to <m> and package
"ip-full" to <*> -  the install command will eventually fail with:

     * check_conflicts_for: The following packages conflict with ip:
     * check_conflicts_for:  ip-full *
     * opkg_install_cmd: Cannot install package ip.

In order to fix the problem, always process the removal requests or the
.install stamp files, even for deselected packages but only write the
package base name into the stamp file if the corresponding package is
marked as builtin.

Signed-off-by: Jo-Philipp Wich <>
5 years agoscripts/ fix handling of virtual (PROVIDES) depends
Jo-Philipp Wich [Thu, 13 Oct 2016 00:48:32 +0000 (02:48 +0200)]
scripts/ fix handling of virtual (PROVIDES) depends

Currently the code emitting dependencies for provide candidates is overwriting
the specification calculated by the previous conditional dependency handling
code, rendering dependencies on virtual PROVIDES packages in conjunction with
conditional dependencies unusable.

Instead of overwriting, append the PROVIDES dependency spec in order to fix
using DEPENDS on virtual provider packages in conjunction with conditions.

Signed-off-by: Jo-Philipp Wich <>
5 years agoinclude/ use STAGING_DIR_HOSTPKG
Jo-Philipp Wich [Tue, 1 Nov 2016 11:09:30 +0000 (12:09 +0100)]

Instead of hardcoding $(STAGING_DIR)/host, use the new $(STAGING_DIR_HOSTPKG)
variable to refer to the directory.

Signed-off-by: Jo-Philipp Wich <>
5 years add STAGING_DIR_HOSTPKG variable
Jo-Philipp Wich [Tue, 1 Nov 2016 11:07:56 +0000 (12:07 +0100)] add STAGING_DIR_HOSTPKG variable

Add a STAGING_DIR_HOSTPKG variable which refers to $(STAGING_DIR)/host in order
to prepare support for relocating that directory in the future.

Signed-off-by: Jo-Philipp Wich <>
5 years agofstools: add build-depends on util-linux
Jo-Philipp Wich [Tue, 1 Nov 2016 08:50:39 +0000 (09:50 +0100)]
fstools: add build-depends on util-linux

The libblkid probe support in fstools git head requires blkid/blkid.h for
compilation, so add a build dependency on util-linux which provides libblkid.

Signed-off-by: Jo-Philipp Wich <>
5 years agoar71xx: Add support to Powerline ac TP-Link WPA8630
Henryk Heisig [Sun, 30 Oct 2016 16:57:27 +0000 (17:57 +0100)]
ar71xx: Add support to Powerline ac TP-Link WPA8630

Signed-off-by: Henryk Heisig <>
5 years agolayerscape: add 64b/32b target for ls1012ardb device
Yutang Jiang [Fri, 28 Oct 2016 16:18:23 +0000 (00:18 +0800)]
layerscape: add 64b/32b target for ls1012ardb device

The QorIQ LS1012A processor, optimized for battery-backed or
USB-powered, integrates a single ARM Cortex-A53 core with a hardware
packet forwarding engine and high-speed interfaces to deliver
line-rate networking performance.
QorIQ LS1012A Reference Design System (LS1012ARDB) is a high-performance
development platform, with a complete debugging environment.
The LS1012ARDB board supports the QorIQ LS1012A processor and is
optimized to support the high-bandwidth DDR3L memory and
a full complement of high-speed SerDes ports.

LEDE/OPENWRT will auto strip executable program file while make. So we
need select CONFIG_NO_STRIP=y while make menuconfig to avoid the ppfe network
fiemware be destroyed, then run make to build ls1012ardb firmware.

The fsl-quadspi flash with jffs2 fs is unstable and arise some failed message.
This issue have noticed the IP owner for investigate, hope he can solve it
earlier. So the ls1012ardb now also provide a xx-firmware.ext4.bin as default
firmware, and the uboot bootcmd will run wrtboot_ext4rfs for "rootfstype=ext4"

Signed-off-by: Yutang Jiang <>
5 years agolayerscape: add 64b/32b target for ls1043ardb device
Yutang Jiang [Fri, 28 Oct 2016 16:14:32 +0000 (00:14 +0800)]
layerscape: add 64b/32b target for ls1043ardb device

Add support for NXP layerscape ls1043ardb 64b/32b Dev board.

LS1043a is an SoC with 4x64-bit up to 1.6 GHz ARMv8 A53 cores.
ls1043ardb support features as: 2GB DDR4, 128MB NOR/512MB NAND, USB3.0, eSDHC,
I2C, GPIO, PCIe/Mini-PCIe, 6x1G/1x10G network port, etc.

64b/32b ls1043ardb target is using 4.4 kernel, and rcw/u-boot/fman images from
NXP QorIQ SDK release.

All of 4.4 kernel patches porting from SDK release or upstream.

QorIQ SDK ISOs can be downloaded from this location:

Signed-off-by: Yutang Jiang <>
5 years agoramips: Archer C50 cleanup
Henryk Heisig [Thu, 13 Oct 2016 19:00:06 +0000 (21:00 +0200)]
ramips: Archer C50 cleanup

- setting read-only flag to important partitions
- enabling PA to improve 2.4 GHz signal strength
- add missing leds
- rename colour led
- add mac adress to 5GHz wlan interface
- included <dt-bindings/input/input.h> and <dt-bindings/gpio/gpio.h>

Signed-off-by: Henryk Heisig
5 years agolantiq: add vpe/watchdog modules to kernel
Stefan Koch [Mon, 3 Nov 2014 08:33:04 +0000 (08:33 +0000)]
lantiq: add vpe/watchdog modules to kernel

(required not-distributable firmware blob - dump it by yourself from original firmware)

Signed-off-by: Eddi De Pieri <>
(cherry picked from commit 064f467264c5c9b6eca0bb96b587f9412b770cc5)
Signed-off-by: Stefan Koch <>
5 years agolantiq: added xrx200 as plattform for ltqvmmc
Stefan Koch [Tue, 4 Oct 2016 19:22:40 +0000 (21:22 +0200)]
lantiq: added xrx200 as plattform for ltqvmmc

Signed-off-by: Stefan Koch <>
5 years agolantiq: added xrx200 as plattform for ltqtapi
Stefan Koch [Tue, 4 Oct 2016 19:22:40 +0000 (21:22 +0200)]
lantiq: added xrx200 as plattform for ltqtapi

Signed-off-by: Stefan Koch <>
5 years agolantiq: vpe softdog
Stefan Koch [Mon, 3 Nov 2014 08:32:32 +0000 (08:32 +0000)]
lantiq: vpe softdog

(required not-distributable firmware blob - dump it by yourself from original firmware)

Signed-off-by: Eddi De Pieri <>
(cherry picked from commit eb0ce57270d0b5b81b224b9336cf54707497eede)

Modified after cherry-pick:
obj in Makefile

Signed-off-by: Stefan Koch <>
5 years agolantiq: added support for VPE1
Stefan Koch [Thu, 20 Oct 2016 19:32:02 +0000 (21:32 +0200)]
lantiq: added support for VPE1

Created minimal patchset based on BB rev 43158 by Eddi De Pieri
14.07/openwrt.git 79472c025449efae9310defad0d3a73cff14d756

If the VR9 based router provides FXS ports and they shoud enabled then
the following must added to the kernel command line:
mem=[TOTALMEMSIZE-2M] vpe1_load_addr=ADDRESS vpe1_mem=2M maxvpes=1

To use FXS 2M of RAM are needed for the VPE firmware. The size is set
by vpe1_mem.
The available RAM must be reduced by this size using the mem argument.
A correct load address (example 0x83e00000) for the firmware must be given,

Signed-off-by: Stefan Koch <>
5 years agolantiq: modify vr9.dts to support vmmc
Stefan Koch [Mon, 3 Nov 2014 08:32:58 +0000 (08:32 +0000)]
lantiq: modify vr9.dts to support vmmc

(required not-distributable firmware blob - dump it by yourself from original firmware)

Signed-off-by: Eddi De Pieri <>
(cherry picked from commit 8d924d43c0ea6839a3a33e54982e8da48b736001)

Modified after cherry-pick:
compatible attribute

Signed-off-by: Stefan Koch <>
5 years agolantiq: fix vmmc build
Stefan Koch [Mon, 3 Nov 2014 08:32:50 +0000 (08:32 +0000)]
lantiq: fix vmmc build

(required not-distributable firmware blob - dump it by yourself from original firmware)

Signed-off-by: Eddi De Pieri <>
(cherry picked from commit 47b1ff965b0cb57013b40fbe2bcd7f3c6eb6b606)

Modified after cherry-pick:
FW_MD5SUM in Makefile

Signed-off-by: Stefan Koch <>
5 years agolantiq: ltq-vmmc add support for ar9-vr9
Stefan Koch [Mon, 3 Nov 2014 08:32:30 +0000 (08:32 +0000)]
lantiq: ltq-vmmc add support for ar9-vr9

(required not-distributable firmware blob - dump it by yourself from original firmware)

Signed-off-by: Eddi De Pieri <>
(cherry picked from commit ea9e61b8eb61a2e362a50541f03466dc7d087947)
Signed-off-by: Stefan Koch <>
5 years agoipq806x: add reserved memory node in Netgear R7800
Pavel Kubelun [Tue, 25 Oct 2016 16:46:04 +0000 (19:46 +0300)]
ipq806x: add reserved memory node in Netgear R7800

KERNEL BUG: BAD_PAGE_STATE in process appears here and there during intensive memory usage.

Signed-off-by: Pavel Kubelun <>
5 years agoramips: Add RTC driver to kernel for working hctosys
Rosen Penev [Sun, 30 Oct 2016 03:50:45 +0000 (20:50 -0700)]
ramips: Add RTC driver to kernel for working hctosys

Build the RTC driver into the kernel, (and remove the optional module), in order
to make hctosys working. (Currently the module is loaded after hctosys has failed previously)

Signed-off-by: Rosen Penev <>
5 years agoipq806x: ArcherC2600: export usb power pins
Henryk Heisig [Sat, 29 Oct 2016 20:19:32 +0000 (22:19 +0200)]
ipq806x: ArcherC2600: export usb power pins

Signed-off-by: Henryk Heisig <>
5 years agoipq806x: ArcherC2600: devictree cleanup
Henryk Heisig [Sat, 29 Oct 2016 19:49:09 +0000 (21:49 +0200)]
ipq806x: ArcherC2600: devictree cleanup

add blank lines and use macros for GPIOs

Signed-off-by: Henryk Heisig <>
5 years agofstools: update to latest git HEAD
John Crispin [Mon, 31 Oct 2016 15:38:26 +0000 (16:38 +0100)]
fstools: update to latest git HEAD

adds f2fs support

Signed-off-by: John Crispin <>
5 years agokernel: update kernel 4.4 to version 4.4.28
Paul Wassi [Sat, 29 Oct 2016 07:56:58 +0000 (09:56 +0200)]
kernel: update kernel 4.4 to version 4.4.28

Refresh patches for all targets that support kernel 4.4.
compile/run-tested on ar71xx, brcm47xx, kirkwood.

Signed-off-by: Paul Wassi <>
Tested-by: Koen Vandeputte <>
5 years agobase-files: sysfixtime: Keep RTC time in UTC timezone
Petr Štetiar [Mon, 31 Oct 2016 10:50:09 +0000 (11:50 +0100)]
base-files: sysfixtime: Keep RTC time in UTC timezone

We need to tell hwclock with -u commandline option, that we would like
to keep our RTC clock in UTC timezone. Linux kernel expects RTC in UTC
timezone anyway.

In current state of things, we don't tell hwclock to load/store time
from/to RTC in UTC timezone so it uses the timezone from the system
time. If it's set to different timezone then UTC, sysfixtime is going to
screw the time in RTC.

I've following in the setup script:

  uci set system.@system[0].timezone='CET-1CEST,M3.5.0,M10.5.0/3'
  uci set system.@system[0].zonename='Europe/Prague'

I've this RTC setup (rtc1 is RTC on i.MX6 SoC, rtc0 is battery backed RTC mcp7941x):

  rtc-ds1307 3-006f: rtc core: registered mcp7941x as rtc0
  snvs_rtc 20cc000.snvs:snvs-rtc-lp: rtc core: registered 20cc000.snvs:snvs-r as rtc1

Then we can experience following (current time is 10:15am):

$ date
Fri Oct 21 10:15:07 CEST 2016

$ hwclock -r -f /dev/rtc0
Fri Oct 21 08:14:46 2016  0.000000 seconds

$ hwclock -u -r -f /dev/rtc0
Fri Oct 21 10:14:46 2016  0.000000 seconds

And after current broken sysfixtime:

$ /etc/init.d/sysfixtime stop

$ date
Fri Oct 21 10:15:25 CEST 2016

$ hwclock -r -f /dev/rtc0
Fri Oct 21 10:15:31 2016  0.000000 seconds

Now we've time in our battery backed RTC in CEST timezone instead of
UTC.  Then once again, but with this patch applied to sysfixtime, where
hwclock is using correctly the -u parameter:

$ /etc/init.d/sysfixtime stop

$ date
Fri Oct 21 10:15:53 CEST 2016

$ hwclock -r -f /dev/rtc0
Fri Oct 21 08:15:55 2016  0.000000 seconds

Signed-off-by: Petr Štetiar <>
Acked-by: Jo-Philipp Wich <>
5 years agokernel: deactivate CONFIG_QCOM_SPMI_TEMP_ALARM
Hauke Mehrtens [Mon, 31 Oct 2016 13:34:08 +0000 (14:34 +0100)]
kernel: deactivate CONFIG_QCOM_SPMI_TEMP_ALARM

This fixes the build of the ipq806x target.

Signed-off-by: Hauke Mehrtens <>
5 years agohostapd: properly package wpa-supplicant-mesh
Alexis Green [Fri, 7 Oct 2016 20:09:13 +0000 (13:09 -0700)]
hostapd: properly package wpa-supplicant-mesh

Ensure that selecting the wpa-supplicant-mesh package actually packages the
wpa_supplicant binary with SAE support and add missing dependency on OpenSSL.

Signed-off-by: Alexis Green <>
[Jo-Philipp Wich: slightly reword commit message for clarity]
Signed-off-by: Jo-Philipp Wich <>
5 years agohostapd support for VLANs through a file in addition to Radius.
Petr Konecny [Fri, 28 Oct 2016 16:29:15 +0000 (09:29 -0700)]
hostapd support for VLANs through a file in addition to Radius.

Signed-off-by: Petr Konecny <>
5 years agouhttpd: Add Basic Auth config
Daniel Dickinson [Sat, 13 Aug 2016 23:24:59 +0000 (19:24 -0400)]
uhttpd: Add Basic Auth config

We add an 'httpauth' section type that contains the options:

prefix: What virtual or real URL is being protected
username: The username for the Basic Auth dialogue
password: Hashed (crypt()) or plaintext password for the Basic Auth dialogue

httpauth section names are given included as list
items to the instances to which they are to be applied.

Further any existing httpd.conf file (really whatever
is configured in the instance, but default of
/etc/httpd.conf) is appended to the per-instance httpd.conf

Signed-off-by: Daniel Dickinson <>
5 years agomusl: fix parsing of quoted time zone names
Hannu Nyman [Thu, 27 Oct 2016 09:22:05 +0000 (12:22 +0300)]
musl: fix parsing of quoted time zone names

fix parsing of the < > quoted time zone names. Compare the correct
character instead of repeatedly comparing the first character.

Patch has been submitted to upstream as

Longer explanation in

Signed-off-by: Hannu Nyman <>
[Jo-Philipp Wich: move references into commit message, shorten title prefix]
Signed-off-by: Jo-Philipp Wich <>
5 years agoar71xx: Add net config for MR12 & MR16
Chris Blake [Sun, 30 Oct 2016 04:16:03 +0000 (23:16 -0500)]
ar71xx: Add net config for MR12 & MR16

Both the MR12 and MR16 are single gigabit ethernet devices, similar to the
MR18. This change gives them the correct network config on a fresh install.

Signed-off-by: Chris Blake <>
[Jo-Philipp Wich: adjust for alphabetical ordering, line wrap commit msg]
Signed-off-by: Jo-Philipp Wich <>
5 years agoar71xx: fix ethernet on wpj344 board
Christian Mehlis [Fri, 28 Oct 2016 12:56:13 +0000 (14:56 +0200)]
ar71xx: fix ethernet on wpj344 board

In 814d70b2 the member mac06_exchange_en of struct
ar8327_pad_cfg was changed to mac06_exchange_dis,
but wpj344 was not adopted to stay in sync.

Signed-off-by: Christian Mehlis <>
Reported-by: Nick Dennis <>
5 years agolldpd: freeze execution of lldpd during reload
Alexandru Ardelean [Tue, 5 Jan 2016 07:40:22 +0000 (09:40 +0200)]
lldpd: freeze execution of lldpd during reload

During reload, we could send invalid information to the other
side and confuse it.

That's why, during reload we'll pause execution, do the reconfig
and resume + update when reload is done.

Signed-off-by: Alexandru Ardelean <>
5 years agolldpd: fix reload function for when interfaces change
Alexandru Ardelean [Wed, 21 Oct 2015 13:11:09 +0000 (16:11 +0300)]
lldpd: fix reload function for when interfaces change

The problem is that interfaces are specified at start as
command line arguments, making them unchange-able via reload.

That means, we have to move (since lldpd allows this) the
interfaces-match-pattern option to be in a config file and reload
the configuration.
It's either that, or do a 'restart'.

Since we're generating the lldpd.conf file, we'll have to
move the 'sysconfdir' of lldpd to /tmp, where the files will
get written ; this will prevent any unncessary flash writes.

Signed-off-by: Alexandru Ardelean <>
5 years agoath10k-firmware: update qca9984 firmware
Pavel Kubelun [Thu, 13 Oct 2016 13:23:43 +0000 (16:23 +0300)]
ath10k-firmware: update qca9984 firmware

Signed-off-by: Pavel Kubelun <>
[Jo-Philipp Wich: add mirror sha256 sum]
Signed-off-by: Jo-Philipp Wich <>
5 years agofirmware-utils: fix compilation on MacOS X
Alexandru Ardelean [Fri, 28 Oct 2016 06:52:12 +0000 (09:52 +0300)]
firmware-utils: fix compilation on MacOS X

Looking into /usr/include/stdlib.h, there's a `devname()` function defined
under some #ifdef's which conflicts with the `static char *devname` definition
in `src/mkwrggimg.c`.

Defining `_ANSI_SOURCE` in the `src/mkwrggimg.c` file, omits that part of the

Another more intrusive approach is to rename `devname` to something like
`g_devname` in `src/mkwrggimg.c`. But I think the `_ANSI_SOURCE` define should
be enough.

Compilation error is:

    src/mkwrggimg.c:64:14: error: redefinition of 'devname' as different kind of symbol
    static char *devname;
    /usr/include/stdlib.h:286:7: note: previous definition is here
    char    *devname(dev_t, mode_t);
    src/mkwrggimg.c:147:12: error: non-object type 'char *(dev_t, mode_t)' (aka 'char *(int, unsigned short)') is not assignable
                            devname = optarg;
                            ~~~~~~~ ^
    src/mkwrggimg.c:192:6: warning: comparison of function 'devname' equal to a null pointer is always false [-Wtautological-pointer-compare]
            if (devname == NULL) {
                ^~~~~~~    ~~~~
    src/mkwrggimg.c:192:6: note: prefix with the address-of operator to silence this warning
            if (devname == NULL) {
    src/mkwrggimg.c:251:27: warning: incompatible pointer types passing 'char *(dev_t, mode_t)' (aka 'char *(int, unsigned short)') to parameter of type 'const char *' [-Wincompatible-pointer-types]
            strncpy(header->devname, devname, sizeof(header->devname));
    /usr/include/secure/_string.h:119:34: note: expanded from macro 'strncpy'
      __builtin___strncpy_chk (dest, src, len, __darwin_obsz (dest))

Signed-off-by: Alexandru Ardelean <>
[Jo-Philipp Wich: slightly reformat commit message]
Signed-off-by: Jo-Philipp Wich <>
5 years agobrcm47xx: image: use append-rootfs step for per-device rootfs support
Rafał Miłecki [Fri, 28 Oct 2016 15:03:55 +0000 (17:03 +0200)]
brcm47xx: image: use append-rootfs step for per-device rootfs support

Signed-off-by: Rafał Miłecki <>
5 years agobrcm47xx: image: make TRX steps work with rootfs passed as $@
Rafał Miłecki [Fri, 28 Oct 2016 14:43:19 +0000 (16:43 +0200)]
brcm47xx: image: make TRX steps work with rootfs passed as $@

Signed-off-by: Rafał Miłecki <>
5 years agobrcm47xx: image: make linksys-pattern-partition leave specific file
Rafał Miłecki [Fri, 28 Oct 2016 14:39:08 +0000 (16:39 +0200)]
brcm47xx: image: make linksys-pattern-partition leave specific file

Thanks to leaving .pattern file we can easily insert extra step between
linksys-pattern-partition and trx-v2-with-loader, e.g. rootfs one.

Signed-off-by: Rafał Miłecki <>
5 years agoapm821xx: fix USB LED trigger for WNDR4700
Rafał Miłecki [Fri, 28 Oct 2016 10:40:01 +0000 (12:40 +0200)]
apm821xx: fix USB LED trigger for WNDR4700

The old usbdev trigger never supported assigning more than 1 USB port.
This code we got was never working as expected and it was missing 2 more
ports. Switch to usbport to have LED working with all ports.

Signed-off-by: Rafał Miłecki <>
Tested-by: Christian Lamparter <>
5 years agoopenvpn: cacert does not exist
John Crispin [Thu, 27 Oct 2016 17:52:33 +0000 (19:52 +0200)]
openvpn: cacert does not exist

cacert is really called ca and already in the script

Signed-off-by: John Crispin <>
5 years agoconfig: ext4: increase x86 rootfs size to 2GB to support online resize2fs
Jo-Philipp Wich [Sat, 15 Oct 2016 23:59:44 +0000 (01:59 +0200)]
config: ext4: increase x86 rootfs size to 2GB to support online resize2fs

The current default rootfs size of 256MB in conjunction with 4K blocks
produces an ext4 filesystem which lacks the appropriate amount of backup GDT
entries to support online-resizing.

For x86 targets, increase the default rootfs size to 2048MB which allows
online resizing the filesystem to up to 2TB which is the current theoretical
maximum for LEDE, due to missing GPT support on the root block device.

Note that the filesystem artefact will not occupy 2GB on the build system as
the make_ext4fs utility uses sparse files to generate the filesystem images,
so the actual disk usage is much lower. Furthermore the filesystem images
are gzip compressed, shrinking them to only a few megabytes on the download

Signed-off-by: Jo-Philipp Wich <>
Acked-by: Michael Heimpold <>
5 years agoconfig: ext4: drop option to set maximum number of inodes
Jo-Philipp Wich [Sat, 15 Oct 2016 23:59:43 +0000 (01:59 +0200)]
config: ext4: drop option to set maximum number of inodes

There is very little practical use to limit the number of available inodes on
an ext4 filesystem and the make_ext4fs utility is able to calculate useful
defaults by itself.

Drop the option to make resulting ext4 filesystems more flexible by default.

Signed-off-by: Jo-Philipp Wich <>
Acked-by: Michael Heimpold <>
5 years agoinclude: make ext4 reserved blocks percentage optional
Jo-Philipp Wich [Sat, 15 Oct 2016 23:59:42 +0000 (01:59 +0200)]
include: make ext4 reserved blocks percentage optional

Allow CONFIG_TARGET_EXT4_RESERVED_PCT to be empty as make_ext4fs is usually
able to figure out a suitable default.

Signed-off-by: Jo-Philipp Wich <>
Reviewed-by: Michael Heimpold <>
5 years agolinux/modules: drop ledtrig-netfilter
John Crispin [Thu, 27 Oct 2016 13:29:36 +0000 (15:29 +0200)]
linux/modules: drop ledtrig-netfilter

this is a duplicate of kmod-ipt-led

Signed-off-by: John Crispin <>
5 years agoopenvpn: add handling for capath and cafile
John Crispin [Thu, 27 Oct 2016 13:19:59 +0000 (15:19 +0200)]
openvpn: add handling for capath and cafile

Signed-off-by: John Crispin <>
5 years agobrcm47xx: add support for per-device rootfs
Rafał Miłecki [Mon, 24 Oct 2016 19:33:21 +0000 (21:33 +0200)]
brcm47xx: add support for per-device rootfs

Signed-off-by: Rafał Miłecki <>
5 years agopackage/network/utils/ipset: Update to 6.30
Daniel Engberg [Mon, 24 Oct 2016 21:16:59 +0000 (23:16 +0200)]
package/network/utils/ipset: Update to 6.30

Updates to 6.30

Signed-off-by: Daniel Engberg <>
5 years agoutil-linux: disc -> Disc and moved some packages
Alberto Bursi [Thu, 27 Oct 2016 06:32:50 +0000 (08:32 +0200)]
util-linux: disc -> Disc and moved some packages

Capitalized "disc" submenu name as all submenu names are capitalized (apart from "database", but I'll fix that

moved "swap-utils" to Filesystem submenu as it is "formatting" a partition as swap so it looks out of place in Disc.

Signed-off-by: Alberto Bursi <>
5 years agoath10k-ct: Add QCA9888/9886 support, fix compat issue.
Ben Greear [Tue, 25 Oct 2016 22:47:37 +0000 (15:47 -0700)]
ath10k-ct: Add QCA9888/9886 support, fix compat issue.

This should fix problems with latest backports, and also adds
driver support for QCA9888 chipset.

Signed-off-by: Ben Greear <>
5 years agobase-files: Ensure reset only works if an overlay exists
Chris Blake [Mon, 24 Oct 2016 18:13:15 +0000 (13:13 -0500)]
base-files: Ensure reset only works if an overlay exists

Currently the reset script will try to run jffs2reset on boards that are
running a rw rootfs, such as ext4. This will cause jffs2reset to fail
and the board to never reboot while the LED blinks until a manual

This commit does two different things:
1. Disables reset on boards that do not have an overlay mount
2. Disables the Blinking LED after 5 seconds if the board does not
support reset

Signed-off-by: Chris Blake <>
5 years agonetifd: update to latest git HEAD
John Crispin [Thu, 27 Oct 2016 10:02:21 +0000 (12:02 +0200)]
netifd: update to latest git HEAD

Signed-off-by: John Crispin <>
5 years agouboot-kirkwood: fix default bootcmd for Seagate Dockstar
Paul Wassi [Tue, 25 Oct 2016 07:57:25 +0000 (09:57 +0200)]
uboot-kirkwood: fix default bootcmd for Seagate Dockstar

Fix the default value for the 'bootcmd' environment variable.
Therefore make the default bootcmd work for buildbot's images.

Signed-off-by: Paul Wassi <>
5 years agouboot-kirkwood: bump to upstream 2016.09.01
Paul Wassi [Wed, 26 Oct 2016 09:14:47 +0000 (11:14 +0200)]
uboot-kirkwood: bump to upstream 2016.09.01

Bump U-Boot for Kirkwood to upstream 2016.09.01. Local patches
cleaned up and reworked.

Signed-off-by: Paul Wassi <>
5 years agoimx6: Add ds1672 RTC to kernel for working hctosys (Gateworks)
Koen Vandeputte [Wed, 26 Oct 2016 14:10:03 +0000 (16:10 +0200)]
imx6: Add ds1672 RTC to kernel for working hctosys (Gateworks)

Fixes the same hctosys init issue as described in commit

The Gateworks Ventana family uses an emulated Dallas 1672
RTC device.

Signed-off-by: Koen Vandeputte <>
5 years agoar71xx: add support for D-Link DAP-2695 rev. A1
Stijn Tintel [Mon, 18 Jul 2016 17:39:57 +0000 (19:39 +0200)]
ar71xx: add support for D-Link DAP-2695 rev. A1

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agoar71xx: build relocate stub for generic and legacy images
Stijn Tintel [Fri, 21 Oct 2016 14:50:39 +0000 (17:50 +0300)]
ar71xx: build relocate stub for generic and legacy images

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agoar71xx/base-files: rename 09_fix-trx-header 09_fix-checksum
Stijn Tintel [Sun, 7 Aug 2016 17:01:20 +0000 (19:01 +0200)]
ar71xx/base-files: rename 09_fix-trx-header 09_fix-checksum

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agomtd: add fixwrgg command
Stijn Tintel [Sun, 7 Aug 2016 16:51:12 +0000 (18:51 +0200)]
mtd: add fixwrgg command

Based on fixseama.

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agomtd: fix endianness detection on musl
Stijn Tintel [Fri, 21 Oct 2016 14:32:03 +0000 (17:32 +0300)]
mtd: fix endianness detection on musl

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agokernel: mtdsplit: add support for WRGG images
Stijn Tintel [Sun, 7 Aug 2016 11:39:00 +0000 (13:39 +0200)]
kernel: mtdsplit: add support for WRGG images

Support splitting WRGG images, found in some D-Link devices (e.g.

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agofirmware-utils: mkwrggimg: new tool for D-Link DAP-2695
Stijn Tintel [Sat, 6 Aug 2016 17:36:24 +0000 (19:36 +0200)]
firmware-utils: mkwrggimg: new tool for D-Link DAP-2695

Based on mkwrgimg.c.

Signed-off-by: Stijn Tintel <>
Acked-by: John Crispin <>
5 years agodnsmasq: Multiple dnsmasq instances support
Hans Dedecker [Wed, 12 Oct 2016 08:33:16 +0000 (10:33 +0200)]
dnsmasq: Multiple dnsmasq instances support

Adds support in uci for configuring multiple dnsmasq instances via
multiple dnsmasq sections.
The uci sections host, boot, mac, tag, vendorclass, userclass,
circuitid, ... will refer to a dnsmasq instance via the instance
parameter defined in the section; if the instance parameter is
not specified backwards compatibility is preserved.

Start/Stopping a dnsmasq instance can be achieved by passing the
dnsmasq instance name as argument to start/stop via the init script.

Multiple dnsmasq instances is usefull in scenarios where you want to
bind a dnsmasq instance to an interface in order to isolate networks.

This patch is a rework of a multiple dnsmasq instance patch by Daniel Dickinson

Signed-off-by: Hans Dedecker <>
5 years agogrub2: add missing SECTION variable and remove non breaking space
Jo-Philipp Wich [Wed, 26 Oct 2016 15:51:44 +0000 (17:51 +0200)]
grub2: add missing SECTION variable and remove non breaking space

Fix metadata scan failure in the grub2 package by removing an unexpected
invisible space character and by adding back the missing SECTION variable
which was removed with d140648.

Signed-off-by: Jo-Philipp Wich <>
5 years agoipip: Support fqdn as remote tunnel endpoint
Hans Dedecker [Tue, 25 Oct 2016 20:50:04 +0000 (22:50 +0200)]
ipip: Support fqdn as remote tunnel endpoint

Signed-off-by: Hans Dedecker <>
5 years agouhttpd: create self-signed certificates with unique subjects
Hannu Nyman [Thu, 6 Oct 2016 17:37:59 +0000 (20:37 +0300)]
uhttpd: create self-signed certificates with unique subjects

Add a partially random O= item to the certificate subject in order
to make the automatically generated certificates' subjects unique.

Firefox has problems when several self-signed certificates
with CA:true attribute and identical subjects have been
seen (and stored) by the browser. Reference to upstream bugs:

Certificates created by the OpenSSL one-liner fall into that category.

Avoid identical certificate subjects by including a new 'O=' item
with CommonName + a random part (8 chars). Example:

That ensures that the browser properly sees the accumulating
certificates as separate items and does not spend time
trying to form a trust chain from them.

Signed-off-by: Hannu Nyman <>