lede.git
7 years agolantiq: remove lantiq_board_name, use the generic function instead
Felix Fietkau [Fri, 17 Mar 2017 15:21:30 +0000 (16:21 +0100)]
lantiq: remove lantiq_board_name, use the generic function instead

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agolantiq: move lantiq_board_detect() to 03_preinit_board.sh
Felix Fietkau [Fri, 17 Mar 2017 15:17:42 +0000 (16:17 +0100)]
lantiq: move lantiq_board_detect() to 03_preinit_board.sh

It is only used there

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agobase-files: add generic board_name function to functions.sh
Felix Fietkau [Fri, 17 Mar 2017 15:12:27 +0000 (16:12 +0100)]
base-files: add generic board_name function to functions.sh

This will be used to replace all those nasty board specific scripts
that do basically the same thing

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agobusybox: enable the "ip neigh" command (FS#618)
Felix Fietkau [Thu, 16 Mar 2017 15:54:16 +0000 (16:54 +0100)]
busybox: enable the "ip neigh" command (FS#618)

It is useful for creating static ARP entries.
Adds about 1.9 KiB after gzip on MIPS.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agomac80211: Fix race condition leading to wifi interfaces not coming up at boot sometimes.
Vittorio Gambaletta [Tue, 14 Mar 2017 13:45:15 +0000 (14:45 +0100)]
mac80211: Fix race condition leading to wifi interfaces not coming up at boot sometimes.

In the drv_mac80211_setup function, mac80211_interface_cleanup
is called to ask the kernel to delete all existing interfaces
for the phy that is being configured via netlink.

Later in the first function, mac80211_prepare_vif is called to
set up the new interfaces as required.

But sometimes, when mac80211_prepare_vif (and so the relevant
`iw phy x interface add y` command) runs, the kernel might still
be cleaning up the old interface with the same ifname. It usually
takes very few time to do that; possibly a few milliseconds of
sleep in the script after detecting this error condition could be
enough, but the busybox sh does not support sub-second sleep
intervals.

When this happens, iw obviously fails to create the new interface;
and the following message is printed in the system log, followed by
subsequent failure messages from hostapd in case this would have been
an AP interface.

Tue Mar 14 04:21:57 2017 daemon.notice netifd: radio1 (2767): command failed: Too many open files in system (-23)

This was a long-standing issue existing since at least OpenWrt Backfire,
and today I finally managed to debug and (hopefully) solve it.
It was happening very few times on most devices; but it was happening
a lot more frequently on fast platforms with multiple radios, such as
the powerpc-based dual-ath9k-radio tl-wdr4900-v1.

Signed-off-by: Vittorio Gambaletta <openwrt@vittgam.net>
7 years agoipq806x: remove v4.4 support
John Crispin [Wed, 22 Mar 2017 08:43:01 +0000 (09:43 +0100)]
ipq806x: remove v4.4 support

Signed-off-by: John Crispin <john@phrozen.org>
7 years agoipq806x: set v4.9 as default
John Crispin [Wed, 15 Mar 2017 10:49:43 +0000 (11:49 +0100)]
ipq806x: set v4.9 as default

Signed-off-by: John Crispin <john@phrozen.org>
7 years agofirmware: add custom IPQ wifi board definitions
Christian Lamparter [Wed, 15 Mar 2017 10:49:14 +0000 (11:49 +0100)]
firmware: add custom IPQ wifi board definitions

On the ath10k-devel ML Michael Kazior stated:

"board-2 is a key-value store of actual board files.
Some devices, notably qca61x4 hw3+ and qca4019 need
distinct board files to be uploaded. Otherwise they
fail in various ways." [0].

Later on Rajkumar Manoharan explained:

"In QCA4019 platform, only radio specific calibration
(pre-cal-data) is stored in flash. Board specific contents
are read from board-2.bin. For each radio appropriate board
data should be loaded. To fetch correct board data from
board-2.bin bundle, pre-cal/radio specific caldata should
be loaded first to get proper board id.

|My understanding until now was that:
|
| * pre-cal data + board-2.bin info == actual calibration data

Correct." [1].
The standard board-2.bin from the ath10k-firmware-qca4019
barely works on the RT-AC58U. Especially 5GHz clients fail
to connect at all and if they do, they have very low
throughput even right next to the router.

Currently, the solution for this problem is to supply a
custom board-2.bin for every device.

To implement this feature, this method makes use of:
Rafał Miłecki's "base-files: add support for overlaying
rootfs content". This comes with a few limitations:
1. Since there can only be one board-2.bin at the right
   location, there can only one board overwrite installed
   at any time. (All packages CONFLICT with each other.
   It's also not possible to "builtin" multiple package.)

2. updating ath10k-firmware-qca4019 will also replace
   the board-2.bin. For this cases the user needs to
   manually reinstall the wifi-board package once the
   ath10k-firmware-qca4019 is updated.

To create the individual board-2.bin: Use the ath10k-bdencoder
utility from the qca-swiss-army-knife repository:
<https://github.com/qca/qca-swiss-army-knife>
The raw board.bin files have to be extracted from the
vendor's source GPL.tar archieves.

Signed-off-by: Alexis Green <agreen@cococorp.com>
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
7 years agofirmware: add IPQ fritz-tools FW extraction tool
Christian Lamparter [Wed, 15 Mar 2017 10:48:56 +0000 (11:48 +0100)]
firmware: add IPQ fritz-tools FW extraction tool

Adds tools to access tffs3 and pre-calibration data required for ipq4019
based FritBoxes.

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
7 years agoipq806x: make the dwc3 driver and required phy drivers built-in
John Crispin [Wed, 22 Mar 2017 08:36:00 +0000 (09:36 +0100)]
ipq806x: make the dwc3 driver and required phy drivers built-in

Signed-off-by: John Crispin <john@phrozen.org>
7 years agoipq806x: enable QCE hardware crypto inside the kernel
John Crispin [Wed, 22 Mar 2017 08:29:11 +0000 (09:29 +0100)]
ipq806x: enable QCE hardware crypto inside the kernel

Signed-off-by: John Crispin <john@phrozen.org>
7 years agoipq806x: add ipq4019 fritz4040 support
Christian Lamparter [Tue, 14 Mar 2017 16:04:28 +0000 (17:04 +0100)]
ipq806x: add ipq4019 fritz4040 support

This patch adds support for AVM FRITZ!Box 4040.

hardware highlights:

SOC: IPQ4018 / QCA Dakota
CPU: Quad-Core ARMv7 Processor rev 5 (v7l) Cortex-A7
DRAM: 256 MiB Nanya NT5CC128M16IP
FLASH: 32 MiB MXIC MX25L25635FMI
ETH: Qualcomm Atheros QCA8075 Gigabit Switch (4 x LAN, 1 x WAN)
USB: 1 x 3.0 (via Synopsys DesignWare DWC3 controller in the SoC)
1 x 2.0 (via Synopsys DesignWare DWC3 controller in the SoC)
WLAN1: Qualcomm Atheros QCA4018 2.4GHz 802.11bgn 2:2x2
WLAN2: Qualcomm Atheros QCA4018 5GHz 802.11a/n/ac 2:2x2
INPUT: one WLAN and one WPS button
LEDS: Power, WAN/Internet, WIFI, INFO (red and amber) and LAN.
Serial:
WARNING: The serial port needs a TTL/RS-232 v3.3 level converter!
        The Serial setting is 115200-8-N-1. The SoC's serial port is right
next to the MXIC FLASH chip. The board has a unpopulated 1x4 0.1"
header for it. Use a multimeter to figure out the pinout!

This board currently needs an additional u-boot image in order to boot
properly. Booting with EVA isn't possible ATM.

Install Procedure:
 0. It's highly recommended to connect to the serial port.
    The serial settings are listed above.
 1. install a u-boot image for AVM Fritz!Box 4040
    (see <https://github.com/chunkeey/FritzBox-4040-UBOOT/releases> and
    <https://github.com/chunkeey/FritzBox-4040-UBOOT/blob/master/upload-to-f4040.sh>)
 2. upload the initramfs.itb image via tftp (u-boot listens to
    192.168.1.1 - use binary transfer mode!)
 3. connect to the FB4040 and use sysupgrade sysupgrade.bin
    to install the image.

Works:
- Switch and Ethernet (99%)
        - Buttons (WLAN, WPS)
        - FLASH (1 x 32MiB NOR Chip)
        - WLAN2G and WLAN5G
        - CPUFREQ scaling
        - PRNG
        - serial
        - Crypto Accelerator
        - sysupgrade (Read the flash instructions to avoid bricking)
        - full LEDE Install (Read the flash instructions to avoid bricking)
- LEDs (Power, WAN, Info (red and amber), LAN)
          The LEDs are connected to the QCA8075 LED ports.
  The AR40xx driver contains a gpio-controller to
  handle these special "GPIOs".
        - USB Both 3.0 and 2.0 ports
        - many packages from other ARMv7 boards
          (This does include the RaspberryPi Model 2!)
        - ...

Not planned:
- WAN<->LAN short-cut
- Qualcomm Secure Execution Environment
        - ...

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John Crispin <john@phrozen.org>
7 years agoipq806x: add ipq4019 support
Christian Lamparter [Tue, 14 Mar 2017 16:02:20 +0000 (17:02 +0100)]
ipq806x: add ipq4019 support

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John Crispin <john@phrozen.org>
7 years agouboot-fritz4040: add an ipq806x u-boot for FritzBox 4040
Felix Fietkau [Tue, 14 Mar 2017 09:19:17 +0000 (10:19 +0100)]
uboot-fritz4040: add an ipq806x u-boot for FritzBox 4040

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoprocd: nand: make it possible to configure kernel and ubi partition
Christian Lamparter [Sun, 20 Nov 2016 02:33:03 +0000 (03:33 +0100)]
procd: nand: make it possible to configure kernel and ubi partition

With this patch, the nand.sh script no longer overwrites
CI_KERNELPART and CI_UBIPART if they have been set before.

This allows platforms specify alternative values for these
variables.

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
7 years agomac80211: enable ath10k AHB support for QCA4019
Christian Lamparter [Mon, 14 Nov 2016 03:57:38 +0000 (04:57 +0100)]
mac80211: enable ath10k AHB support for QCA4019

This patch enables the ATH10K_AHB support for the QCA4019
devices on the AHB bus.

This patch also removes 936-ath10k_skip_otp_check.patch
because it breaks the AHB device identification.
"Patch is wrong. I find it frustrating OpenWRT/LEDE doesn't
try to work with upstream on ixing these things right."

[1] <https://www.mail-archive.com/ath10k@lists.infradead.org/msg05896.html>

It also limits ath10k memory hunger (This is a problem with 128MiB RAM)

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
7 years agoath10k-firmware: firmware package for QCA4019 devices
Christian Lamparter [Wed, 9 Nov 2016 16:05:42 +0000 (17:05 +0100)]
ath10k-firmware: firmware package for QCA4019 devices

This patch adds a new firmware package to ath10k-firmware
called ath10k-firmware-qca4019.

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
7 years agoath10k-firmware: bump to latest version
Christian Lamparter [Wed, 9 Nov 2016 16:05:42 +0000 (17:05 +0100)]
ath10k-firmware: bump to latest version

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
7 years agoinclude: add KERNEL_LOAD_ADDR to TARGET_VARS
John Crispin [Wed, 15 Mar 2017 06:40:08 +0000 (07:40 +0100)]
include: add KERNEL_LOAD_ADDR to TARGET_VARS

This will allow us to override the variable from within a Device template.

Signed-off-by: John Crispin <john@phrozen.org>
7 years agoopenvpn: add extra respawn parameters
Martin Schiller [Thu, 16 Mar 2017 10:25:53 +0000 (11:25 +0100)]
openvpn: add extra respawn parameters

This change protects the openvpn instances to be marked as "in a crash
loop" and thereby the connection retries will run infinitely.

When the remote site of an openvpn connection goes down for some time
(network failure etc.) the openvpn instance in an openwrt/lede device
should not stop retrying to establish the connection.

With the current limit of 5 retries, there is a user interaction
required, which isn't really what you want when the device should
simply do everything to keep the vpn connection up.

Signed-off-by: Martin Schiller <ms@dev.tdt.de>
7 years agoutils/tcpdump: Rework URLs
Daniel Engberg [Wed, 22 Mar 2017 07:01:04 +0000 (08:01 +0100)]
utils/tcpdump: Rework URLs

Add actual mirror and use main site as last resport
Source: http://www.tcpdump.org/mirrors.html

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agolibs/libpcap: Rework URLs
Daniel Engberg [Wed, 22 Mar 2017 07:00:40 +0000 (08:00 +0100)]
libs/libpcap: Rework URLs

Add mirror and use main site as last resort.
Source: http://www.tcpdump.org/mirrors.html

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agoutils/util-linux: Update to 2.29.2
Daniel Engberg [Mon, 20 Mar 2017 08:51:26 +0000 (09:51 +0100)]
utils/util-linux: Update to 2.29.2

Update util-linux to 2.29.2
Remote 0001-fix-uClibc-ng-scanf-check.patch as it's been merged upstream.
Refresh patches
Change ncurses to ncursesw to fix compilation and avoid confusion

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agolibs/openssl: Refresh mirror list
Daniel Engberg [Wed, 22 Mar 2017 06:31:20 +0000 (07:31 +0100)]
libs/openssl: Refresh mirror list

Refresh mirror list, some doesn't offer OpenSSL and add main site as last resort.
Source: https://www.openssl.org/source/mirror.html

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agokernel: fix builds with kmod-drm-vc4 and kmod-backlight enabled
Rafał Miłecki [Wed, 22 Mar 2017 07:47:43 +0000 (08:47 +0100)]
kernel: fix builds with kmod-drm-vc4 and kmod-backlight enabled

This fixes:
JDI LT070ME05000 WUXGA DSI panel (DRM_PANEL_JDI_LT070ME05000) [N/m/?] (NEW) aborted!

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agobrcm2708: add kmod-drm-vc4 package
Rafał Miłecki [Tue, 21 Mar 2017 07:49:15 +0000 (08:49 +0100)]
brcm2708: add kmod-drm-vc4 package

It includes vc4 driver for Broadcom VideoCore IV GPU.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agoumdns: update to the version 2017-03-21
Rafał Miłecki [Tue, 21 Mar 2017 21:57:56 +0000 (22:57 +0100)]
umdns: update to the version 2017-03-21

This includes following changes:
480d7bc Fix sending unicast questions on cache expire
a0403cd Keep source sockaddr for every cached DNS record
1478293 Fix code freeing cached non-A(AAA) records too early
9f1cc22 Fix replying to "QU" questions received on unicast interface
943bedb Fix reading port of incoming packets
c725494 Use MCAST_PORT define for port 5353
ce7e9e9 Use one define for DNS-Based Service Discovery service name
e1bacef Drop entries cached for interface we're going to delete
496aeba Fix comment typo in cache_gc_timer
f89986b Fix refreshing cached A(AAA) records that expire

Previous updates made umdns work as expected on startup but there were
still many bugs. They were mostly related to runtime - cache management
and requests + responses. E.g. umdns was never able to send question on
DNS record expire. It was also ignoring all incoming unicast questions.

Since these issues are quite serious it makes sense to backport this
update to the stable branch.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agonetifd: update to git HEAD version
Hans Dedecker [Tue, 21 Mar 2017 21:54:47 +0000 (22:54 +0100)]
netifd: update to git HEAD version

a032166 interface-ip: set prefix indicator flag when IPv6 prefix lifetime changes
b4f8984 system-linux: parse vti specific settings as nested json data object
7e3b89a system-linux: parse gre specific settings as nested json data object

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
7 years agovti: add vti specific settings as nested json object
Hans Dedecker [Tue, 14 Mar 2017 21:39:43 +0000 (22:39 +0100)]
vti: add vti specific settings as nested json object

Add vti specific settings ikey and okey as a nested data json object

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
7 years agogre: add gre specific settings as nested json object
Hans Dedecker [Tue, 14 Mar 2017 21:34:18 +0000 (22:34 +0100)]
gre: add gre specific settings as nested json object

Add gre specific settings ikey, okey, iseqno, oseqno, icsum
and ocsum as a nested data json object

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
7 years agomac80211: brcmfmac: backport patches pushed on 2017-03-20
Rafał Miłecki [Tue, 21 Mar 2017 09:45:30 +0000 (10:45 +0100)]
mac80211: brcmfmac: backport patches pushed on 2017-03-20

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agokernel: update bcma including recent changes for 4.12
Rafał Miłecki [Tue, 21 Mar 2017 08:26:02 +0000 (09:26 +0100)]
kernel: update bcma including recent changes for 4.12

This adds support for describing GPIO chips placed on PCIe cards. Thanks
to this we get working 2.4 GHz LED on Tenda AC9.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agokernel: move disabling DRM symbols to the config file
Rafał Miłecki [Tue, 21 Mar 2017 07:21:34 +0000 (08:21 +0100)]
kernel: move disabling DRM symbols to the config file

This is the standard way we handle this. Please note (it seems) I could
drop few symbols as they are hidden under (disabled) DRM_LEGACY now.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agokernel: video: add kmod-i2c-core dependency to kmod-drm
Rafał Miłecki [Sun, 19 Mar 2017 19:26:34 +0000 (20:26 +0100)]
kernel: video: add kmod-i2c-core dependency to kmod-drm

For targets with i2c not built-in this fixes following error:
Package kmod-drm is missing dependencies for the following libraries:
i2c-core.ko

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agokernel: disable CONFIG_BCM_IPROC_ADC by default
Rafał Miłecki [Tue, 21 Mar 2017 06:12:56 +0000 (07:12 +0100)]
kernel: disable CONFIG_BCM_IPROC_ADC by default

This fixes compiling issue on bcm53xx target with kmod-iio-core:
Broadcom IPROC ADC driver (BCM_IPROC_ADC) [N/m/?] (NEW)

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agoinclude: Add nomips16 CPU_SUBTYPE
Hauke Mehrtens [Tue, 14 Mar 2017 17:55:03 +0000 (18:55 +0100)]
include: Add nomips16 CPU_SUBTYPE

This can be used to indicate that a target does not support the optional mips16
extension even when it is a mips32r2 or later CPU.

This will generate a separate toolchain and a separate package folder,
e.g. mips_24kc_nomips16

Signed-off-by: Hauke Mehrtens <hauke.mehrtens@intel.com>
7 years agoscripts: only generate config from feature flag if fully match
Hauke Mehrtens [Tue, 14 Mar 2017 17:55:02 +0000 (18:55 +0100)]
scripts: only generate config from feature flag if fully match

Without this change the code checked if the string was contained in the
feature option and not if the string matches the complete word. This only
removes the nand option from the omap24xx target, the other changes are
only removing options which were added twice.

Signed-off-by: Hauke Mehrtens <hauke.mehrtens@intel.com>
7 years agotools/upx: Remove from repo
Daniel Engberg [Sun, 12 Mar 2017 18:58:08 +0000 (19:58 +0100)]
tools/upx: Remove from repo

Remove upx from repo, it hasn't been used for a very long time and serves little to no purpose.

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agolzo: Update to 2.10
Daniel Engberg [Sun, 12 Mar 2017 18:43:50 +0000 (19:43 +0100)]
lzo: Update to 2.10

Update lzo to 2.10

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agoiperf3: Update to 3.1.7
Daniel Engberg [Sun, 12 Mar 2017 19:48:01 +0000 (20:48 +0100)]
iperf3: Update to 3.1.7

Update iperf3 to 3.1.7

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agotools/coreutils: Update to 8.27
Daniel Engberg [Tue, 14 Mar 2017 06:53:38 +0000 (07:53 +0100)]
tools/coreutils: Update to 8.27

Update coreutils to 8.27

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agotools/libressl: Update to 2.5.1
Daniel Engberg [Tue, 14 Mar 2017 06:46:39 +0000 (07:46 +0100)]
tools/libressl: Update to 2.5.1

Update libressl to 2.5.1

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agotools/dosfstools: Update to 4.1
Daniel Engberg [Tue, 14 Mar 2017 07:23:10 +0000 (08:23 +0100)]
tools/dosfstools: Update to 4.1

Update to 4.1
Remove 0001-mkfs-Default-to-64-32-heads-sectors-for-targets-smal.patch
as it's included in upstream repo.
Refresh patch

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agotoolchain/binutils: Add binutils 2.28
Daniel Engberg [Mon, 13 Mar 2017 23:38:59 +0000 (00:38 +0100)]
toolchain/binutils: Add binutils 2.28

Add binutils 2.28 to repo

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agopackage/boot: update uboot-kirkwood to 2017.03
Paul Wassi [Wed, 15 Mar 2017 06:37:48 +0000 (07:37 +0100)]
package/boot: update uboot-kirkwood to 2017.03

Bump uboot-kirkwood to upstream release 2017.03

Signed-off-by: Paul Wassi <p.wassi@gmx.at>
7 years agoramips: clean kernel config symbol conflicts for MT7688 PWM
Furong Xu [Wed, 15 Mar 2017 08:47:59 +0000 (16:47 +0800)]
ramips: clean kernel config symbol conflicts for MT7688 PWM

pwm-mediatek should be selected in Other Modules submenu, not default to yes.

Signed-off-by: Furong Xu <xfr@outlook.com>
7 years agotools/sed: Update to 4.4
Daniel Engberg [Wed, 15 Mar 2017 07:52:29 +0000 (08:52 +0100)]
tools/sed: Update to 4.4

Update sed to 4.4

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agoramips: fix switch port mapping for ZBT-WE2026
Vaclav Svoboda [Sat, 18 Mar 2017 16:11:14 +0000 (17:11 +0100)]
ramips: fix switch port mapping for ZBT-WE2026

This patch fixes switch port mapping for Zbtlink ZBT-WE2026.

Signed-off-by: Vaclav Svoboda <svoboda@neng.cz>
7 years agoipq806x: support independent core clocks in cpufreq with kernel 4.9+
Hannu Nyman [Thu, 16 Mar 2017 13:32:16 +0000 (15:32 +0200)]
ipq806x: support independent core clocks in cpufreq with kernel 4.9+

Add back support for the independent_clocks definition that has been
removed between kernel 4.4 and 4.9 by upstream commits
eb96924acddc709db58221c210ca05cd9effb1df and
e86eee6bc2aaa6b3637f6497b26beee09a91bde9

* extend the new cpufreq_dt_platform_data definition in cpufreq-dt.h
* revert the removal of its usage in cpufreq-dt.c
* use new cpufreq-dt.h in qcom-cpufreq.c

Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
7 years agoipq806x: route gpio interrupts to APPS processor through scm firmware
Pavel Kubelun [Sun, 19 Mar 2017 11:59:26 +0000 (07:59 -0400)]
ipq806x: route gpio interrupts to APPS processor through scm firmware

For IPQ806x targets, TZ protects the registers that are used to
configure the routing of interrupts to a target processor.
To resolve this, this patch uses scm call to route GPIO interrupts
to application processor. Also the scm call interface is changed.

Signed-off-by: Pavel Kubelun <be.dissent@gmail.com>
7 years agoipq806x: remove scm firmware clocks
Pavel Kubelun [Sun, 19 Mar 2017 01:24:45 +0000 (04:24 +0300)]
ipq806x: remove scm firmware clocks

At the moment as a workaround definition for scm firmware in DT is used as if it is
apq8064 board. This leads to incomplete scm firmware initialization and as a result
cpuidle driver fails to configure.

By design unlike other qcom boards ipq do not use clocks to connect to scm.

Considering this we're removing from DT and scm driver clocks for ipq boards.

As a result cpuidle does not produce errors about failed configuration anymore.

Signed-off-by: Pavel Kubelun <be.dissent@gmail.com>
7 years agoiproute2: fix ip monitor can't work when NET_NS is not enabled
Yousong Zhou [Fri, 17 Mar 2017 09:49:29 +0000 (17:49 +0800)]
iproute2: fix ip monitor can't work when NET_NS is not enabled

The bug appeared in v4.1.0 and was fixed since v4.8.0

Fixes FS#620

Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
7 years agobcm53xx: switch to kernel 4.9
Rafał Miłecki [Sat, 18 Mar 2017 22:00:00 +0000 (23:00 +0100)]
bcm53xx: switch to kernel 4.9

With iproc regression fixed and GIC interrupts in DTS updated it seems
we can safely start using 4.9.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agobuild: fix high cpu usage / hang in prereq-build.mk
Felix Fietkau [Sat, 18 Mar 2017 20:50:19 +0000 (21:50 +0100)]
build: fix high cpu usage / hang in prereq-build.mk

host-build.mk should not be included for prereq

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agobcm53xx: backport BCM5301X patches
Rafał Miłecki [Sat, 18 Mar 2017 17:31:36 +0000 (18:31 +0100)]
bcm53xx: backport BCM5301X patches

This fixes GIC interrupts (required before switching to 4.9), adds few
new entires & introduces DTS for Archer C5.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agodnsmasq: don't point --resolv-file to default location unconditionally
Philip Prindeville [Tue, 14 Mar 2017 18:58:37 +0000 (12:58 -0600)]
dnsmasq: don't point --resolv-file to default location unconditionally

If noresolv is set, we should not generate a --resolv-file parameter.

Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [minor cleanup]
7 years agobuild: add devel option to store build config in firmware
Vitaly Chekryzhev [Thu, 5 Jan 2017 07:06:26 +0000 (12:06 +0500)]
build: add devel option to store build config in firmware

Store config.seed in firmware /etc/build.config

Signed-off-by: Vitaly Chekryzhev <13hakta@gmail.com>
7 years agobuild: prepare config.seed before package compilation
Vitaly Chekryzhev [Fri, 6 Jan 2017 11:47:18 +0000 (16:47 +0500)]
build: prepare config.seed before package compilation

Signed-off-by: Vitaly Chekryzhev <13hakta@gmail.com>
7 years agolldpd: bump to 0.9.6
Stijn Tintel [Wed, 15 Mar 2017 07:30:26 +0000 (08:30 +0100)]
lldpd: bump to 0.9.6

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
7 years agomt76: update to the latest version, fixes Mesh/IBSS crypto support
Felix Fietkau [Fri, 17 Mar 2017 12:56:05 +0000 (13:56 +0100)]
mt76: update to the latest version, fixes Mesh/IBSS crypto support

9f02db7 mt76x2: fall back to software crypto for IBSS/Mesh per-sta GTK
4a54ab3 mt7603: fall back to software crypto for IBSS/Mesh per-sta GTK
712b8e8 mac80211: claim RSN IBSS support

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoperf: fix build with kernel 4.9
Mathias Kresin [Fri, 17 Mar 2017 07:29:42 +0000 (08:29 +0100)]
perf: fix build with kernel 4.9

The musl workaround isn't required anymore and causes build errors with
kernel 4.9 when applied.

Signed-off-by: Mathias Kresin <dev@kresin.me>
7 years agolantiq: Shortcut non-pppoa interfaces in dsl_notify
Oswald Buddenhagen [Wed, 15 Mar 2017 10:40:12 +0000 (11:40 +0100)]
lantiq: Shortcut non-pppoa interfaces in dsl_notify

No need to query 'up' and 'auto' when they are not going to be used.

Signed-off-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
7 years agoramips: add support for Zbtlink ZBT-WE2026
Vaclav Svoboda [Wed, 15 Mar 2017 08:37:05 +0000 (09:37 +0100)]
ramips: add support for Zbtlink ZBT-WE2026

This patch adds support for the Zbtlink ZBT-WE2026.

Specification:
- SoC: MediaTek MT7620N (580MHz)
- RAM: 64 MiB
- Flash: 8 MiB SPI
- LAN: 4x100M
- WAN: 1x100M

Installation through bootloader webserver:
- With the power unplugged press and hold reset button.
- Plug power and hold reset button until LED starts to blink.
- Install sysupgrade image using web interface on 192.168.1.1.

Signed-off-by: Vaclav Svoboda <svoboda@neng.cz>
7 years agolantiq: vr9 fxs support: fixed check of SMP state within vpe-mt.c
Stefan Koch [Mon, 13 Mar 2017 22:42:10 +0000 (23:42 +0100)]
lantiq: vr9 fxs support: fixed check of SMP state within vpe-mt.c

SMP state is generally affected by
- CONFIG_SMP build-time kernel configuration option and
- 'nosmp' runtime kernel commandline option

The SMP state within vpe-mt.c is determined by CONFIG_SMP option.
A runtime check is needed if VPE functionality
should be used with a kernel image that supports SMP.

This fix introduces a check for 'nosmp' command line option
if CONFIG_SMP kernel configuration option is enabled.

Note: This patch is needed to use lantiq FXS if CONFIG_MIPS_MT_SMP
(that activates CONFIG_SMP) is enabled within kernel configuration
and the 'nosmp' command line argument is given to disable SMP at runtime.

Without this patch CONFIG_MIPS_MT_SMP must be disabled before using FXS.
With this patch setting the 'nosmp' parameter is enough.

In general, concurrent usage of FXS and SMP
is incompatible and will cause kernel panics.

Signed-off-by: Stefan Koch <stefan.koch10@gmail.com>
7 years agoodhcp6c: update to git HEAD version
Hans Dedecker [Thu, 16 Mar 2017 20:38:12 +0000 (21:38 +0100)]
odhcp6c: update to git HEAD version

7e0d8b8 CMakeLists: don't enable libubox md5 implementation by default

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
7 years agokernel: do not build modules.order
Felix Fietkau [Thu, 16 Mar 2017 09:23:29 +0000 (10:23 +0100)]
kernel: do not build modules.order

It is not needed for anything on the system and skipping this saves some
build time, especially in cases where there is nothing to do.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agokernel: speed up build system by getting rid of redundant work
Felix Fietkau [Thu, 16 Mar 2017 08:33:18 +0000 (09:33 +0100)]
kernel: speed up build system by getting rid of redundant work

KERNELRELEASE contains a $(shell) call which is evaluated over and over
again.
The call to checksyscalls.sh is unnecessary for LEDE and also takes a
few seconds to complete.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agobuild: improve performance by avoiding lazy-eval for make shell calls
Felix Fietkau [Wed, 15 Mar 2017 17:12:46 +0000 (18:12 +0100)]
build: improve performance by avoiding lazy-eval for make shell calls

Avoids lots of redundant calls to mkhash on things like
package/kernel/linux

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agokernel: add missing config symbols for 4.9
Felix Fietkau [Fri, 10 Mar 2017 15:13:14 +0000 (16:13 +0100)]
kernel: add missing config symbols for 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agokernel: Fix the incorrect i_nlink count after jffs2's RENAME_EXCHANGE operations.
Jing Qiu [Tue, 14 Mar 2017 07:32:08 +0000 (15:32 +0800)]
kernel: Fix the incorrect i_nlink count after jffs2's RENAME_EXCHANGE operations.

The problem is caused by the incorrect handling of the parent inode's
i_nlink count for the dentry to be RENAME_EXCHANGED. There are 3 cases
to consider. Assume we want to RENAME_EXCHANGE struct dentry *a and
struct dentry *b, and inode_a is pointed to by dentry_a, inode_b is
pointed to by dentry_b:

1. If inode_a is a directory, but inode_b isn't, then we must decrease
   the i_nlink count of old_dir_i, and increase the i_nlink of new_dir_i.
2. If inode_a isn't a directory, but inode_b is a directory, then we
   must increase the i_nlink of old_dir_i, and decrease the i_nlink count
   of new_dir_i.
3. If the types of inode_a and inode_b are the same, we don't change the
   i_nlink for either old_dir_i or new_dir_i.

Signed-off-by: Jing Qiu <aqiu0720@gmail.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agomac80211: Update wireless-regdb to master-2017-03-07
Ryan Mounce [Tue, 7 Mar 2017 13:41:42 +0000 (00:11 +1030)]
mac80211: Update wireless-regdb to master-2017-03-07

The short log of changes since the 2016-06-10 release is below.

Jouni Malinen (1):
      wireless-regdb: Remove DFS requirement for India (IN)

Ryan Mounce (1):
      wireless-regdb: Update rules for Australia (AU) and add 60GHz rules

Seth Forshee (2):
      wireless-regdb: Update 5 GHz rules for Canada
      wireless-regdb: update regulatory.bin based on preceding changes

Signed-off-by: Ryan Mounce <ryan@mounce.com.au>
7 years agox86: image: drop duplicated copy of bzImage into vmlinuz
Rafał Miłecki [Sun, 12 Mar 2017 17:46:59 +0000 (18:46 +0100)]
x86: image: drop duplicated copy of bzImage into vmlinuz

It's already done once few lines earlier in the Image/Build/iso.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agolibnl: Fix building with uClibc
Alexey Brodkin [Mon, 13 Mar 2017 09:47:45 +0000 (12:47 +0300)]
libnl: Fix building with uClibc

uClibc doesn't implement strerror_l() and thus libnl starting from
3.2.29 couldn't be compiled with it any longer, see
https://github.com/thom311/libnl/commit/6c2d111177e91184073c44f83d4a6182aaba06d7

To work-around that problem we'll just do a check on strerror_l()
availability during configuration and if it's not there just fall back
to locale-less strerror().

Patch for libnl is alreadfy merged upstream, see
https://github.com/thom311/libnl/commit/e15966ac7f3b43df2acf869f98089762807d0568
and once the next libnl release happens this one must be removed from
Lede/OpenWrt.

Signed-off-by: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Felix Fietkau <nbd@nbd.name>
Cc: John Crispin <john@phrozen.org>
Cc: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agobrcm63xx: diag.sh: use common status_led_blink_* functions
Piotr Dymacz [Thu, 9 Mar 2017 14:07:52 +0000 (15:07 +0100)]
brcm63xx: diag.sh: use common status_led_blink_* functions

Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
7 years agoar7: diag.sh: use common status_led_* functions
Piotr Dymacz [Thu, 9 Mar 2017 14:03:18 +0000 (15:03 +0100)]
ar7: diag.sh: use common status_led_* functions

Switch to the common LED related status_led_blink_* functions in
set_state() as it's done in most of other targets.

Also, use "status_led_on" instead of "status_led_set_heartbeat"
and incorrect "status_led_set_on" in the "done" state.

Moreover, remove outdated comment from the script.

Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
7 years agosysntpd: restore support for peer-less (standalone) mode
Piotr Dymacz [Mon, 6 Mar 2017 22:20:37 +0000 (23:20 +0100)]
sysntpd: restore support for peer-less (standalone) mode

ntpd from Busybox supports peer-less (standalone) mode when it's started
with option -l and without any peer provided with option -p. In this
mode ntpd uses local time as reference and acts as stratum 1 server.

This mode can be used in isolated networks, where Internet access and/or
other NTP server/s are not available, but the device has some other way
of getting correct time, like e.g. GPS (ugps supports setting local time
by default).

Support for this mode was incorrectly disabled/removed in:
1527f96ca6e196fa17c96fdb3ae520158fa5943f

Signed-off-by: Piotr Dymacz <pepe2k@gmail.com>
Acked-by: Jo-Philipp Wich <jo@mein.io>
7 years agoar71xx: mikrotik: prepend "RouterBOARD" to machine name for NOR devices
Thibaut VARENE [Mon, 13 Mar 2017 13:51:54 +0000 (14:51 +0100)]
ar71xx: mikrotik: prepend "RouterBOARD" to machine name for NOR devices

When the device name doesn't already contain "RouterBOARD", this patch adds
this string to the machine name.

Most NOR devices already have "RouterBOARD" in their hardware-stored device name,
but not all of them.

This patch also makes the code more robust against buffer overflows.

Signed-off-by: Thibaut VARENE <hacks@slashdirt.org>
7 years agoar71xx: add support for RB LHG 5nD
Thibaut VARENE [Mon, 6 Mar 2017 22:06:13 +0000 (23:06 +0100)]
ar71xx: add support for RB LHG 5nD

This patch adds support for the MikroTik RouterBOARD LHG 5
https://routerboard.com/RBLHG-5nD

Specifications:
- SoC: Atheros AR9344 (600MHz)
- RAM: 64MB
- Storage: 16MB NOR SPI flash
- Wireless: builtin AR9344 5GHz, 2x2:2
- Ethernet: 1x100M

Initial flash via initramfs, like all other RB devices.

Signed-off-by: Thibaut VARENE <hacks@slashdirt.org>
7 years agocurl: Adjust URLs
Daniel Engberg [Wed, 15 Mar 2017 17:59:43 +0000 (18:59 +0100)]
curl: Adjust URLs

Update mirror list, add main site as last resort
Source: https://github.com/curl/curl-www/blob/master/latest.pl

Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
7 years agotoolchain: add musl libc.so to external toolchain
Hauke Mehrtens [Tue, 14 Mar 2017 18:18:57 +0000 (19:18 +0100)]
toolchain: add musl libc.so to external toolchain

musl provides a /lib/libc.so file which should be integrated into the libc
package when the external toolchain with musl is used.

Signed-off-by: Hauke Mehrtens <hauke.mehrtens@intel.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jo-Philipp Wich <jo@mein.io>
7 years agomt76: update to the latest version, improves mt7603 support
Felix Fietkau [Wed, 15 Mar 2017 19:56:06 +0000 (20:56 +0100)]
mt76: update to the latest version, improves mt7603 support

03e44dc mt76x2: remove unnecessary headroom check
cc70498 mt76x2: clarify queue selection field
b056a78 dma: fix endian issue in mt76_dma_get_buf
f020a60 mt7603: support loading the entire EEPROM from OTP
29b08d3 mt7603: fix endian issue in mt7603_mcu_set_timing
dce8aac mt7603: fix endian issue in mt7603_mac_fill_rx
f22273b mt7603: init WTBL entry before setting capabilities
da8e796 mt7603: check wtbl busy status and stop/start tx queues when clearing sta entry
e54add5 mt7603: move napi/tasklet enable/disable outside of the locked section
59ce2b4 mt7603: set tx vif own MAC index (needed for beacons)
93ce124 mt7603: enable beacons for other virtual interfaces
c91e660 mt7603: set secondary beacon time offsets

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoopkg: gracefully handle missing $PATH, fix build warnings
Jo-Philipp Wich [Wed, 15 Mar 2017 00:54:55 +0000 (01:54 +0100)]
opkg: gracefully handle missing $PATH, fix build warnings

Update to latest Git in order to import the following fixes:

cfe46c7 opkg_active_list_test: mark functions static, swap strcmp() with memcmp()
0b60d6a libopkg_test: mark functions static, remove unused opkg_test()
28b5e15 pkg_parse: fix bad message level for opkg_msg() call
d9e5814 pkg_parse: remove unused variable in pkg_parse_line()
1654bcc pkg_parse: remove const requirement from pkg_parse_line()
53e199e pkg_hash: remove unused variable in pkg_hash_load_feeds()
da8d599 pkg_depends: add missing parse_replacelist() prototype
f0c0a80 opkg_utils: add missing header
e450488 libopkg: remove unused progress parameters from opkg_download()
f4a55d1 libopkg: remove unused progress callback
7cbc466 libopkg: remove unused variable from opkg_install_package()
6fbf8a9 libbb: remove unused variables
7975c73 libbb: change gzip_read() buffer argument to void *
6e28563 libbb: fix gzip_thread() warnings
1ff2475 libopkg/opkg_cmd.c: more robust PATH handling
5dc3e33 libopkg: specify "/bin/sh" instead of "sh"

Signed-off-by: Jo-Philipp Wich <jo@mein.io>
7 years agoiwcap: fix handling kill signal during dump
Felix Fietkau [Thu, 9 Mar 2017 07:53:52 +0000 (08:53 +0100)]
iwcap: fix handling kill signal during dump

Do not run another loop iteration before checking the stop flag

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agokernel: Add intel_idle driver to x86_64 build
Michael Marley [Sat, 11 Mar 2017 22:49:26 +0000 (17:49 -0500)]
kernel: Add intel_idle driver to x86_64 build

This driver supports CPU-specific idle features on recent Intel
processors.  It does not conflict with the ACPI idle driver and
that driver will continue to be used for unsupported and non-Intel
processors.

Signed-off-by: Michael Marley <michael@michaelmarley.com>
7 years agokernel: Add coretemp driver to x86_64 build
Michael Marley [Sat, 11 Mar 2017 22:47:27 +0000 (17:47 -0500)]
kernel: Add coretemp driver to x86_64 build

The x86_64 build already has the k10temp driver for AMD processors
built in, so this patch adds the coretemp driver for the same
functionality on Intel processors.

Signed-off-by: Michael Marley <michael@michaelmarley.com>
7 years agokernel: add missing config symbols for 4.9
Felix Fietkau [Fri, 10 Mar 2017 07:11:46 +0000 (08:11 +0100)]
kernel: add missing config symbols for 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoumdns: update to the version 2017-03-14
Rafał Miłecki [Tue, 14 Mar 2017 10:59:53 +0000 (11:59 +0100)]
umdns: update to the version 2017-03-14

This includes 3 cleanups:
fd5a160 Don't cache hosts as services
80dd246 Refresh DNS records A and AAAA directly
6515101 Access cached records (instead of services) to read list of hosts

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
7 years agoipq806x: fix ethernet DMA parameters with linux 4.9
Felix Fietkau [Mon, 13 Mar 2017 20:42:40 +0000 (21:42 +0100)]
ipq806x: fix ethernet DMA parameters with linux 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: fix pcie with linux 4.9
Felix Fietkau [Mon, 13 Mar 2017 20:10:13 +0000 (21:10 +0100)]
ipq806x: fix pcie with linux 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agocurl: update to version 7.53.1
Hauke Mehrtens [Mon, 13 Mar 2017 19:07:14 +0000 (20:07 +0100)]
curl: update to version 7.53.1

This fixes the following security problem:
* CVE-2017-2629 SSL_VERIFYSTATUS ignored

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
7 years agombedtls: update to version 2.4.2
Hauke Mehrtens [Mon, 13 Mar 2017 19:04:32 +0000 (20:04 +0100)]
mbedtls: update to version 2.4.2

This fixes the following security problems:
* CVE-2017-2784: Freeing of memory allocated on stack when validating a public key with a secp224k1 curve
* SLOTH vulnerability
* Denial of Service through Certificate Revocation List

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
7 years agokernel: tpm: add Infineon i2c support
Ian Pozella [Mon, 6 Mar 2017 14:12:36 +0000 (14:12 +0000)]
kernel: tpm: add Infineon i2c support

(based on openwrt "kernel: add tpm support")

Signed-off-by: Ian Pozella <Ian.Pozella@imgtec.com>
7 years agolantiq: vr9 fxs support: fixed check of SMP state within vpe-mt.c
Stefan Koch [Tue, 28 Feb 2017 18:09:00 +0000 (19:09 +0100)]
lantiq: vr9 fxs support: fixed check of SMP state within vpe-mt.c

SMP state is generally affected by
- CONFIG_SMP build-time kernel configuration option and
- 'nosmp' runtime kernel commandline option

The SMP state within vpe-mt.c is determined by CONFIG_SMP option.
A runtime check is needed if VPE functionality
should be used with a kernel image that supports SMP.

This fix introduces a check for 'nosmp' command line option
if CONFIG_SMP kernel configuration option is enabled.

Note: This patch is needed to use lantiq FXS if CONFIG_MIPS_MT_SMP
(that activates CONFIG_SMP) is enabled within kernel configuration
and the 'nosmp' command line argument is given to disable SMP at runtime.

Without this patch CONFIG_MIPS_MT_SMP must be disabled before using FXS.
With this patch setting the 'nosmp' parameter is enough.

In general, concurrent usage of FXS and SMP
is incompatible and will cause kernel panics.

Signed-off-by: Stefan Koch <stefan.koch10@gmail.com>
7 years agoipq806x: re-enable the RAM clock for gcc on 4.9
Felix Fietkau [Mon, 13 Mar 2017 12:25:19 +0000 (13:25 +0100)]
ipq806x: re-enable the RAM clock for gcc on 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: fix crash on 4.9 if the rpm clock probe happens before cpuidle init
Felix Fietkau [Mon, 13 Mar 2017 12:25:00 +0000 (13:25 +0100)]
ipq806x: fix crash on 4.9 if the rpm clock probe happens before cpuidle init

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: fix NAND support for linux 4.9
Felix Fietkau [Mon, 13 Mar 2017 11:33:07 +0000 (12:33 +0100)]
ipq806x: fix NAND support for linux 4.9

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: do not allocate coherent memory in dma engine hotpath
Felix Fietkau [Mon, 13 Mar 2017 11:31:19 +0000 (12:31 +0100)]
ipq806x: do not allocate coherent memory in dma engine hotpath

The available amount of coherent DMA memory is very limited. On Linux
4.4 this issue was worked around by increasing the pool size.

It turns out that using coherent memory here is completely unnecessary.
This change reworks the driver code to use kzalloc+dma_map_single
instead.

Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: clean up patches, port missing patches from 4.4
John Crispin [Mon, 13 Mar 2017 12:30:10 +0000 (13:30 +0100)]
ipq806x: clean up patches, port missing patches from 4.4

Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
7 years agoipq806x: sync 4.9 kernel config with 4.4
John Crispin [Mon, 13 Mar 2017 09:04:37 +0000 (10:04 +0100)]
ipq806x: sync 4.9 kernel config with 4.4

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: John Crispin <john@phrozen.org>
7 years agoipq806x: clean up dts patching in 4.9
Felix Fietkau [Mon, 13 Mar 2017 12:32:02 +0000 (13:32 +0100)]
ipq806x: clean up dts patching in 4.9

Do not patch upstream files, overwrite them entirely. The upstream files
are buggy for a number of devices and this significantly simplifies the
patch structure

Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: John Crispin <john@phrozen.org>