21 months agouboot-mvebu: add missing UBOOT_MAKE_FLAGS variable
Marko Ratkaj [Mon, 25 Sep 2017 10:28:52 +0000 (12:28 +0200)]
uboot-mvebu: add missing UBOOT_MAKE_FLAGS variable

This patch removes "/bin/sh: HOSTCPPFLAGS: command not found" errors douring build.

Signed-off-by: Marko Ratkaj <>
21 months agoRevert "toolchain/gdb: update to version 8.0.1"
Stijn Tintel [Mon, 25 Sep 2017 20:16:37 +0000 (23:16 +0300)]
Revert "toolchain/gdb: update to version 8.0.1"

Since version 8.0, gdb requires at least gcc 4.8. Unfortunately some of
the buildbot slaves don't meet this requirement, and fail to build LEDE
after the gdb upgrade. Revert to the previous gdb version for now.

This reverts commit 592abe9ef53f921554d48085d6482d4507b3e142.

Signed-off-by: Stijn Tintel <>
21 months agosamba36: add Package/samba/Default
Stijn Tintel [Sun, 12 Mar 2017 02:09:36 +0000 (03:09 +0100)]
samba36: add Package/samba/Default

Signed-off-by: Stijn Tintel <>
21 months agoipset: bump to 6.34
Stijn Tintel [Mon, 25 Sep 2017 02:55:27 +0000 (05:55 +0300)]
ipset: bump to 6.34

Signed-off-by: Stijn Tintel <>
21 months agotoolchain/gdb: update to version 8.0.1
Ryan Mounce [Sun, 6 Aug 2017 04:52:18 +0000 (14:22 +0930)]
toolchain/gdb: update to version 8.0.1

Fixes CVE-2017-9778.

Signed-off-by: Ryan Mounce <>
[reference fixed CVE]
Signed-off-by: Stijn Tintel <>
21 months agocurl: bump to 7.55.1
Stijn Tintel [Mon, 25 Sep 2017 03:00:51 +0000 (06:00 +0300)]
curl: bump to 7.55.1

Update 200-no_docs_tests.patch.
Refresh patches.

Fixes the following CVEs:
- CVE-2017-1000099
- CVE-2017-1000100
- CVE-2017-1000101

Signed-off-by: Stijn Tintel <>
21 months agoiperf: bump to 2.0.10
Stijn Tintel [Mon, 25 Sep 2017 02:59:23 +0000 (05:59 +0300)]
iperf: bump to 2.0.10

Signed-off-by: Stijn Tintel <>
21 months agobrcm47xx: Fix sysupgrade with E1200v1
Rosen Penev [Fri, 15 Sep 2017 23:09:19 +0000 (16:09 -0700)]
brcm47xx: Fix sysupgrade with E1200v1

Entry was missing for some reason.

Signed-off-by: Rosen Penev <>
21 months agosunxi: add Olimex A20-OLinuXino-LIME2-eMMC
Lucian Cristian [Mon, 18 Sep 2017 22:13:44 +0000 (01:13 +0300)]
sunxi: add Olimex A20-OLinuXino-LIME2-eMMC

Signed-off-by: Lucian Cristian <>
[replaced u-boot patch with original version from u-boot git]
Signed-off-by: Hauke Mehrtens <>
21 months agotools: flex: fix segfault with glibc 2.26+
Marko Ratkaj [Tue, 19 Sep 2017 07:35:39 +0000 (09:35 +0200)]
tools: flex: fix segfault with glibc 2.26+

Fix segmentation fault caused by implicit declaration of function 'reallocarray'. Added patch will enable
reallocarray() prototype in glibc 2.26+ on Linux systems. This fix will be included in flex 2.6.5.

Fixes LEDE issue: FS#1003 (Flex does not build with GCC 7.2)

Signed-off-by: Marko Ratkaj <>
21 months agokernel: don't scrimp on memory on big iron
Philip Prindeville [Tue, 19 Sep 2017 23:49:13 +0000 (17:49 -0600)]
kernel: don't scrimp on memory on big iron

x86_64 platforms typically don't lack memory, so don't needlessly
economize memory if fq_codel on capable platforms.

Signed-off-by: Philip Prindeville <>
[Add a comment to the patch]
Signed-off-by: Hauke Mehrtens <>
21 months agokernel: kmod-usb-storage-uas
James Christopher Adduono [Wed, 20 Sep 2017 05:30:45 +0000 (01:30 -0400)]
kernel: kmod-usb-storage-uas

This will allow you to build and package the uas.ko module.
With more routers supporting USB 3.0 host this could help
speed up activities like DLNA and Samba, as well as reduce
CPU utilization over BOT mass storage drivers.

Signed-off-by: James Christopher Adduono <>
21 months agoramips: fix missing mediatek wdt
Kevin Darbyshire-Bryant [Wed, 20 Sep 2017 14:10:42 +0000 (15:10 +0100)]
ramips: fix missing mediatek wdt

mediatek MT7621 soc watchdog DTS id was renamed from "mtk,mt7621-wdt" to
"mediatek,mt7621-wdt" when driver upstreamed to kernel 4.5

Update mt7621.dtsi & mt7628an.dtsi definitions to match upstreamed

Restores hardward watchdog functionality on mt7621 devices under linux

Tested on: MIR3G

Signed-off-by: Kevin Darbyshire-Bryant <>
21 months agouhttp: update to latest version
Adrian Panella [Tue, 12 Sep 2017 18:29:09 +0000 (13:29 -0500)]
uhttp: update to latest version

3fd58e9 2017-08-19 uhttpd: add manifest support
88c0b4b 2017-07-09 file: fix basic auth regression
99957f6 2017-07-02 file: remove unused "auth" member from struct
c0a569d 2017-07-02 proc: expose HTTP_AUTH_USER and HTTP_AUTH_PASS
ad93be7 2017-07-02 auth: store parsed username and password
fa51d7f 2017-07-02 proc: do not declare empty process variables
a8bf9c0 2017-01-26 uhttpd: Add TCP_FASTOPEN support
e6cfc91 2016-10-25 lua: ensure that PATH_INFO starts with a slash

Signed-off-by: Adrian Panella <>
21 months agolibubox: fix uloop race condition
Hans Dedecker [Thu, 21 Sep 2017 20:42:28 +0000 (22:42 +0200)]
libubox: fix uloop race condition

7a10576 uloop: Fix race condition in SIGCHLD handling

Signed-off-by: Hans Dedecker <>
21 months agokernel: update 4.9 to 4.9.51
Stijn Tintel [Wed, 20 Sep 2017 20:18:24 +0000 (23:18 +0300)]
kernel: update 4.9 to 4.9.51

Refresh patches.
Compile-tested on octeon and x86/64.
Runtime-tested on octeon and x86/64.

Fixes the following CVEs:
- CVE-2017-14106
- CVE-2017-14497

Signed-off-by: Stijn Tintel <>
21 months agoarm-trusted-firmware-sunxi: depend on sunxi target
Stijn Tintel [Tue, 19 Sep 2017 18:34:47 +0000 (21:34 +0300)]
arm-trusted-firmware-sunxi: depend on sunxi target

The arm-trusted-firmware-sunxi package is only used by the Allwinner
A64, so only make it selectable for its subtarget sunxi/cortexa53.

Signed-off-by: Stijn Tintel <>
Acked-by: Hauke Mehrtens <>
Reviewed-by: Jonas Gorski <>
21 months agoat91bootstrap: New package at91bootstrap
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:23 +0000 (11:51 -0700)]
at91bootstrap: New package at91bootstrap

at91bootstrap is a second-level bootloader for Microchip(Atmel AT91) SoCs.
It provides a set of algorithms to manage the hardware initialization and
to download the main application or a third-level bootloader(i.e. uboot)
from specified boot media to main memory and execute it.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agouboot-at91: Add support for SAMA5D4 Xplained board
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:22 +0000 (11:51 -0700)]
uboot-at91: Add support for SAMA5D4 Xplained board

Add support for SAMA5D4 Xplained board and options to select & build
u-boot configs for different media storage.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agouboot-at91: Add support for SAMA5D2 Xplained board
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:21 +0000 (11:51 -0700)]
uboot-at91: Add support for SAMA5D2 Xplained board

Add support for SAMA5D2 Xplained board and options to select & build
u-boot configs for different media storage.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agouboot-at91: Add support for SAMA5D3 Xplained board
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:20 +0000 (11:51 -0700)]
uboot-at91: Add support for SAMA5D3 Xplained board

Add support for SAMA5D3 Xplained board and options to select & build
u-boot configs for different media storage.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agouboot-at91: move BUILD_SUBTARGET from U-Boot/Default to devices
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:19 +0000 (11:51 -0700)]
uboot-at91: move BUILD_SUBTARGET from U-Boot/Default to devices

currenlty U-Boot/Default supports only at91 legacy devices.To add
sama5 support, moving BUILD_SUBTARGET from U-Boot/Default to target

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agoat91: Add UBI parameters for sama5d4.
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:18 +0000 (11:51 -0700)]
at91: Add UBI parameters for sama5d4.

Add UBIFS_OPTS & UBINIZE_OPTS parameters for sama5d4 Xplained board.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agoat91: Add SAMA5D4 device
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:17 +0000 (11:51 -0700)]
at91: Add SAMA5D4 device

Add support for SAMA5D4 with target device as at91-sama5d4_xplained
in SAMA5 subtarget and build images for SAMA5D4 Xplained board.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agoat91: Add SAMA5D2 device
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:16 +0000 (11:51 -0700)]
at91: Add SAMA5D2 device

Add support for SAMA5D2 with target device as at91-sama5d2_xplained
in SAMA5 subtarget and build images for SAMA5D2 Xplained board.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agoat91: Install zImage.
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:15 +0000 (11:51 -0700)]
at91: Install zImage.

Installing zImage to bin folder of device target.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agobuild: add image command for installing zImage file.
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:14 +0000 (11:51 -0700)]
build: add image command for installing zImage file.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agoat91: Renaming subtarget sama5d3 to sama5
Sandeep Sheriker Mallikarjun [Wed, 13 Sep 2017 18:51:13 +0000 (11:51 -0700)]
at91: Renaming subtarget sama5d3 to sama5

Renaming at91 subtarget sama5d3 to sama5 and using at91-sama5d3_xplained
as a target device in sama5 subtarget.This will enable to add other
sama5d2 & sama5d4 target devices in sama5 subtraget.This will avoid
code duplication when sama5d2 & sama5d4 added as different subtarget.

Signed-off-by: Sandeep Sheriker Mallikarjun <>
21 months agobuild: remove @ as it's causing an error
Philip Prindeville [Tue, 19 Sep 2017 21:17:09 +0000 (15:17 -0600)]
build: remove @ as it's causing an error

Since $(DownloadMethod/unknown) is being invoked in the expansion of
$(call locked ...) anyway, you can't have an @ because the shell
doesn't know what to do with it.

Signed-off-by: Philip Prindeville <>
21 months agoipq-wifi: fix missing define of PKG_NAME
Chen Minqiang [Wed, 20 Sep 2017 02:20:09 +0000 (10:20 +0800)]
ipq-wifi: fix missing define of PKG_NAME

Signed-off-by: Chen Minqiang <>
21 months agousbutils: avoid duplicating the git revision
Philip Prindeville [Tue, 19 Sep 2017 20:47:54 +0000 (14:47 -0600)]
usbutils: avoid duplicating the git revision

Signed-off-by: Philip Prindeville <>
21 months agoopenvpn: add "extra-certs" option
Sven Roederer [Tue, 5 Sep 2017 16:27:02 +0000 (18:27 +0200)]
openvpn: add "extra-certs" option

This option is used to specify a file containing PEM certs, to complete the
local certificate chain. Which is quite usefull for "split-CA" setups.

Signed-off-by: Sven Roederer <>
Signed-off-by: Yousong Zhou <>
22 months agosunxi: add Olimex A20-OlinuXino-LIME2
Lucian Cristian [Fri, 19 May 2017 01:14:08 +0000 (04:14 +0300)]
sunxi: add Olimex A20-OlinuXino-LIME2

Signed-off-by: Lucian Cristian <>
22 months agohostapd: ft_over_ds support
Lorenzo Santina [Sat, 16 Sep 2017 09:14:27 +0000 (11:14 +0200)]
hostapd: ft_over_ds support

Add support for ft_over_ds flag in ieee80211r

Signed-off-by: Lorenzo Santina <>
22 months agohostapd: ft_psk_generate_local support
Lorenzo Santina [Sat, 16 Sep 2017 09:07:24 +0000 (11:07 +0200)]
hostapd: ft_psk_generate_local support

Add support for ft_psk_generate_local flag in ieee80211r

Signed-off-by: Lorenzo Santina <>
[original author]
Signed-off-by: Sergio <>
22 months agoath10k-firmware: use firmware from git instead of extra download
Hauke Mehrtens [Sun, 17 Sep 2017 19:33:20 +0000 (21:33 +0200)]
ath10k-firmware: use firmware from git instead of extra download

Instead of manually downloading the files again we can also take the
same files directly from the ath10k-firmware git which was cloned

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: Add A64 support with cortex53 subtarget
Hauke Mehrtens [Sat, 15 Jul 2017 20:53:20 +0000 (22:53 +0200)]
sunxi: Add A64 support with cortex53 subtarget

This adds initial support for the A64 Allwinner SoC to LEDE.
It will be build in the new cortexa53 subtarget.

Currently it only supports the pine64 and the image is able to boot on
this SoC.

Camera, Ethernet, HDMI and other parts are currently not working.

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: Backport patches from kernel 4.11 for A64
Hauke Mehrtens [Mon, 17 Jul 2017 20:48:31 +0000 (22:48 +0200)]
sunxi: Backport patches from kernel 4.11 for A64

This backports some more patches from kernel 4.11 adding more devices
to the device tree of the A64 SoC.

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: Backport patches needed for A64
Hauke Mehrtens [Sat, 15 Jul 2017 20:50:41 +0000 (22:50 +0200)]
sunxi: Backport patches needed for A64

This backports multiple patches from kernel 4.10 which are adding
missing support for the A64 and the pine64 board. These are the device
tree files, the pinctlk and the clock driver.

Signed-off-by: Hauke Mehrtens <>
22 months agouboot-sunxi: build A64 SoC and pine64 U-Boot
Hauke Mehrtens [Sat, 15 Jul 2017 20:53:35 +0000 (22:53 +0200)]
uboot-sunxi: build A64 SoC and pine64 U-Boot

This creates a U-Boot for the aarch64 SoC A64 on the pine64 board.

Signed-off-by: Hauke Mehrtens <>
22 months agoarm-trusted-firmware-sunxi: add new package
Hauke Mehrtens [Sat, 15 Jul 2017 22:01:21 +0000 (00:01 +0200)]
arm-trusted-firmware-sunxi: add new package

This is needed for the Boot loader of the A64 SoC.

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: split into cortex A8 and A7 subtarget
Hauke Mehrtens [Sat, 15 Jul 2017 18:35:57 +0000 (20:35 +0200)]
sunxi: split into cortex A8 and A7 subtarget

Now we can activate some compiler optimizations for the cortex A7.

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: fix build of rtc package when module not available
Hauke Mehrtens [Sat, 15 Jul 2017 20:51:25 +0000 (22:51 +0200)]
sunxi: fix build of rtc package when module not available

If the Kconfig option CONFIG_RTC_DRV_SUNXI is not selected this package
should be be build.

Signed-off-by: Hauke Mehrtens <>
22 months agouboot-sunxi: revert the usage of binman
Hauke Mehrtens [Thu, 20 Jul 2017 21:27:31 +0000 (23:27 +0200)]
uboot-sunxi: revert the usage of binman

This will avoid the usage of swig.

Signed-off-by: Hauke Mehrtens <>
22 months agouboot-sunxi: do not depend on dtc being install on host
Hauke Mehrtens [Wed, 19 Jul 2017 20:46:34 +0000 (22:46 +0200)]
uboot-sunxi: do not depend on dtc being install on host

make mkimage check the DTC environment variable first.

Signed-off-by: Hauke Mehrtens <>
22 months agouboot-sunxi: update to version 2017.07
Hauke Mehrtens [Sat, 15 Jul 2017 16:41:57 +0000 (18:41 +0200)]
uboot-sunxi: update to version 2017.07

The deleted patches are already integrated in the upstream U-Boot

Signed-off-by: Hauke Mehrtens <>
22 months agosunxi: add support for kernel 4.9
Hauke Mehrtens [Sat, 15 Jul 2017 12:57:08 +0000 (14:57 +0200)]
sunxi: add support for kernel 4.9

Most of the patches were backpoprts from the mainline kernel and are
integrated upstream now.

Signed-off-by: Hauke Mehrtens <>
22 months agokernel: add some config options
Hauke Mehrtens [Sat, 15 Jul 2017 13:57:29 +0000 (15:57 +0200)]
kernel: add some config options

These are needed for the sunxi target.

Signed-off-by: Hauke Mehrtens <>
22 months agorb532: drop 4.4 support
Stijn Tintel [Mon, 18 Sep 2017 09:13:18 +0000 (12:13 +0300)]
rb532: drop 4.4 support

Signed-off-by: Stijn Tintel <>
22 months agorb532: add myself as maintainer
Roman Yeryomin [Sun, 17 Sep 2017 18:35:19 +0000 (21:35 +0300)]
rb532: add myself as maintainer

Signed-off-by: Roman Yeryomin <>
22 months agorb532: switch to 4.9
Roman Yeryomin [Sun, 17 Sep 2017 18:34:36 +0000 (21:34 +0300)]
rb532: switch to 4.9

Signed-off-by: Roman Yeryomin <>
22 months agorb532: add support for 4.9
Roman Yeryomin [Sun, 17 Sep 2017 18:34:00 +0000 (21:34 +0300)]
rb532: add support for 4.9

Includes latest korina fixes.

Signed-off-by: Roman Yeryomin <>
[rewrite commit message (subject <= 50 characters)]
Signed-off-by: Stijn Tintel <>
22 months agorb532: set lan interface type to brigde
Roman Yeryomin [Sun, 17 Sep 2017 18:33:20 +0000 (21:33 +0300)]
rb532: set lan interface type to brigde

Signed-off-by: Roman Yeryomin <>
22 months agorb532: increase kernel size limit
Roman Yeryomin [Sun, 17 Sep 2017 18:32:42 +0000 (21:32 +0300)]
rb532: increase kernel size limit

This is required to support kernel 4.9.

Signed-off-by: Roman Yeryomin <>
[rewrite commit message (subject <= 50 characters)]
Signed-off-by: Stijn Tintel <>
22 months agodnsmasq: fix dhcp "ignore" option on wwan interfaces
Marcin Jurkowski [Thu, 14 Sep 2017 12:49:10 +0000 (14:49 +0200)]
dnsmasq: fix dhcp "ignore" option on wwan interfaces

Init script won't append --no-dhcp-interface option if interface
protocol is one of: ncm, directip, qmi, mbim.
This is caused by IP address assigned to dynamically created netifd
interfaces. As a result there's no netmask assigned to the main
interface and dhcp_add() function returns prematurely.

By moving network subnet check we can ensure that --no-dhcp-interface is
properly generated for wwan interfaces.

Signed-off-by: Marcin Jurkowski <>
Signed-off-by: Hans Dedecker <> [PKG_RELEASE increase; move network checks]
22 months agobase-files: fix wan6 interface config generation for pppoe
Hans Dedecker [Mon, 18 Sep 2017 07:18:36 +0000 (09:18 +0200)]
base-files: fix wan6 interface config generation for pppoe

Setting ipv6 to auto in case of a pppoe interface will trigger the
creation of a dynamic wan_6 interface meaning two IPv6 interfaces
(wan6 and wan_6) will be active on top of the pppoe interface.
This leads to unpredictable behavior in the network; therefore set
ipv6 to 1 which will prevent the dynamic creation of the wan_6
Further alias the wan6 interface on top of the wan interface for pppoe
as the wan6 interface can only be started when the link local address is
ready. In case of pppoe the link local address is negotiated during the
Internet Protocol Control Protocol when the PPP link is setup meaning
all the IP address info is only available when the wan interface is up.

Signed-off-by: Hans Dedecker <>
22 months agokernel: update 4.9 to 4.9.50
Stijn Tintel [Mon, 18 Sep 2017 00:58:25 +0000 (03:58 +0300)]
kernel: update 4.9 to 4.9.50

Refresh patches.
Compile-tested on ipq8065/nbg6817 and x86/64.
Runtime-tested on ipq8065/nbg6817 and x86/64.

Fixes CVE-2017-1000251.

Signed-off-by: Stijn Tintel <>
[adapt qcom_nandc.c patches to match upstream changes, test ipq8065/nbg6817]
Signed-off-by: Stefan Lippers-Hollmann <>
22 months agokernel: update to 3.18.71
Florian Fainelli [Sat, 16 Sep 2017 23:41:48 +0000 (16:41 -0700)]
kernel: update to 3.18.71

Delete a bunch of fixes that are already included.
Refresh patches.
Compile-tested on malta/mipsel
Runtime-tested on malta/mipsel

Signed-off-by: Florian Fainelli <>
Acked-by: Hauke Mehrtens <>
22 months agotcpdump: noop commit to refer CVEs fixed in 4.9.2
Stijn Tintel [Sun, 17 Sep 2017 22:26:44 +0000 (01:26 +0300)]
tcpdump: noop commit to refer CVEs fixed in 4.9.2

When bumping tcpdump from 4.9.1 to 4.9.2, I did not include the fixed
CVEs in the commit message. As the list of fixed CVEs is quite long,
we should probably mention them in the changelogs of the releases to
come. This commit will make sure this happens.

The following CVEs were fixed in 21014d9708d586becbd62da571effadb488da9fc:


Signed-off-by: Stijn Tintel <>
22 months agomac80211: make iwlwifi select AC support
Hauke Mehrtens [Sat, 16 Sep 2017 23:35:00 +0000 (01:35 +0200)]
mac80211: make iwlwifi select AC support

Some NICs supported by this driver support ieee80211 AC.

Signed-off-by: Hauke Mehrtens <>
22 months agolinux-firmware: pack Intel iwl FW separately
Hauke Mehrtens [Sat, 16 Sep 2017 23:33:35 +0000 (01:33 +0200)]
linux-firmware: pack Intel iwl FW separately

Do not create one big package with all the Intel firmware files
supported by the iwlwifi driver, but use a separate package for each

This also updates some 7000 and 8000 series firmware files to more
recent version. The older versions shipped are not supported by the
current driver any more.

Signed-off-by: Hauke Mehrtens <>
22 months agomac80211: add dev_coredumpm() function
Hauke Mehrtens [Sun, 17 Sep 2017 12:50:01 +0000 (14:50 +0200)]
mac80211: add dev_coredumpm() function

dev_coredumpm() was added with kernel 4.7, but it is used by iwlwifi.
When the dev coredump framework form compat-wireless is used this is not
a problem because it already contains this, but this is deactivated if
the build system finds out that it is already included in the kernel we
compile against. This option was now activated by the bluetooth driver
btmrvl. Having dev coredump in the kernel adds about 400 bytes to the
lzma compressed kernel for brcm47xx.

This is copied from a more recent backports version to add the
dev_coredumpm() function when the internal core devdump is not used.

Fixes: a5922f6 ("kernel: bluetooth: add marvell sdio bluetooth module")
Signed-off-by: Hauke Mehrtens <>
22 months agokernel: kmod-btmrvl: Add kmod-mmc as dependency
Daniel Engberg [Sun, 17 Sep 2017 06:17:54 +0000 (08:17 +0200)]
kernel: kmod-btmrvl: Add kmod-mmc as dependency

This fixes the build of this module and should fix the build bots.

Fixes: a5922f6 ("kernel: bluetooth: add marvell sdio bluetooth module")
Signed-off-by: Daniel Engberg <>
[removed mveub dependency and update commit comment]
Signed-off-by: Hauke Mehrtens <>
22 months agogeneric: drop 704-phy-no-genphy-soft-reset.patch
Florian Fainelli [Sat, 16 Sep 2017 22:16:09 +0000 (15:16 -0700)]
generic: drop 704-phy-no-genphy-soft-reset.patch

4.4.80+ contains 71a165f6397df07a06ce643de5c2dbae29bd3cfb, 4.9.41+ contains
6c78197e4a69c19e61dfe904fdc661b2aee8ec20 which are all backports of upstream
commit 0878fff1f42c18e448ab5b8b4f6a3eb32365b5b6 ("net: phy: Do not perform
software reset for Generic PHY").

Our local patch is no longer needed, all this patch was doing was utilizing
gen10g_soft_reset which does nothing either, so just keep the code unchanged.

Signed-off-by: Florian Fainelli <>
22 months agodnsmasq: Pass TARGET_CPPFLAGS to Makefile
Florian Fainelli [Sat, 16 Sep 2017 23:27:37 +0000 (16:27 -0700)]
dnsmasq: Pass TARGET_CPPFLAGS to Makefile

With the introduction of the ubus notifications, we would now fail building
dnsmasq with external toolchains that don't automatically search for headers.
Pass TARGET_CPPFLAGS to the Makefile to resolve that.

Fixes: 34a206bc1194 ("dnsmasq: add ubus notifications for new leases")
Signed-off-by: Florian Fainelli <>
22 months agoarmvirt: Enable CONFIG_ARM_PMU
Florian Fainelli [Sat, 16 Sep 2017 21:49:37 +0000 (14:49 -0700)]
armvirt: Enable CONFIG_ARM_PMU

We will be prompted with this config symbol when performance monitoring is
enabled in the kernel.

Signed-off-by: Florian Fainelli <>
22 months agomvebu: WRT3200ACM: add bluetooth module
Henryk Heisig [Fri, 7 Jul 2017 07:53:12 +0000 (09:53 +0200)]
mvebu: WRT3200ACM: add bluetooth module

Signed-off-by: Henryk Heisig <>
22 months agokernel: bluetooth: add marvell sdio bluetooth module
Henryk Heisig [Fri, 7 Jul 2017 07:43:10 +0000 (09:43 +0200)]
kernel: bluetooth: add marvell sdio bluetooth module

This commit add support for Marvell bluetooth with SDIO interface.

Signed-off-by: Henryk Heisig <>
[Fix KCONFIG and FILES option]
Signed-off-by: Hauke Mehrtens <>
22 months agolinux-firmware: update to the commit from 2017-09-06
Henryk Heisig [Fri, 7 Jul 2017 07:35:20 +0000 (09:35 +0200)]
linux-firmware: update to the commit from 2017-09-06

update firmware mrvl/sd8887_uapsta.bin

Signed-off-by: Henryk Heisig <>
[update to version 2017-09-06]
Signed-off-by: Hauke Mehrtens <>
22 months agoutils/e2fsprogs: Update to 1.43.6
Daniel Engberg [Thu, 7 Sep 2017 22:44:26 +0000 (00:44 +0200)]
utils/e2fsprogs: Update to 1.43.6

Update e2fsprogs to 1.43.6
Disable compilation of fuse2fs (we don't package it)
Disable thread support (only affects fuse2fs)
Enable linking with libblkid instead of using private (included) version.
The libblkid is ~210KBytes in size, but with using the shared library
the binaries are ~25KBytes smaller. This also brings it in sync with
most other Linux distributions.

Signed-off-by: Daniel Engberg <>
22 months agokernel: add packaging for Xeon iTCO watchdog timer
Philip Prindeville [Fri, 15 Sep 2017 17:03:52 +0000 (11:03 -0600)]
kernel: add packaging for Xeon iTCO watchdog timer

Signed-off-by: Philip Prindeville <>
22 months agomwlwifi: update to version / 2017-08-10
Kabuli Chana [Thu, 10 Aug 2017 16:48:28 +0000 (10:48 -0600)]
mwlwifi: update to version / 2017-08-10

Update mwlwifi

Signed-off-by: Kabuli Chana <>
22 months agolibs/wolfssl: bump to version 3.12.0 ; add myself as maintainer
Alexandru Ardelean [Fri, 25 Aug 2017 13:22:36 +0000 (16:22 +0300)]
libs/wolfssl: bump to version 3.12.0 ; add myself as maintainer

Signed-off-by: Alexandru Ardelean <>
22 months agolibs/wolfssl: adjust symbol defaults against libwolfssl defaults
Alexandru Ardelean [Fri, 25 Aug 2017 11:23:42 +0000 (14:23 +0300)]
libs/wolfssl: adjust symbol defaults against libwolfssl defaults

Some symbols have been renamed.
Some are default enabled/disabled, so we need
to adjust semantics against that.

Signed-off-by: Alexandru Ardelean <>
22 months agolibs/wolfssl: disable hardening check in `settings.h`
Alexandru Ardelean [Fri, 25 Aug 2017 11:15:15 +0000 (14:15 +0300)]
libs/wolfssl: disable hardening check in `settings.h`

This seems to cause a false-positive warning/error
while building `libwebsockets-cyassl`.

make[6]: Leaving directory '/home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1'
make[6]: Entering directory '/home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1'
[  2%] Building C object CMakeFiles/websockets.dir/lib/base64-decode.c.o
In file included from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/wolfssl/ssl.h:31:0,
                 from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/cyassl/ssl.h:33,
                 from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/cyassl/openssl/ssl.h:30,
                 from /home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1/lib/private-libwebsockets.h:256,
                 from /home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1/lib/base64-decode.c:43:
/home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/wolfssl/wolfcrypt/settings.h:1642:14: error: #warning "For timing resistance / side-channel attack prevention consider using harden options" [-Werror=cpp]
             #warning "For timing resistance / side-channel attack prevention consider using harden options"


Hardening is enabled by default in libwolfssl at build-time.

However, the `settings.h` header is exported (along with other headers)
for build (via Build/InstallDev).

This looks like a small bug/issue with wolfssl.

Signed-off-by: Alexandru Ardelean <>
22 months agocyassl,curl,libustream-ssl: rename every `cyassl` to `wolfssl`
Alexandru Ardelean [Thu, 24 Aug 2017 05:56:40 +0000 (08:56 +0300)]
cyassl,curl,libustream-ssl: rename every `cyassl` to `wolfssl`

This is to eliminate any ambiguity about the cyassl/wolfssl lib.

The rename happened some time ago (~3+ years).
As time goes by, people will start to forget cyassl and
start to get confused about the wolfSSL vs cyassl thing.

It's a good idea to keep up with the times (moving forward).

Signed-off-by: Alexandru Ardelean <>
22 months agolibs/wolfssl: add libcysassl to PROVIDES field (for backwards compat)
Alexandru Ardelean [Fri, 25 Aug 2017 11:25:28 +0000 (14:25 +0300)]
libs/wolfssl: add libcysassl to PROVIDES field (for backwards compat)

Until other packages from feeds decide to rename the
dependency of `+libcyassl` to `+libwolfssl`, this allows
for a bit of backwards compatibility with those packages.

Signed-off-by: Alexandru Ardelean <>
22 months agowwan: json format in some modem definitions
Alexandru Ardelean [Mon, 4 Sep 2017 13:16:00 +0000 (16:16 +0300)]
wwan: json format in some modem definitions

Method used:
cd package/network/utils/wwan/files/data
sed -e 's/}}/}/g' -i *
sed -e 's/}\t"acm": 1/\t"acm": 1/g' -i *
sed -e 's/}\t"generic": 1/\t"generic": 1/g' -i *

Manually adjusted commas.
Validated with
for f in `ls` ; do echo $f ; python -m json.tool < $f || break ; done

Thanks to @lynxis for pointing out the commas.

Signed-off-by: Alexandru Ardelean <>
22 months agotools/e2fsprogs: Update to 1.43.6
Daniel Engberg [Fri, 8 Sep 2017 07:42:23 +0000 (09:42 +0200)]
tools/e2fsprogs: Update to 1.43.6

Update e2fsprogs to 1.43.6
* Remove FreeBSD patch as it's not needed, FreeBSD 9.1 is EoL and this
  is compiling on FreeBSD 11.1.
* Remove libmagic patch, RHEL 5 is EoL (End of Production Phase) since
  March 31, 2017.

Signed-off-by: Daniel Engberg <>
22 months agotools/expat: Update to 2.2.4
Daniel Engberg [Wed, 6 Sep 2017 09:56:15 +0000 (11:56 +0200)]
tools/expat: Update to 2.2.4

Update (lib)expat to 2.2.4

Signed-off-by: Daniel Engberg <>
22 months agotoolchain: gcc: update 7.x to 7.2.0
Kevin Darbyshire-Bryant [Tue, 22 Aug 2017 10:01:07 +0000 (11:01 +0100)]
toolchain: gcc: update 7.x to 7.2.0

Bump gcc from 7.1 to 7.2

Compile & run tested: ar71xx

Trace history of current patches and update with commit ref & comment
to give more clue as to why they're still around/needed.  Some have
changed form since the original commit but some clue is better than no
clue at all.

Signed-off-by: Kevin Darbyshire-Bryant <>
Tested-by: Koen Vandeputte <>
22 months agokernel: update 4.4 to 4.4.88
Kevin Darbyshire-Bryant [Fri, 15 Sep 2017 10:41:09 +0000 (11:41 +0100)]
kernel: update 4.4 to 4.4.88

Refresh patches.
Compile & run tested: ar71xx  Archer C7 v2

Signed-off-by: Kevin Darbyshire-Bryant <>
22 months agoodhcpd: don't enable server mode on non-static lan port
Karl Palsson [Fri, 1 Sep 2017 11:22:11 +0000 (11:22 +0000)]
odhcpd: don't enable server mode on non-static lan port

Instead of blindly enabling the odhcpd v6 server and RA server on the
lan port, only do that if the lan port protocol is "static"

This prevents the unhelpful case of a device being a dhcpv4 client and
v6 server on the same ethernet port.

Signed-off-by: Karl Palsson <>
[PKG_SOURCE_DATE increase; odhcpd.defaults script cleanup]
Signed-off-by: Hans Dedecker <>
22 months agoramips: fix used MAC addresses for Phicomm K2P
Jiawei Wang [Wed, 13 Sep 2017 06:04:22 +0000 (14:04 +0800)]
ramips: fix used MAC addresses for Phicomm K2P

The factory partition of the Phicomm K2P contains two MAC addresses.
The lower MAC address is at offset 0xe006 and the higher one is at
offset 0xe000.

Use the lower MAC address as base mac-address which the switch driver
increments by one for the second (wan) vlan.

The MAC addresses are still inverted in contrast to the stock firmware
where the lower MAC address is used for wan. But at least the use of a
MAC address not intended/reserved for this particular board is fixed.

Signed-off-by: Jiawei Wang <>
22 months agoodhcpd: update to git HEAD version
Hans Dedecker [Wed, 13 Sep 2017 20:18:08 +0000 (22:18 +0200)]
odhcpd: update to git HEAD version

f0bce9c dhcpv4: fix memset compile issue
0ba3278 dhcpv4: rework assignment lookup
e3b49f3 dhcpv4: cleanup dhcpv4_test usage
47fe122 dhcpv4: rework lease expire handling logic
028ab85 dhcpv4: force renew nonce authentication support
a827fca dhcpv4: avoid segfault when there's no IPv4 prefix
bea088b ndp: detect ifindex changes via interface netlink events
f66103e ubus: display accept reconf status for DHCPv6 assignments
f0e354b treewide: replace RELAYD prefix naming in macros
1a313f9 dhcpv4: fix possible segfault when lease is not created
e2d6eb4 dhcpv4: dhcpv4: move interface lease list insertion out of dhcpv4_assign

Signed-off-by: Hans Dedecker <>
22 months agoramips: fix D240 mini-PCIe power control GPIOs
Kristian Evensen [Tue, 12 Sep 2017 16:27:10 +0000 (18:27 +0200)]
ramips: fix D240 mini-PCIe power control GPIOs

In commit b11c51916cb9 ("ramips: Improve Sanlinking D240 config") I made
a mistake with regards GPIO numbering. And in addition to specifying the
wrong GPIO for controling the power of one of the mini-PCIe, I recently
discovered that the power of both slots can be controlled.

This patch specifies the correct GPIO for the left-most mini-PCIe slot
of the D240 (labeled power_mpcie2 since the slot is attached to SIM2),
and adds a GPIO that can be used to control the power of the other
mini-PCIe slot (labeled power_mpcie1).

Signed-off-by: Kristian Evensen <>
[do not use the gpio active macros for the gpio-export value]
Signed-off-by: Mathias Kresin <>
22 months agoramips: add support for TP-Link Archer C20 v1
Maxim Anisimov [Fri, 8 Sep 2017 09:39:17 +0000 (12:39 +0300)]
ramips: add support for TP-Link Archer C20 v1

TP-Link Archer C20 v1 is a router with 5-port FE switch and
non-detachable antennas. It's very similiar to TP-Link Archer C50.
Also it's based on MediaTek MT7620A+MT7610EN.

- MediaTek MT7620A (580 Mhz)
- 64 MB of RAM
- 8 MB of FLASH
- 2T2R 2.4 GHz and 1T1R 5 GHz
- 5x 10/100 Mbps Ethernet
- 2x external, non-detachable antennas
- UART (J1) header on PCB (115200 8n1)
- 8x LED (GPIO-controlled*), 2x button, power input switch
- 1 x USB 2.0 port

* WAN LED in this devices is a dual-color, dual-leads type which isn't
  (fully) supported by gpio-leds driver. This type of LED requires both
  GPIOs state change at the same time to select color or turn it off.
  For now, we support/use only the blue part of the LED.
* MT7610EN ac chip isn't not supported by LEDE. Therefore 5Ghz won't

Factory image notes:

These devices use version 3 of TP-Link header, fortunately without RSA
signature (at least in case of devices sold in Europe). The difference
lays in the requirement for a non-zero value in "Additional Hardware
Version" field. Ideally, it should match the value stored in vendor
firmware header on device.

We are able to prepare factory firwmare file which is accepted and
(almost) correctly flashed from the vendor GUI. As it turned out, it
accepts files without U-Boot image with second header at the beginning
but due to some kind of bug in upgrade routine, flashed image gets
corrupted before it's written to flash. So, to flash this device we must
to prepare image using original firmware from tp-link site with uboot.

Flash instruction:

Until (if at all) TP-Link fixes described problem, the only way to flash
LEDE image in these devices is to use tftp recovery mode in U-Boot.
There are two ways to flash the device to LEDE:

1) Using tftp mode with UART connection and original LEDE image

 - Place lede-ramips-mt7620-ArcherC20-squashfs-factory.bin in tftp
   server directory
 - Configure PC with static IP and tftp server.
 - Connect PC with one of LAN ports, power up the router and press
   key "4" to access U-Boot CLI.
 - Use the following commands to update the device to LEDE:

    setenv serverip
    tftp 0x80060000 lede-ramips-mt7620-ArcherC20-squashfs-factory.bin
    erase tplink 0x20000 0x7a0000
    cp.b 0x80060000 0x20000 0x7a0000

 - After that the device will reboot and boot to LEDE

2) Using tftp mode without UART connection but require some
   manipulations with target image

 - Download and unpack TP-Link Archer C20 v1 firmware from original web
 - Split uboot.bin from original firmware by this command (example):

    dd if=Archer_C20v1_0.9.1_4.0_up_boot(160427)_2016-04-27_13.53.59.bin of=uboot.bin bs=512 count=256 skip=1

 - Create ArcherC20V1_tp_recovery.bin using this command:

    cat uboot.bin lede-ramips-mt7620-ArcherC20-squashfs-factory.bin > ArcherC20V1_tp_recovery.bin

 - Place ArcherC20V1_tp_recovery.bin in tftp server directory.
 - Configure PC with static IP and tftp server.
 - Connect PC with one of LAN ports, press the reset button, power up
   the router and keep button pressed for around 6-7 seconds, until
   device starts downloading the file.
 - Router will download file from server, write it to flash and reboot.

Signed-off-by: Maxim Anisimov <>
22 months agogeneric: drop support for get_port_stats() on ar8xxx
Thibaut VARENE [Tue, 12 Sep 2017 21:22:37 +0000 (23:22 +0200)]
generic: drop support for get_port_stats() on ar8xxx

The implementation is not efficient on ar8xxx switches. It triggers high
CPU load and degrades device performance.

The high CPU load has been traced down to the ar8xxx_reg_wait() call in
ar8xxx_mib_op(), which has to usleep_range() till the MIB busy flag set
by the request to update the MIB counter is cleared.

This commit removes the get_port_stats() code introduced in 4d8a66d and
leaves a note for future hacker's beware.

Fixes: FS#1004

Signed-off-by: Thibaut VARENE <>
22 months agotreewide: fix shellscript syntax errors/typos
Lorenzo Santina [Mon, 11 Sep 2017 13:27:53 +0000 (15:27 +0200)]
treewide: fix shellscript syntax errors/typos

Fix multiple syntax errors in shelscripts (of packages only)
These errors were causing many conditions to not working properly

Signed-off-by: Lorenzo Santina <>
[increase PKG_RELEASE, drop command substitution from]
Signed-off-by: Mathias Kresin <dev@kresin.em>
22 months agoramips: fix hg255d LED status support
David Yang [Sat, 9 Sep 2017 13:16:11 +0000 (21:16 +0800)]
ramips: fix hg255d LED status support

Use the green power LED for boot status indication.

Source: Para 3

Signed-off-by: David Yang <>
22 months agobasefiles: allow suid coredumps
Kevin Darbyshire-Bryant [Mon, 4 Sep 2017 12:13:24 +0000 (13:13 +0100)]
basefiles: allow suid coredumps

Set sysctl fs.suid_dumpable = 2

This allows suid processes to dump core according to kernel.core_pattern
setting.  LEDE typically uses suid to drop root priviledge rather than
gain it but without this setting any suid process would be unable to
produce coredumps (e.g. dnsmasq)

Processes still need to set a non zero core file process limit ('ulimit
-c unlimited' or if procd used 'procd_set_param limits
core="unlimited"') in order to produce a core.  This setting removes an
obscure stumbling block along the way.



This value can be used to query and set the core dump mode for setuid
or otherwise protected/tainted binaries. The modes are

0 - (default) - traditional behaviour. Any process which has changed
privilege levels or is execute only will not be dumped.
1 - (debug) - all processes dump core when possible. The core dump is
owned by the current user and no security is applied. This is
intended for system debugging situations only. Ptrace is unchecked.
This is insecure as it allows regular users to examine the memory
contents of privileged processes.
2 - (suidsafe) - any binary which normally would not be dumped is dumped
anyway, but only if the "core_pattern" kernel sysctl is set to
either a pipe handler or a fully qualified path. (For more details
on this limitation, see CVE-2006-2451.) This mode is appropriate
when administrators are attempting to debug problems in a normal
environment, and either have a core dump pipe handler that knows
to treat privileged core dumps with care, or specific directory
defined for catching core dumps. If a core dump happens without
a pipe handler or fully qualifid path, a message will be emitted
to syslog warning about the lack of a correct setting.

Signed-off-by: Kevin Darbyshire-Bryant <>
Signed-off-by: Hans Dedecker <> [PKG_RELEASE increase]
22 months agoar71xx: fix MAC addresses on TP-Link TL-WR1043ND v4
Matthias Schiffer [Mon, 11 Sep 2017 17:41:41 +0000 (19:41 +0200)]
ar71xx: fix MAC addresses on TP-Link TL-WR1043ND v4

The addresses were read from the 'config' partition, which would not always
contain the addresses at the same offsets, depending on the stock firmware
version used before flashing LEDE. Change this to get the addresses from
the 'product-info' partition, which is read-only.

Reported-and-tested-by: Andreas Ziegler <>
Signed-off-by: Matthias Schiffer <>
22 months agoramips: Add support for ZBT WE1026-5G
Kristian Evensen [Sun, 10 Sep 2017 12:44:47 +0000 (14:44 +0200)]
ramips: Add support for ZBT WE1026-5G

The ZBT WE1026-5G
( is the follow-up
to the ZBT WE1026 and is based on MT7620. For the previous WE1026, the
ZBT WE826 image could be used. However, as the name implies, the -5G
comes equipped with a 5GHz wifi radio. As the WE826 only has a 2.4GHz
radio, the addition of 5GHz means that a separate image is needed for
the WE1026-5G. I suspect that this image will also work on the previous
WE1026, but I don't have a device to test with.

The WE1026-5G has following specifications:
* CPU: MT7620A
* 1x 10/100Mbps Ethernet.
* 16 MB Flash.
* 64 MB RAM.
* 1x USB 2.0 port.
* 1x mini-PCIe slots.
* 1x SIM slots.
* 1x 2.4Ghz WIFI.
* 1x 5GHz wifi (MT7612)
* 1x button.
* 3x controllable LEDs.

* Wifi.
* Switch.
* mini-PCIe slot. Only tested with a USB device (a modem).
* SIM slot.
* Sysupgrade.
* Button (reset).

Not working:
* The 5GHz WIFI LED is completely dead. I suspect the issue is the same
as on other devices with Mediatek 5Ghz wifi-cards/chips. The LED is
controlled by the driver, and mt76 (currently) does not support this.

Not tested:
* SD card reader.

* The modem (labeled 3G/4G) and power LEDs are controlled by the
* There is a 32MB version of this device available, but I do not have
access to it. I have therefor only added support for the 16MB version,
but added all the required infrastructure to make adding support for the
32MB version easy.

The router comes pre-installed with OpenWRT, including a variant of
Luci. The initial firmware install can be done through this UI,
following normal procedure. I.e., access the UI and update the firmware
using the sysupgrade-image. Remember to select that you do not want to
keep existing settings.

If you brick the device, the WE1026-5G supports recovery using HTTP. Keep the
reset button pressed for ~5sec when booting to start the web server. Set the
address of the network interface on your machine to, and
point your browser to to access the recovery UI. From the
recovery UI you can upload a firmware image.

Signed-off-by: Kristian Evensen <>
22 months agoath10k: Re-enable intermediate softqueues for all devices
Toke Høiland-Jørgensen [Wed, 6 Sep 2017 12:53:19 +0000 (14:53 +0200)]
ath10k: Re-enable intermediate softqueues for all devices

The upstream ath10k driver disables the intermediate softqueues for some
devices. This patch reverts that behaviour and always enables the
softqueues (and associated bufferbloat fixes). We have had reports of people
running this with good results:

This also refreshes mac80211 patches.

Signed-off-by: Toke Høiland-Jørgensen <>
22 months agoscripts/ fail loudly if provided hash is unsupported
Baptiste Jonglez [Sun, 3 Sep 2017 12:01:28 +0000 (14:01 +0200)]
scripts/ fail loudly if provided hash is unsupported

Currently, if the provided hash is unsupported (length different from 32
or 64 bytes), we happily download the requested file without any kind of
checksum verification.

This is quite dangerous and may provide a false sense of security, because
a single typo in the hash (e.g. one character deleted by mistake) may skip
checksum verification entirely.

Instead, fail immediately if we don't support the provided hash.
In particular, if an external package repository decides to change the
hash algorithm one day, we will now fail loudly instead of skipping
checksum verification without complaints.

Note: if some users of scripts/ knowingly provide an empty hash
because they don't need checksum verification, this change will break
them.  This does not seem to be the case currently, but if this feature is
ever needed, an option should be added to instead of relying
on the hash being empty.

Fixes: eaa4eba10a89 ("scripts/ add SHA-256 support")

Signed-off-by: Baptiste Jonglez <>
22 months agoar71xx: Add GRO support to ag71xx
Rosen Penev [Sun, 3 Sep 2017 18:35:46 +0000 (11:35 -0700)]
ar71xx: Add GRO support to ag71xx

On a TL-WN710N, this patch increases iperf performance from ~92.5 to ~93.5 mbps. Keep in mind the WN710N is a 100mbps device. I expect greater numbers from gigabit devices.

Signed-off-by: Rosen Penev <>
22 months agoramips: Change ethernet driver to use napi_complete_done.
Rosen Penev [Thu, 7 Sep 2017 02:57:41 +0000 (19:57 -0700)]
ramips: Change ethernet driver to use napi_complete_done.

Backport of mailine linux commit. Speeds up ethernet slightly and reduces latency.

Signed-off-by: Rosen Penev <>
22 months agocns3xxx: fix GPIO controller interrupt enable
Tim Harvey [Fri, 8 Sep 2017 18:56:59 +0000 (11:56 -0700)]
cns3xxx: fix GPIO controller interrupt enable

The cns3xxx interrupt controller uses a single register and as such
the 'mask' reg/functions must be used as opposed to the 'enable'/'disable'

This fixes an issue that occurs if more than one GPIO on a specific controller
(there is GPIOA and GPIOB each having 32 GPIO's) uses interrupts. When one
would get enabled all others would be disabled prior to this patch.

Signed-off-by: Tim Harvey <>
Acked-by: Koen Vandeputte <>
22 months agokernel: update 4.9 to 4.9.49
Stijn Tintel [Sun, 10 Sep 2017 21:37:48 +0000 (23:37 +0200)]
kernel: update 4.9 to 4.9.49

Refresh patches.
Compile-tested on octeon and x86/64.
Runtime-tested on octeon and x86/64.

Fixes CVE-2017-11600.

Signed-off-by: Stijn Tintel <>
22 months agostrace: bump to 4.19
Stijn Tintel [Sun, 10 Sep 2017 20:38:07 +0000 (22:38 +0200)]
strace: bump to 4.19

Signed-off-by: Stijn Tintel <>