3 years agoar71xx: set GPIO reset line for Ubiquiti NanoStation Loco XW
Felix Fietkau [Wed, 30 Nov 2016 10:56:48 +0000 (11:56 +0100)]
ar71xx: set GPIO reset line for Ubiquiti NanoStation Loco XW

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: add support for configuring at803x PHY GPIO reset via platform data
Felix Fietkau [Wed, 30 Nov 2016 10:55:39 +0000 (11:55 +0100)]
ar71xx: add support for configuring at803x PHY GPIO reset via platform data

Needed to work around ethernet hang issues on Ubiquiti NanoStation Loco XW,
because ar71xx is not converted to device tree yet.

Signed-off-by: Felix Fietkau <>
3 years agokernel: add support for the AT8032 PHY
Felix Fietkau [Wed, 30 Nov 2016 10:34:32 +0000 (11:34 +0100)]
kernel: add support for the AT8032 PHY

This PHY is used on the Ubiquiti Networks NanoStation Loco XW

Signed-off-by: Felix Fietkau <>
3 years agokernel: backport some upstream at803x fixes
Felix Fietkau [Tue, 29 Nov 2016 16:55:41 +0000 (17:55 +0100)]
kernel: backport some upstream at803x fixes

Signed-off-by: Felix Fietkau <>
3 years agoramips: prevent packet forwarding on mt7620 between switch ports during init (FS...
Felix Fietkau [Tue, 29 Nov 2016 10:59:48 +0000 (11:59 +0100)]
ramips: prevent packet forwarding on mt7620 between switch ports during init (FS#103)

By default, forwarding between all ports is allowed on init. This is
problematic in cases where some ports are supposed to be isolated from
each other, most commonly LAN/WAN separation.

REG_ESW_PORT_PCR(port) has a destination mask for a particular port,
controlling what other ports it is allowed to send packets to.
Instead of initializing all to 0xff (all ports), allow each physical
port to send to the CPU port, and the CPU port to send to all other

Signed-off-by: Felix Fietkau <>
3 years agobase-files: add a hint in sysupgrade that shows what to do when the image metadata...
Felix Fietkau [Sat, 26 Nov 2016 17:43:59 +0000 (18:43 +0100)]
base-files: add a hint in sysupgrade that shows what to do when the image metadata check fails

Signed-off-by: Felix Fietkau <>
3 years agoramips: fix LAN port order of WeVO W2914NS v2
Yong-hyu, Ban [Sat, 3 Dec 2016 11:44:18 +0000 (20:44 +0900)]
ramips: fix LAN port order of WeVO W2914NS v2

Signed-off-by: Yong-hyu, Ban <>
3 years agoramips: fix PBR-D1 button definition
BangLang Huang [Fri, 2 Dec 2016 01:13:46 +0000 (09:13 +0800)]
ramips: fix PBR-D1 button definition

Due to the product specification, the button on PBR-D1 should be
reset, not wps.

Signed-off-by: BangLang Huang <>
3 years agoramips: add support for PCI based OHCI/EHCI support for F5D8235 V1
Tobias Wolf [Fri, 18 Nov 2016 14:52:38 +0000 (15:52 +0100)]
ramips: add support for PCI based OHCI/EHCI support for F5D8235 V1

This router uses an PCI attached NEC EHCI controller to support the single USB port.

Signed-off-by: Tobias Wolf <>
3 years agoramips: introduce CONFIG_PCI and CONFIG_OF_*PCI for rt288x
Tobias Wolf [Tue, 15 Nov 2016 21:31:16 +0000 (22:31 +0100)]
ramips: introduce CONFIG_PCI and CONFIG_OF_*PCI for rt288x

This is a prerequisite for F5D8235 V1 to use PCI based OHCI/EHCI.

Signed-off-by: Tobias Wolf <>
3 years agoramips: improve F5D8235 V1 support
Tobias Wolf [Thu, 1 Dec 2016 14:08:22 +0000 (15:08 +0100)]
ramips: improve F5D8235 V1 support

This fixes the partition name for the firmware splitter, the cfi
address and adds the mtd-eeprom address for wmac. It adds additional
LEDs and make use of them in and 01_leds.

Please note that the ":blue:wired" LED is used because the
":blue:router" behaviour is unpredictable for failsafe indication. The
issue with the router LED is that you have two states only.
"off" is steady on and "on" blinks. Therefore the wired LED is more

Furthermore it reuses the correct switch configuration definition to
reflect the device ports and numbering. Additionally fixes the issue
that the default configuration is not applied as no port 6 exists on
this device.

Signed-off-by: Tobias Wolf <>
3 years agomvebu: fix sysupgrade for Linksys WRT3200ACM
Vignesh Balasubramaniam [Tue, 29 Nov 2016 17:53:13 +0000 (12:53 -0500)]
mvebu: fix sysupgrade for Linksys WRT3200ACM

sysupgrade command fails due to missing U-Boot environment-processing
binaries on sysupgrade ramdisk. The missing binaries result in the
following output:

Switching to ramdisk...
Performing system upgrade...
ash: /usr/sbin/fw_printenv: not found
ash: fw_setenv: not found
ash: touch: not found
cannot find target partition

Signed-off-by: Vignesh Balasubramaniam <>
3 years agomvebu: fix image validation error
Mathias Kresin [Thu, 1 Dec 2016 07:57:25 +0000 (08:57 +0100)]
mvebu: fix image validation error

The name from the Device define will be used in the metadata. Due to
typo/different spelling, this name might not match the one exported in

Signed-off-by: Mathias Kresin <>
3 years agolantiq: fix image validation errors
Mathias Kresin [Thu, 1 Dec 2016 06:29:49 +0000 (07:29 +0100)]
lantiq: fix image validation errors

The boards did not have the name set that is expected during metadata
validation on sysupgrade.

Signed-off-by: Mathias Kresin <>
3 years agocyassl: update to wolfssl version 3.9.10
Hauke Mehrtens [Sat, 3 Dec 2016 20:33:48 +0000 (21:33 +0100)]
cyassl: update to wolfssl version 3.9.10

This fixes the following security problems:
CVE-2016-7440: Software AES table lookups do not properly consider cache-bank access times
CVE-2016-7439: Software RSA does not properly consider cache-bank monitoring
CVE-2016-7438: Software ECC does not properly consider cache-bank monitoring
SWEET32 Attack

Signed-off-by: Hauke Mehrtens <>
3 years agovalgrind: update to 3.12.0
Hauke Mehrtens [Sat, 1 Oct 2016 17:54:22 +0000 (19:54 +0200)]
valgrind: update to 3.12.0

Support for MIPS toolchains without FPU support was added upstream,
so remove our patch.

patches/310-mips-link-tool.patch was a backport form this version of valgrind
src/abort.c is not referenced anywhere

Signed-off-by: Hauke Mehrtens <>
3 years agocurl: update to version 7.51.0
Hauke Mehrtens [Sat, 3 Dec 2016 15:07:47 +0000 (16:07 +0100)]
curl: update to version 7.51.0

This fixes the following security problems:
CVE-2016-8615: cookie injection for other servers
CVE-2016-8616: case insensitive password comparison
CVE-2016-8617: OOB write via unchecked multiplication
CVE-2016-8618: double-free in curl_maprintf
CVE-2016-8619: double-free in krb5 code
CVE-2016-8620: glob parser write/read out of bounds
CVE-2016-8621: curl_getdate read out of bounds
CVE-2016-8622: URL unescape heap overflow via integer truncation
CVE-2016-8623: Use-after-free via shared cookies
CVE-2016-8624: invalid URL parsing with '#'
CVE-2016-8625: IDNA 2003 makes curl use wrong host

Signed-off-by: Hauke Mehrtens <>
3 years agombedtls: update to version 2.4.0
Hauke Mehrtens [Sat, 3 Dec 2016 15:07:31 +0000 (16:07 +0100)]
mbedtls: update to version 2.4.0

This fixes two minor security problems.

Signed-off-by: Hauke Mehrtens <>
3 years agopolarssl: update to version 1.3.18
Hauke Mehrtens [Sat, 3 Dec 2016 15:07:14 +0000 (16:07 +0100)]
polarssl: update to version 1.3.18

This fixes two minor security problems.

Signed-off-by: Hauke Mehrtens <>
3 years agoscripts: append Git short hash to revision
Jo-Philipp Wich [Fri, 2 Dec 2016 11:07:19 +0000 (12:07 +0100)]
scripts: append Git short hash to revision

Change to append a short Git commit hash to the end of the artifical
revision number. This way we still have order- and comparable commit numbers
but also a direct relation to the Git commit.

The new output format will look like "r2400+2-882472e" for dirty trees or like
"r2402-882472e" for clean ones.

Signed-off-by: Jo-Philipp Wich <>
3 years agobuild: adjust version number handling
Jo-Philipp Wich [Thu, 1 Dec 2016 16:40:43 +0000 (17:40 +0100)]
build: adjust version number handling

Move the revision info to the VERSION_CODE variable and default VERSION_NUMBER
to CURRENT for master branch builds.

Also introduce a new menuconfig option CONFIG_VERSION_CODE which allows users
to override the revision value put into VERSION_CODE and adjust the template
files used by the base-files package to accomodate for the changed semantics.

While we're at it, also adjust the various URLs to match the current web site.

After this commit, the relevent files will look like the examples given below:

    # cat /etc/openwrt_version

    # cat /etc/openwrt_release
    DISTRIB_TAINTS='no-all override'

    # cat /usr/lib/os-release
    ID_LIKE="lede openwrt"
    LEDE_TAINTS="no-all override"
    LEDE_RELEASE="LEDE Reboot CURRENT r2398+1"

On a release branch, those files would look like:

    # cat /etc/openwrt_version

    # cat /etc/openwrt_release
    DISTRIB_DESCRIPTION='LEDE Test Release 16.12-CURRENT r2399'
    DISTRIB_TAINTS='no-all override'

    # cat /usr/lib/os-release
    VERSION="16.12-CURRENT, Test Release"
    ID_LIKE="lede openwrt"
    PRETTY_NAME="LEDE Test Release 16.12-CURRENT"
    LEDE_TAINTS="no-all override"
    LEDE_RELEASE="LEDE Test Release 16.12-CURRENT r2399"

On a release tag, those files would look like:

    # cat /etc/openwrt_version

    # cat /etc/openwrt_release
    DISTRIB_DESCRIPTION='LEDE Mighty Unicorn 17.02.1 r2500'
    DISTRIB_TAINTS='no-all override'

    # cat /usr/lib/os-release
    VERSION="17.02.1, Mighty Unicorn"
    ID_LIKE="lede openwrt"
    PRETTY_NAME="LEDE Mighty Unicorn 17.02.1"
    LEDE_TAINTS="no-all override"
    LEDE_RELEASE="LEDE Mighty Unicorn 17.02.1 r2500"

Signed-off-by: Jo-Philipp Wich <>
Acked-by: Felix Fietkau <>
3 years agoprocd: update to the latest version, fixes killing jailed processes
Felix Fietkau [Fri, 2 Dec 2016 12:55:12 +0000 (13:55 +0100)]
procd: update to the latest version, fixes killing jailed processes

Signed-off-by: Felix Fietkau <>
3 years agolibubox: update to the latest version
Felix Fietkau [Thu, 1 Dec 2016 19:13:48 +0000 (20:13 +0100)]
libubox: update to the latest version

- Improves C++ compatibility
- Adds static initializers for the kvlist API

Signed-off-by: Felix Fietkau <>
3 years agoramips: fix sdhci support on mt7621
John Crispin [Thu, 1 Dec 2016 17:29:50 +0000 (18:29 +0100)]
ramips: fix sdhci support on mt7621

the IRQ level was wrong. refresh the patch while at it.

Signed-off-by: John Crispin <>
3 years agotools: cmake: fix compatibility with LibreSSL as well
Jo-Philipp Wich [Thu, 1 Dec 2016 15:32:39 +0000 (16:32 +0100)]
tools: cmake: fix compatibility with LibreSSL as well

Sadly, LibreSSL claims to be OpenSSL v2.0.0 while not providing the new
interfaces introduced with OpenSSL v1.1.x, so extend the pre-OpenSSL 1.1.x
compat checks to consider LibreSSL as well.

Signed-off-by: Jo-Philipp Wich <>
3 years agoramips: RB750Gr3: Add pwr LED and buzzer to DTS
Andrew Yong [Sat, 26 Nov 2016 14:15:21 +0000 (22:15 +0800)]
ramips: RB750Gr3: Add pwr LED and buzzer to DTS

use pwr LED in
Expose unused pinmux pins as GPIOs
export power LED and buzzer pins
Use rb750gr3:blue:pwr LED in for boot status instead of rb750gr3:green:usr

Signed-off-by: Andrew Yong <>
3 years agoipq806x: add thermal sensor driver
dissent1 [Sat, 5 Nov 2016 13:15:33 +0000 (16:15 +0300)]
ipq806x: add thermal sensor driver

Allows to check cpu temperature.

Huge thanks to @hnyman for valuable assistance!

Signed-off-by: Pavel Kubelun <>
3 years agoipq806x: increase coherent dma pool size
Pavel Kubelun [Tue, 15 Nov 2016 18:28:36 +0000 (21:28 +0300)]
ipq806x: increase coherent dma pool size

Cherry-picked and rebased from

Signed-off-by: Pavel Kubelun <>
3 years agoipq806x: add enable reg and masks for PRNG
Pavel Kubelun [Tue, 15 Nov 2016 16:27:55 +0000 (19:27 +0300)]
ipq806x: add enable reg and masks for PRNG

Cherry-picked from

Signed-off-by: Pavel Kubelun <>
3 years agonet: ar8327: modify some configuration of switch
Pavel Kubelun [Thu, 12 Feb 2015 01:49:08 +0000 (09:49 +0800)]
net: ar8327: modify some configuration of switch

Imported from
Signed-off-by: Pavel Kubelun <>
Enable flow control of LAN and WAN ports to
get better performance.
Setup pvid as 0 for all ports during initialisation
to avoid confusion during system or switch INIT.
Disable PORT MAC before config MAC to avoid it work abnormal.
This change is for IR-054144, IR-057315.

Change-Id: I345f3dffa59ad3f97150e09692723da12a7b1067
Signed-off-by: Zou Shunxiang <>
Signed-off-by: xiaofeis <>
3 years agonet: ar8216: address security vulnerabilities in swconfig & ar8216
Pavel Kubelun [Mon, 28 Nov 2016 15:10:05 +0000 (18:10 +0300)]
net: ar8216: address security vulnerabilities in swconfig & ar8216

Imported from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: net: ar8216: address security vulnerabilities in swconfig & ar8216

This patch does the following changes:
*address the security vulnerabilities in both swconfig framework and in
 ar8216 driver (many bound check additions, and turned swconfig structure
 signed element into unsigned when applicable)
*address a couple of whitespaces and indendation issues


Change-Id: I94ea78fcce8c1932cc584d1508c6e3b5dfb93ce9
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Toshi Kikuchi <>
Commit-Queue: Toshi Kikuchi <>
Tested-by: Toshi Kikuchi <>
3 years agonet: ar8216: prevent device duplication in ar8xxx_dev_list
Pavel Kubelun [Mon, 28 Nov 2016 12:31:38 +0000 (15:31 +0300)]
net: ar8216: prevent device duplication in ar8xxx_dev_list

Import from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: drivers: ar8216: prevent device duplication in ar8xxx_dev_list

If probe is called twice, once for PHY0 and a second time for PHY4,
the same switch device will be added twice to ar8xxx_dev_list, while
supposedly this list should have one element per hardware switch present
in the system.

While no negative impact have been observed, it does happen if a
platform instanciates these two PHYs from device-tree, as an example.

Change-Id: Iddcbdf7d4adacb0af01975b73f8e56b4582e894e
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Matthias Kaehlcke <>
Reviewed-by: Toshi Kikuchi <>
Tested-by: Toshi Kikuchi <>
3 years agonet: ar8216: hold ar8xxx_dev_list_lock during use_count--
Pavel Kubelun [Mon, 28 Nov 2016 12:21:42 +0000 (15:21 +0300)]
net: ar8216: hold ar8xxx_dev_list_lock during use_count--

Import from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: drivers: ar8216: hold ar8xxx_dev_list_lock during use_count--

It is possible for the remove() callback to run twice in parallel, which
could result into --use_count returning only 1 in both cases and the
rest of the unregistration path to never be reached.

This case has never been observed in practice, but we will fix
preventively to make the code more robust.


Change-Id: If09abe27fdb2037f514f8674418bafaab3cbdef6
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Matthias Kaehlcke <>
Reviewed-by: Toshi Kikuchi <>
Tested-by: Toshi Kikuchi <>
3 years agonet: ar8327: replace sprintf() by scnprintf()
Pavel Kubelun [Mon, 28 Nov 2016 12:14:50 +0000 (15:14 +0300)]
net: ar8327: replace sprintf() by scnprintf()

Import from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: drivers: ar8216: replace sprintf() by scnprintf()


Change-Id: Ib82035c9f2769a86d3e90f9573a09e5700ff5676
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Matthias Kaehlcke <>
Tested-by: Toshi Kikuchi <>
Reviewed-by: Toshi Kikuchi <>
3 years agonet: ar8327: remove unnecessary spinlocks
Pavel Kubelun [Mon, 28 Nov 2016 12:11:14 +0000 (15:11 +0300)]
net: ar8327: remove unnecessary spinlocks

Import from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: drivers: ar8216: remove unnecessary spinlocks


Change-Id: Ia1b51258504501863fd3298717cc923a1baf34ca
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Matthias Kaehlcke <>
Reviewed-by: Toshi Kikuchi <>
Tested-by: Toshi Kikuchi <>
3 years agonet: ar8216: sync mib_work cancellation
Pavel Kubelun [Mon, 28 Nov 2016 11:58:39 +0000 (14:58 +0300)]
net: ar8216: sync mib_work cancellation

Import from
Signed-off-by: Pavel Kubelun <>
CHROMIUM: drivers: ar8216: sync mib_work cancellation

ar8xxx_mib_stop() is called from ar8xxx_phy_remove(), so we want to make
sure the work doesn't run after priv is freed / the device ceases to


Change-Id: Iafb44ce93a87433adc4576e5fea5fda58d1f43a9
Signed-off-by: Mathieu Olivari <>
Reviewed-by: Matthias Kaehlcke <>
Reviewed-by: Toshi Kikuchi <>
Reviewed-by: Grant Grundler <>
Tested-by: Toshi Kikuchi <>
3 years agoipq806x: refactor rpm clock controller patches
Pavel Kubelun [Wed, 23 Nov 2016 15:53:52 +0000 (18:53 +0300)]
ipq806x: refactor rpm clock controller patches

RPM clock controller driver had made its way upstream and previous
approach of directly redoing a driver to support ipq806x is a no go anymore.

Thus reverting mentioned patches to upstream state and renaming
in correct patch numbering accordance.

To make the driver work on ipq806x boards we introduce a custom patch.

Signed-off-by: Pavel Kubelun <>
3 years agotools: cmake: import another upstream commit for OpenSSL backwards compatibility
Jo-Philipp Wich [Thu, 1 Dec 2016 14:27:07 +0000 (15:27 +0100)]
tools: cmake: import another upstream commit for OpenSSL backwards compatibility

Signed-off-by: Jo-Philipp Wich <>
3 years agokernel: add missing config symbols
Felix Fietkau [Thu, 1 Dec 2016 13:52:39 +0000 (14:52 +0100)]
kernel: add missing config symbols

Signed-off-by: Felix Fietkau <>
3 years agotools: cmake: import upstream patch for OpenSSL 1.1.x compatibility
Jo-Philipp Wich [Thu, 1 Dec 2016 13:47:30 +0000 (14:47 +0100)]
tools: cmake: import upstream patch for OpenSSL 1.1.x compatibility

Signed-off-by: Jo-Philipp Wich <>
3 years agoxtables-addons: add CONFIG_NF_CONNTRACK_MARK=y to all kmod-* packages
Felix Fietkau [Thu, 1 Dec 2016 13:23:05 +0000 (14:23 +0100)]
xtables-addons: add CONFIG_NF_CONNTRACK_MARK=y to all kmod-* packages

Not all kmod packages depends on kmod-ipt-compat-xtables, but this
kernel config option is required for building the whole package

Signed-off-by: Felix Fietkau <>
3 years agoe2fsprogs: avoid picking up incompatible
Jo-Philipp Wich [Wed, 30 Nov 2016 22:25:17 +0000 (23:25 +0100)]
e2fsprogs: avoid picking up incompatible

Commit 2728512 ("e2fsprogs: List all libraries explicitly") forced the
e2fsprogs build system to link in various places.

Unfortunately, the krb5 package also ships a with a totally
different ABI and puts it into the global staging directory which causes
e2fsprogs to pick up this wrong library, leading to the following failure:

     LD blkid
    ../lib/ undefined reference to `_et_list'
    collect2: error: ld returned 1 exit status
    Makefile:504: recipe for target 'blkid' failed

Change the SYSLIBS specification to explicitely link
which in order to work around the problem.

In the long run, the libcom_err clash between e2fsprogs and krb5 needs to
be solved properly.

Signed-off-by: Jo-Philipp Wich <>
3 years agomkimage: fix openssl 1.1.x compat fix with libressl
Felix Fietkau [Wed, 30 Nov 2016 19:24:31 +0000 (20:24 +0100)]
mkimage: fix openssl 1.1.x compat fix with libressl

libressl sets OPENSSL_VERSION_NUMBER to 0x20000000L, which breaks API
checks based on it.

Signed-off-by: Felix Fietkau <>
3 years agotools: mkimage: fix build with OpenSSL 1.1.x (FS#182)
Jo-Philipp Wich [Wed, 30 Nov 2016 17:09:05 +0000 (18:09 +0100)]
tools: mkimage: fix build with OpenSSL 1.1.x (FS#182)

The OpenSSL 1.1.x version series undergone some major API changes which made
the RSA structure opaque and deprecated a number of methods, so add some
conditional compat code to make the u-boot source build again.

Signed-off-by: Jo-Philipp Wich <>
3 years agortc-rv5c386a: fix include path for bcm47xx_nvram.h
Felix Fietkau [Wed, 30 Nov 2016 11:37:15 +0000 (12:37 +0100)]
rtc-rv5c386a: fix include path for bcm47xx_nvram.h

Signed-off-by: Felix Fietkau <>
3 years agodocs: remove all refrences in Makefiles/scripts
Mathias Kresin [Wed, 30 Nov 2016 09:00:41 +0000 (10:00 +0100)]
docs: remove all refrences in Makefiles/scripts

The SDK Makefile still trys to copy the docs folder which was removed
with 882f4d2d63272abce8c1966983aa10178e2e971f. This causes an SDK build

All other removals are just cleanup.

Signed-off-by: Mathias Kresin <>
3 years agobase-files: add a wrapper for init scripts in profile
Alberto Bursi [Mon, 28 Nov 2016 11:48:59 +0000 (12:48 +0100)]
base-files: add a wrapper for init scripts in profile

"service" is a simple wrapper that will allow to call init.d scripts

current method:     #  /etc/init.d/network reload
with the wrapper:   #  service network reload

If the wrapper is called without arguments or with a wrong init script name, it will print an error and list the content of /etc/init.d/ folder

Signed-off-by: Alberto Bursi <>
3 years agokernel: bump to 4.4.35
Kevin Darbyshire-Bryant [Mon, 21 Nov 2016 18:18:47 +0000 (18:18 +0000)]
kernel: bump to 4.4.35

Refresh patches on all 4.4 supported platforms.

removed as now upstream.

Compile & run tested: ar71xx - Archer C7 v2

Signed-off-by: Kevin Darbyshire-Bryant <>
3 years agodocs: deleting docs because they are obsolete
Alberto Bursi [Mon, 28 Nov 2016 19:28:12 +0000 (20:28 +0100)]
docs: deleting docs because they are obsolete

the docs in /docs folder are pretty much obsolete and in a not very friendly format (latex, that requires to be
compiled), leaving them there only causes confusion.
LEDE documentation's place is the wiki, or the site.

Signed-off-by: Alberto Bursi <>
3 years agolantiq: drop obsolete patch
Mathias Kresin [Tue, 29 Nov 2016 05:10:35 +0000 (06:10 +0100)]
lantiq: drop obsolete patch

ifxhcd never had roothub support but since kernel 3.x it was expected
that a roothub always exists.

The patched fixed a null pointer deref in the usb subsystem because of
the missing roothub.

Since ifxhcd is gone, this whole patch isn't required any longer.

Signed-off-by: Mathias Kresin <>
3 years agouboot-lantiq: Add BT Home Hub 5A support
Mathias Kresin [Fri, 20 May 2016 21:45:59 +0000 (23:45 +0200)]
uboot-lantiq: Add BT Home Hub 5A support

Based on the submission to the uboot-lantiq repo by Martin Blumenstingl.

Use the ddr_settings.h from the GPL tarball. The NAND boot optimized
one (with memory tuning enabled) doesn't work for the UART boot image.

Use the same mtd layout as the stock u-boot. Add add UBI support.

Use the leds to indicate boot status like it is done with the stock
u-boot. Switch on the red power led if kernel image can't be loaded.
Otherwise switch the green led on.

Make only the ramboot u-boot available. Only this image is required for
the first installation of LEDE.

Signed-off-by: Mathias Kresin <>
3 years agolantiq: use BT HomeHub 5 Type A OEM partition layout
Mathias Kresin [Sat, 3 Oct 2015 11:57:54 +0000 (13:57 +0200)]
lantiq: use BT HomeHub 5 Type A OEM partition layout

This way the on nand bad block table is preserved and used. Add support
for nand OOB ECC checksums as well. It should fix all reported ubi
errors, which were all related to bad nand blocks and a purged on nand
bad block table.

The existing ubi partition will be reused, which eliminates the need
to touch the caldata during initial install. The BT u-boot has support
for loading a kernel from an ubi volume. It isn't necessary any longer
to replace the BT u-boot with a custom compiled one to use LEDE.

It is required to restore the BT Firmware and install LEDE from scratch
to switch to the new partition layout.

An image for restoring the BT firmware and installing LEDE is provided

Signed-off-by: Mathias Kresin <>
3 years agolantiq: backport kernel patch to pass of node to nand_dt_init
Mathias Kresin [Tue, 31 May 2016 06:39:53 +0000 (08:39 +0200)]
lantiq: backport kernel patch to pass of node to nand_dt_init

Backport upstream commit a61ae81a1907af1987ad4c77300508327bc48b23.

The actually purpose of the patch was to do some cleanup. As a side
effect of this cleanup, the device node is now passed to nand_dt_init.

This allows to use the common nand device tree properties

 - nand-bus-width
 - nand-on-flash-bbt
 - nand-ecc-mode
 - nand-ecc-step-size
 - nand-ecc-strength

for the plat_nand driver.

Signed-off-by: Mathias Kresin <>
3 years agolantiq: drop ath9k device tree binding & ath9k pci fixup
Mathias Kresin [Sat, 6 Aug 2016 07:19:45 +0000 (09:19 +0200)]
lantiq: drop ath9k device tree binding & ath9k pci fixup

Due to the addition of the bindings to the ath9k driver, this code
isn't used any longer. The fixup is now done by the owloader.

Rename the ath_eep.c file to ath5k_eep.c to indicate that this files
includes ath5k related code only.

Signed-off-by: Mathias Kresin <>
3 years agolantiq: use ath9k device tree bindings binding/owl-loader
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:15 +0000 (01:01 +0100)]
lantiq: use ath9k device tree bindings binding/owl-loader

This moves the extraction of the eeprom/calibration data to a hotplug
firmware script. Additionally it modifies all .dts to configure ath9k
directly from within the .dts.

The owl-loader approach enables support on devices with exotic eeprom
data locations (such as unaligned positions on the flash or data
inside an UBI volume).

Signed-off-by: Martin Blumenstingl <>
[add ath9k caldata mac address patcher]
[fixes DGN3500 wifi mac]
[fixes BTHOMEHUBV3A wifi mac]
[set invalid mac for BTHOMEHUB2B, FRITZ3370, FRITZ7320 & FRITZ7360SL to restore previous random mac behavior]
Signed-off-by: Mathias Kresin <>
3 years agokernel: mac80211: disable ath9k bands via device tree
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:14 +0000 (01:01 +0100)]
kernel: mac80211: disable ath9k bands via device tree

These properties allow overriding the settings from the EEPROM
which indicate whether a band is enabled or not.
Setting this property is only needed when the RF circuit does not
support the 2.4GHz or 5GHz band while it is enabled nevertheless in the

These patches will be replaced with a future upstream version which
will introduces an ieee80211 device tree property to disable bands.

Signed-off-by: Martin Blumenstingl <>
3 years agokernel: mac80211: add pending ath9k EEPROM swapping patches
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:13 +0000 (01:01 +0100)]
kernel: mac80211: add pending ath9k EEPROM swapping patches

There are two types of swapping the EEPROM data in the ath9k driver.
Before this series one type of swapping could not be used without the

The first type of swapping looks at the "magic bytes" at the start of
the EEPROM data and performs swab16 on the EEPROM contents if needed.
The second type of swapping is EEPROM format specific and swaps
specific fields within the EEPROM itself (swab16, swab32 - depends on
the EEPROM format).

With this series the second part now looks at the EEPMISC register
inside the EEPROM, which uses a bit to indicate if the EEPROM data
is Big Endian (this is also done by the FreeBSD kernel).
This has a nice advantage: currently there are some out-of-tree hacks
(in OpenWrt and LEDE) where the EEPROM has a Big Endian header on a
Big Endian system (= no swab16 is performed) but the EEPROM itself
indicates that it's data is Little Endian. Until now the out-of-tree
code simply did a swab16 before passing the data to ath9k, so ath9k
first did the swab16 - this also enabled the format specific swapping.
These out-of-tree hacks are still working with the new logic, but it
is recommended to remove them. This implementation is based on a
discussion with Arnd Bergmann who raised concerns about the
robustness and portability of the swapping logic in the original OF
support patch review, see [0].

After a second round of patches (= v1 of this series) neither Arnd
Bergmann nor I were really happy with the complexity of the EEPROM
swapping logic. Based on a discussion (see [1] and [2]) we decided
that ath9k should use a defined format (specifying the endianness
of the data - I went with __le16 and __le32) when accessing the
EEPROM fields. A benefit of this is that we enable the EEPMISC based
swapping logic by default, just like the FreeBSD driver, see [3]. On
the devices which I have tested (see below) ath9k now works without
having to specify the "endian_check" field in ath9k_platform_data (or
a similar logic which could provide this via devicetree) as ath9k now
detects the endianness automatically. Only EEPROMs which are mangled
by some out-of-tree code still need the endian_check flag (or one can
simply remove that mangling from the out-of-tree code).


Signed-off-by: Martin Blumenstingl <>
3 years agokernel: mac80211: backport ath9k device tree support patches
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:12 +0000 (01:01 +0100)]
kernel: mac80211: backport ath9k device tree support patches

These patches add support for configuring ath9k based devices via
devicetree. This was tested on PCI(e) based devices. This should work
for AHB based devices as well (adding more AHB specific properties may
still be needed) as soon as the ath79 platform is ready to populate the
ath9k wmac via devicetree.
This patchset was accepted upstream, more information can be found on
the linux-wireless list:

Signed-off-by: Martin Blumenstingl <>
3 years agoar71xx: fix syntax error in /lib/
Matthias Schiffer [Tue, 29 Nov 2016 18:53:27 +0000 (19:53 +0100)]
ar71xx: fix syntax error in /lib/

Remove all comments in ubnt_xm_board_detect(), as it's not possible to add
comments to specific lines of the case construct anyways.

Fixes: 9a5801e7 "ar71xx: add model detection for UBNT Rocket Ti"
Signed-off-by: Matthias Schiffer <>
3 years agoar71xx: add model detection for UBNT Rocket Ti
Matthias Schiffer [Tue, 29 Nov 2016 14:35:46 +0000 (15:35 +0100)]
ar71xx: add model detection for UBNT Rocket Ti

Based-on-patch-by: Neal Oakey <>
Signed-off-by: Matthias Schiffer <>
3 years agokernel: have kmod-ipsec depend on kmod-crypto-echainiv
Stijn Tintel [Thu, 17 Nov 2016 14:33:20 +0000 (15:33 +0100)]
kernel: have kmod-ipsec depend on kmod-crypto-echainiv

The ESP algorithms in CBC mode require echainiv, so have kmod-ipsec
depend on kmod-crypto-echainiv.

See upstream commit 32b6170ca59ccf07d0e394561e54b2cd9726038c.

Signed-off-by: Stijn Tintel <>
3 years agobrcm47xx: fix initramfs image build error
Felix Fietkau [Fri, 25 Nov 2016 12:10:38 +0000 (13:10 +0100)]
brcm47xx: fix initramfs image build error

Signed-off-by: Felix Fietkau <>
3 years agobrcm47xx: merge cpu cache workaround patches into one, ensure they get compiled out...
Felix Fietkau [Fri, 25 Nov 2016 12:09:58 +0000 (13:09 +0100)]
brcm47xx: merge cpu cache workaround patches into one, ensure they get compiled out on mips74k

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove squashfs-64k rootfs image from bin directory, the generic one is enough
Felix Fietkau [Tue, 29 Nov 2016 08:58:33 +0000 (09:58 +0100)]
ar71xx: remove squashfs-64k rootfs image from bin directory, the generic one is enough

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove split kernel/rootfs images where the sysupgrade image can be written...
Felix Fietkau [Tue, 29 Nov 2016 09:05:46 +0000 (10:05 +0100)]
ar71xx: remove split kernel/rootfs images where the sysupgrade image can be written to flash directly

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove 2MB flash variant of WP543
Felix Fietkau [Tue, 29 Nov 2016 08:54:53 +0000 (09:54 +0100)]
ar71xx: remove 2MB flash variant of WP543

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove legacy devices that cannot be supported due to kernel partition size...
Felix Fietkau [Tue, 29 Nov 2016 08:43:36 +0000 (09:43 +0100)]
ar71xx: remove legacy devices that cannot be supported due to kernel partition size limits

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove legacy gzip images
Felix Fietkau [Wed, 23 Nov 2016 09:27:17 +0000 (10:27 +0100)]
ar71xx: remove legacy gzip images

Signed-off-by: Felix Fietkau <>
3 years agoar71xx: remove obsolete jffs2 image building code
Felix Fietkau [Wed, 23 Nov 2016 09:21:07 +0000 (10:21 +0100)]
ar71xx: remove obsolete jffs2 image building code

Signed-off-by: Felix Fietkau <>
3 years agoiw: drop TX power patch that is part of upstream version now
Rafał Miłecki [Tue, 29 Nov 2016 07:38:24 +0000 (08:38 +0100)]
iw: drop TX power patch that is part of upstream version now

Applying it again was resulting in duplicated TX info like:
Interface wlan0
        ifindex 6
        wdev 0x1
        addr 00:23:6a:a3:7d:00
        ssid LEDE2
        type AP
        wiphy 0
        channel 11 (2462 MHz), width: 20 MHz, center1: 2462 MHz
        txpower 31.00 dBm
        txpower 31.00 dBm

Signed-off-by: Rafał Miłecki <>
3 years agokernel: use upstream accepted bcm47xxpart parsing fix
Rafał Miłecki [Mon, 28 Nov 2016 12:14:35 +0000 (13:14 +0100)]
kernel: use upstream accepted bcm47xxpart parsing fix

Signed-off-by: Rafał Miłecki <>
3 years agobcm53xx: use upstream accepted fix for ARM core aborts
Rafał Miłecki [Mon, 28 Nov 2016 12:07:29 +0000 (13:07 +0100)]
bcm53xx: use upstream accepted fix for ARM core aborts

Signed-off-by: Rafał Miłecki <>
3 years agofirmware-utils: replace md5 code with Alexander Peslyak's implementation
Rafał Miłecki [Sat, 19 Nov 2016 20:09:53 +0000 (21:09 +0100)]
firmware-utils: replace md5 code with Alexander Peslyak's implementation

Our current implementation is pretty old and uses some pre-standard/old
ANSI C style that triggers warnings like:
warning: call to function 'MD5_Init' without a real prototype [-Wunprototyped-calls]

This is caused by declarations specified in a following way:
src/md5.h:60:6: note: 'MD5_Init' was declared here
 void MD5_Init ();

Having these warnings makes it harded to notice real problems. We could
try hiding them but it makes more sense to just use a cleaner code.
Another tiny gain from this switch is slightly reduced binary size, on
x86_64 tplink-safeloader's size 48104 became 48003.

The new code is public domain, uses "heavily cut-down BSD license".

Signed-off-by: Rafał Miłecki <>
3 years agolantiq: add Falcon support
Hauke Mehrtens [Sun, 2 Oct 2016 16:39:50 +0000 (18:39 +0200)]
lantiq: add Falcon support

This adds support for the Intel Falcon SoC for GPON.

Support for the Falcon SoC was removed in commit c8218363951 svn rev:
40392 from OpenWrt, this commit adds it again.

This patch adds a new subtarget for the Falcon SoC, but it still misses
all the drivers needed to control the GPON part.

Signed-off-by: Hauke Mehrtens <>
3 years agolantiq: add swconfig to the default packages
Mathias Kresin [Sun, 27 Nov 2016 13:59:20 +0000 (14:59 +0100)]
lantiq: add swconfig to the default packages

The DEVICE_PACKAGES are not included in the initramfs image in case
TARGET_PER_DEVICE_ROOTFS is set. This might produces initramfs images
with a not working network configuration because of a missing swconfig.

Workaround the issues by adding the essential swconfig package to the
default packages selection and deselect swconfig for boards not having
swconfig included before via device packages.

Signed-off-by: Mathias Kresin <>
3 years agoltq-hcd: drop package
Mathias Kresin [Sun, 27 Nov 2016 12:15:23 +0000 (13:15 +0100)]
ltq-hcd: drop package

Since ffd7c155008bff4b08197dfd710d7a1386313f98 it isn't used anymore on
any supported target.

Signed-off-by: Mathias Kresin <>
3 years agogpio-button-hotplug: add more buttons
Mathias Kresin [Wed, 19 Oct 2016 19:08:14 +0000 (21:08 +0200)]
gpio-button-hotplug: add more buttons

The keycodes are used by some boards.

Signed-off-by: Mathias Kresin <>
3 years agotreewide: fix button keys codes used in dts
Mathias Kresin [Sun, 27 Nov 2016 08:47:51 +0000 (09:47 +0100)]
treewide: fix button keys codes used in dts

Use keycodes that matches the intended use case based on the label.

Signed-off-by: Mathias Kresin <>
3 years agoramips: add size checks/append metadata where missing
Mathias Kresin [Sat, 26 Nov 2016 09:43:34 +0000 (10:43 +0100)]
ramips: add size checks/append metadata where missing

Signed-off-by: Mathias Kresin <>
3 years agoramips: fix Airlink AR670W factory image
Mathias Kresin [Fri, 25 Nov 2016 12:03:01 +0000 (13:03 +0100)]
ramips: fix Airlink AR670W factory image

Move the mkwrgimg build step to the makefile. It will be required by
other boards as well.

Signed-off-by: Mathias Kresin <>
3 years agomvebu: fix typo in image metadata support
Mathias Kresin [Fri, 25 Nov 2016 16:44:24 +0000 (17:44 +0100)]
mvebu: fix typo in image metadata support

Signed-off-by: Mathias Kresin <>
3 years agoipq806x: more dts cleanup
Mathias Kresin [Sat, 26 Nov 2016 09:20:03 +0000 (10:20 +0100)]
ipq806x: more dts cleanup

Remove the wifi5g LED from the the d7800, r7500 and r7800. Albeit this
GPIO is mentioned in the GPL tarball, it doesn't do anything. The
2.4/5 GHz LEDs are connected to the wifi chips and not be controlled
from the the userspace.

Use the LEDs names/colours as they are used in the board manuals. Merge
redundant LED configurations. Use the phy[0|1]tpt trigger for the
wireless LEDs. Remove the workarounds for the not controllable wireless

Fix spi compatible strings and remove superfluous spi-max-frequency

If there are two power leds, use one for indicating normal operation and
one for failsafe/upgrade. Keep the on/off state of the main power led
during boot.

Use the usb pinmux settings from the nbg6817 gpl sources.

Signed-off-by: Mathias Kresin <>
3 years agoipq806x: clean up dts files
Henryk Heisig [Mon, 21 Nov 2016 20:04:52 +0000 (21:04 +0100)]
ipq806x: clean up dts files

Set the pinmux to the values found in the GPL tarballs of the boards.
Remove pinmux which are is not used (like nand pinmux for spi
flash boards).

This allows to use the wan orange led of the C2600 which had a wrong
pinmux before. Might fix buttons or leds of other boards as well.

Fix the LED color and the ledswitch key code of the C2600. Rename the
ledgnr to ledswitch.

Add support for indication the boot state using LEDs to the D7800,
NBG6817, R7500 and R7500v2.

Change GPIO active to readable values in D7800, EA8500, R7500,
R7500v2 and R7800.

Change gpioexport to gpio pinmux.

Add proper "drive strenght" to i2c4_pins and use it for RPM on
C2600, D7800, EA8500, R7500, R7500v2.

Remove pcie pinmux from D7800.

Move pinctrl to correct place in NBG6817 and R7800.

Signed-off-by: Henryk Heisig <>
3 years agofirmware-utils: Fix build failure in mkmerakifw.c FS#298
Florian Fainelli [Wed, 23 Nov 2016 20:50:25 +0000 (12:50 -0800)]
firmware-utils: Fix build failure in mkmerakifw.c FS#298

Earlier compilers such as GCC 4 do not like anynomous unions, make it a named
union "u", and update the code where relevant.

Fixes FS#298
Fixes: a72e1692b8b3 ("firmware-utils: Add support for the Cisco Meraki MX60/MX60W")
Signed-off-by: Florian Fainelli <>
3 years agoipset: Add InstallDev to provide libipset as library
Julian Kornberger [Sat, 26 Nov 2016 10:15:34 +0000 (11:15 +0100)]
ipset: Add InstallDev to provide libipset as library

3 years agolantiq: specify the PCIe controller's interrupt, size and address cells
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:11 +0000 (01:01 +0100)]
lantiq: specify the PCIe controller's interrupt, size and address cells

This allows adding devices to the PCIe controller in the .dts files.

Signed-off-by: Martin Blumenstingl <>
3 years agolantiq: Register the device tree node with the PCIe controller
Martin Blumenstingl [Sat, 26 Nov 2016 00:01:10 +0000 (01:01 +0100)]
lantiq: Register the device tree node with the PCIe controller

This allows specifying PCI devices as children of the PCIe controller
node to pass configuration data to them.

Signed-off-by: Martin Blumenstingl <>
3 years agoramips: Add missing chunked-io directive to W2914NSV2.dtsi
Yong-hyu, Ban [Sat, 26 Nov 2016 12:20:14 +0000 (21:20 +0900)]
ramips: Add missing chunked-io directive to W2914NSV2.dtsi

Fix bug that LEDE failed to boot with this message:

WARNING: CPU: 2 PID: 1 at drivers/spi/spi-mt7621.c:214

Signed-off-by: Yong-hyu, Ban <>
3 years agoramips: SamKnows SK-WB8 DTS cleanup
Andrew Yong [Mon, 21 Nov 2016 15:24:57 +0000 (23:24 +0800)]
ramips: SamKnows SK-WB8 DTS cleanup

Use gpio.h definition of GPIO_ACTIVE_HIGH and GPIO_ACTIVE_LOW. Remove unused backup partition to increase available JFFS space. As long as U-Boot env variable "bootcount" is < 3 (reset to 0 after boot by init script) SamKnow's U-Boot will not attempt to boot from the backup flash address (0xe30000).

Signed-off-by: Andrew Yong <>
3 years agokernel: fix typo in input-gpio-encoder package title
Avi H. D [Thu, 24 Nov 2016 20:49:35 +0000 (02:19 +0530)]
kernel: fix typo in input-gpio-encoder package title

Line 115, 'rotary' was misspelled as 'rotay'

Signed off by: Avinash Duduskar <>

3 years agotools/cmake: update to 3.7.0
Hannu Nyman [Fri, 25 Nov 2016 14:10:09 +0000 (16:10 +0200)]
tools/cmake: update to 3.7.0

Release notes at

Signed-off-by: Hannu Nyman <>
3 years agotools/quilt: update to 0.65
Hannu Nyman [Fri, 25 Nov 2016 14:08:31 +0000 (16:08 +0200)]
tools/quilt: update to 0.65

User-visible changes since 0.64 are as follows:

  - Translation fixes
  - Project settings have priority
  - Reject binary files in patches
  - Fix a race condition in diff_file
  - Performance: Optimizations to the setup command
  - Performance: Optimizations to the bash completion script
  - Test suite: Improve the edit test case
  - Test suite: Make the symlink test more robust
  - Test suite: Test backup failure
  - Test suite: Test the header command with hard links
  - diff: Report diff failures
  - edit: Fix a corner case bug
  - mail: Fix the help text
  - push: Fix the synopsis
  - refresh: Do not remove symlinks
  - refresh: Break links to read-only patch files
  - refresh: Always preserve modification time
  - setup: Report failed look-ups in inspect-wrapper
  - quilt.el: Fix quilt-editable when patches are stored in subdirs
  - bash_completion: Handle spaces in file names
  - bash_completion: Update the list of commands
  - bash_completion: Add new command options
  - bash_completion: Fix handling of mail, new, push options
  - guards: Simplify the help text
  - guards: Include the file name also in the "Not found" case
  - guards: Add support for an external filelist in --check mode
  - guards: Report which config file has problem in --check mode
  - guards: Documentation update
  - guards: Clarify a user message

Note to packagers: the Makefile was modified to take configure's
--sysconfdir into account as other projects do. As a result, setting
--prefix=/usr will no longer put the configuration files under /etc.
You now need to explicitly pass --sysconfdir=/etc. If you don't,
configuration files will go under /usr/etc, which is not what you want.
This is somewhat less intuitive, but also more consistent with what
other projects are doing.

Signed-off-by: Hannu Nyman <>
3 years agotools/scons: update to 2.5.1
Hannu Nyman [Fri, 25 Nov 2016 14:07:11 +0000 (16:07 +0200)]
tools/scons: update to 2.5.1

RELEASE 2.5.1 - Mon, 03 Nov 2016 13:37:42 -0400
    - Add to packaging. It was omitted
    - Use memoization to optimize PATH evaluation across
      all dependencies per node.

Signed-off-by: Hannu Nyman <>
3 years agoRevert "mvebu: simplify etc/board.d/02_network"
Felix Fietkau [Fri, 25 Nov 2016 19:02:56 +0000 (20:02 +0100)]
Revert "mvebu: simplify etc/board.d/02_network"

This reverts commit 539ae47103b7de5cecb1a94beab0acb97ef7df1d.
According to reports, this introduces a regression on WRT3200ACM

Signed-off-by: Felix Fietkau <>
3 years agoramips: remove kmod-ledtrig-usbdev from recently added devices
Mathias Kresin [Thu, 24 Nov 2016 07:03:22 +0000 (08:03 +0100)]
ramips: remove kmod-ledtrig-usbdev from recently added devices

None of the devices uses an usb led at all.

Signed-off-by: Mathias Kresin <>
3 years agoar71xx: move DomyWifi DW33D to nand subtarget
Piotr Dymacz [Wed, 23 Nov 2016 15:34:31 +0000 (16:34 +0100)]
ar71xx: move DomyWifi DW33D to nand subtarget

This device uses NAND FLASH, so it should be kept in nand subtarget.

Also, inlcude in packages kmod-usb-ledtrig-usbport instead of
obsolete kmod-ledtrig-usbdev.

Signed-off-by: Piotr Dymacz <>
3 years agoipq806x: add support for TP-Link Archer VR2600v
Sebastian Quilitz [Wed, 23 Nov 2016 17:51:15 +0000 (18:51 +0100)]
ipq806x: add support for TP-Link Archer VR2600v

This router is similar to the C2600. Ethernet on WAN + LAN, switch,
sysupgrade, LEDs, buttons and WiFi on 2G + 5G do work. The xDSL modem
and the POTS/DECT interface are not supported yet.

It is not possible to flash LEDE via the TP-Link webinterface. The
image need to be signed. The first 0x200 bytes of the image is the
TP-Link header including the signature. The signature is not validated
by the bootloader. The LEDE image is zeroed in this area.

To install LEDE it is necessary to solder a four pin header to JP2.
Connect a serial interface to this header and interrupt the autostart
of kernel. Transfer the sysupgrade image via TFTP and write it to the
serial flash at 0x320000.

Signed-off-by: Sebastian Quilitz <>
3 years agoremove bogus dev/null that was accidentally introduced while applying a patch
Felix Fietkau [Thu, 24 Nov 2016 12:42:58 +0000 (13:42 +0100)]
remove bogus dev/null that was accidentally introduced while applying a patch

Signed-off-by: Felix Fietkau <>
3 years agokernel: remove another redundant KCONFIG entry in
Felix Fietkau [Thu, 24 Nov 2016 12:41:10 +0000 (13:41 +0100)]
kernel: remove another redundant KCONFIG entry in

Signed-off-by: Felix Fietkau <>
3 years agokernel: fix virtualization kmod dependencies and kconfig symbols
Felix Fietkau [Thu, 24 Nov 2016 12:29:05 +0000 (13:29 +0100)]
kernel: fix virtualization kmod dependencies and kconfig symbols

Signed-off-by: Felix Fietkau <>