ramips: add LinkIt Smart7688 support

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47203

SVN-Revision: 47227

ramips: add mt7688 subtarget

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47203

SVN-Revision: 47226

CC: polarssl: update to version 1.3.14
This fixes CVE-2015-5291 and some other smaller security issues.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 47202

opkg: backport 'opkg list --size' option

Backport from trunk to CC15.05 the --size option to optionally show also
the *.ipk size in the opkg package listing.

* Default behaviour is to list the available packages as earlier:
  "name - version - description"
* with "--size" the output of is "name - version - size - description".

Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
SVN-Revision: 47196

generic: add 8139cp fixes from 4.3

This contains only the fixes from the 4.3-rc4 kernel.

Additional improvements are going into 4.4 which will fix and enable
hardware checksum/TSO offload, but backporting those to older kernels
is non-trivial.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
SVN-Revision: 47195

mcs814x: use firmware partition splitter on dLAN USB Extender

to avoid editing the dts every time the kernel size changes.
uImage is now bigger than 1MB. Pad uImage to 64k erase block size.

Signed-off-by: Günther Kelleter <guenther.kelleter@devolo.de>
Backport of r46648

SVN-Revision: 47194

mcs814x: fix interrupt handling

Switch to generich chip irqs/irq domains.
Interrupts were broken since kernel 3.14. dLAN USB extender is now
booting again.

Signed-off-by: Günther Kelleter <guenther.kelleter@devolo.de>
Backport of r46647

SVN-Revision: 47193

mcs814x: fix debug UART

Switch to new 8250 debug uart code because the old
mach-mcs814x/include/mach/debug-macro.S tries to include
asm/hardware/debug-8250.S which no longer exists since kernel 3.14

Signed-off-by: Günther Kelleter <guenther.kelleter@devolo.de>
Backport of r46646

SVN-Revision: 47192

ar71xx: add support for D-LINK DIR-615 rev. Ix board

backport of r46248.

Signed-off-by: Jaehoon You <teslamint@gmail.com>
rebased on r46767.

SVN-Revision: 47191

lantiq: several ARV7519RW22 fixes

Fixes ARV7519RW22 flash register (only 1 flash).
Power LED is green, not red.
Rename message LEDs to internet.
Add LEDs configuration (lan, wan, power).
Enable switch VLANs.
Fix secondary USB.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: José Vázquez Fernández <ppvazquezfer@gmail.com>
SVN-Revision: 47190

mt76: update to the latest version, fixes tx related crashes

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47187

SVN-Revision: 47188

mvebu: add support for the Linksys WRT1900ACS (Shelby)

Backport of r47164.

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 47165

generic/3.18: refresh patches

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 47163

CC: uhttpd: fix keep-alive bug (#20607, #20661)
The two commits

  5162e3b0ee7bd1d0fd6e75e1ca7993a1834b5291
"allow request handlers to disable chunked reponses"

and

  618493e378e2239f0d30902e47adfa134e649fdc
"file: disable chunked encoding for file responses"

broke the chunked transfer encoding handling for proc responses in keep-alive
connections that followed a file response with http status 204 or 304.

The effect of this bug is that cgi responses following a 204 or 304 one where
sent neither in chunked encoding nor with a content-length header, causing
browsers to stall until the keep alive timeout was reached.

Fix the logic flaw by inverting the chunk prevention flag in the client state
and by testing the chunked encoding preconditions every time instead of
once upon client (re-)initialization.

Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
Backport of r47161

SVN-Revision: 47162

hostapd: wait longer for inactive client probe (empty data frame)

One second is not enough for some devices to ackowledge null data frame
which is sent at the end of ap_max_inactivity interval. In particular,
this causes severe Wi-Fi instability with Apple iPhone which may take
up to 3 seconds to respond.

Signed-off-by: Dmitry Ivanov <dima@ubnt.com>
Backport of r47149

SVN-Revision: 47150

mt76: sync with trunk r47143, add backports for compat with the older mac80211 version

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 47145

mac80211: add a fix to initialize the tid field in the tx queue struct

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47142

SVN-Revision: 47144

build: include homebrew include/library directory in cflags/ldflags on mac os x

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47139

SVN-Revision: 47141

build: use host cflags/ldflags for openssl prereq check

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47138

SVN-Revision: 47140

ramips: various mt7688 pinmux fixes

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46474

SVN-Revision: 47101

ralink: add irq to mt7628 gpio node

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47076

SVN-Revision: 47100

ralink: the mmc driver can now handle CD lines that are active low

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47074

SVN-Revision: 47099

ramips: add CD polling to sd driver

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46475

SVN-Revision: 47098

ralink: speed selection was broken in spi-mt7621

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47073

SVN-Revision: 47097

ramips: remove debug code from usb phy driver

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46659

SVN-Revision: 47096

ramips: add mt7621/3/8 support to the I2C driver

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46959

SVN-Revision: 47095

ramips: fix reported vendor name

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46958

SVN-Revision: 47094

ramips: add proper mt7688 detection

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46957

SVN-Revision: 47093

ramips: add speed and mode settings to spi-mt7621

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46956

SVN-Revision: 47092

ramips: add get_direction() callback and irq support to gpio-mt7621

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46955

SVN-Revision: 47091

ramips: more m7628 pinmux fixes

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46954

SVN-Revision: 47090

ramips: add mt7628 pwm driver

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46953

SVN-Revision: 47089

ramips: make pinctrl work on newer socs

newer socs have 2 mux registers

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46952

SVN-Revision: 47088

ramips: make the mt7628 spi driver work for both cs lines

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46951

SVN-Revision: 47087

ramips: make the early_printk code detect which uart is used

only tested on mt7628

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46950

SVN-Revision: 47086

rpcd: update to latest git HEAD

Provides ubus calls for sysupgrade and password changing.

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r47075

SVN-Revision: 47085

uci: update to the latest version, adds various fixes

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46738

SVN-Revision: 47084

fstools: update to latest git revision

jffs2reset was following symlinks making it delete mounts for example.

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46960

SVN-Revision: 47083

mountd: move code to a git repo

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46935

SVN-Revision: 47082

CC: ramips: Added WIZnet WizFi630A Platfrom based on Ralink RT5350
This patch is to add the WIZnet WizFi630A board as a new platform. The board is in mini pci express form factor.

Signed-off-by: Tobias Welz <tw@wiznet.eu>
This is a backport of patch r46921

SVN-Revision: 47072

bzip2: add host build

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46470

SVN-Revision: 47071

hostapd: check for banned client on association event

When using FullMAC drivers (e.g. brcmfmac) we don't get mgmt frames so
check for banned client in probe request handler won't ever be used.
Since cfg80211 provides us info about STA associating let's put a check
there.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Backport of r47064

SVN-Revision: 47065

igmpproxy: fix spurious restarts on interface events, pass used netdevs to procd instead

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47055

SVN-Revision: 47060

netifd: update to the latest version, fixes spurious client isolation in unbridged AP configurations (#20574)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47053

SVN-Revision: 47054

bison: replace the yacc script with one that does not hardcode the path (fixes #20605)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47050

SVN-Revision: 47051

kernel: backport some SPI layer improvements from 4.1 to 3.18, significantly improves flash speed on ramips

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47045

SVN-Revision: 47046

ath9k: fix handling of tx headroom for padding

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47042

SVN-Revision: 47044

base-files: sysupgrade: quote source image name

get_image() eval a filename without quoting it, resulting in errors with
filenames containing characters that need to be escaped.
    $ sysupgrade -T -f ./cfg\(12\).tar.gz img.bin; echo $?
    Image check 'platform_check_image' failed.
    Invalid config file. Please use only .tar.gz files
    1
    $ mv cfg\(12\).tar.gz cfg_12.tar.gz
    $ mv img\(1\).bin img.bin
    sysupgrade -T -f ./cfg_12.tar.gz img.bin; echo $?
    0

Enclose the content of $from in double quotes.

Signed-off-by: Adrien Schildknecht <adrien+dev@schischi.me>
Backport of r46919

SVN-Revision: 47028

kernel: add a patch to fix crashes on pppoe disconnect/reconnect

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r47026

SVN-Revision: 47027

base-files: fix ramoverlay function with kernels 3.18+

Even though there are not many users left within the OpenWrt tree it seems
this function broke during the kernel 3.18 transition.

Fix it by providing a workdir as required by overlayfs.

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Backport of r47015

SVN-Revision: 47016

tools/mtd-utils: fix PKG_MIRROR_MD5SUM

The checksum doesn't match the file on mirror2.openwrt.org.

Signed-off-by: Matthias Schiffer <mschiffer@universe-factory.net>
Backport of r47010

SVN-Revision: 47011

mips: Free memory when load_module fails (#14453)

The mips reloc patch introduced new allocations which were done before
add_unformed_module but never freed them in case of an error. A new hook in
Linux 3.19 called module_arch_freeing_init can be used for freeing memory
which were allocated during this init phase.

The problem can be seen when trying to load a module (via busybox insmod)
when it was already loaded.

    free -m
    for i in `seq 1 100`; do
        /sbin/insmod /lib/modules/*/ath9k.ko >& /dev/null
    done
    free -m

This simple loop would leak ~3.2 MB.

Signed-off-by: Sven Eckelmann <sven@open-mesh.com>
Backport of r46247

SVN-Revision: 47004

kernel: bridge, multicast-to-unicast: fix echoes on STA

Currently, multicast packets from an STA are sent to any according
multicast listener directly through the bridge multicast-to-unicast
feature. Unfortunately, so far this includes the originating STA, too,
resulting in multicast packets being echo'ed back to the originating STA
if it itself is a multicast listener for that group.

This behaviour breaks IPv6 duplicate address detection: An IPv6 Neighbor
Solicitation for IPv6 Duplicate Address Detection is being echo'ed back,
resulting in the host falsely detecting an address collision, which
makes the node unable to claim an IPv6 address and use IPv6 in general.

Mac80211 unfortunately only prevents the echoes for us for multicast
frames. For the multicast frames cast to a unicast destination we'll
need to take care of excluding the originator ourselves.

Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue>
Backport of r46765

SVN-Revision: 47003

generic: Fix per interface nf_call_iptables setting

commit r30917 ("kernel: bypass all netfilter hooks if the sysctls for that
functionality have been disabled - eliminates the overhead of enabling
CONFIG_BRIDGE_NETFILTER in the kernel config") introduced an optimization
which should reduce/eliminate the overhead for traffic send over bridges on
kernels compiled with CONFIG_BRIDGE_NETFILTER=y. But this optimization
breaks the nf_call_iptables per bridge setting which is more fine grained
than the global sysctl net.bridge.bridge-nf-call-iptables setting.

A test reflecting a real world setup was created to identify if this really
eliminates the overhead and if per-bridge nf_call_iptables could be used in
some setups to increase the throughput. A Qualcomm Atheros QCA9558 based
system with one ethernet and an ath9k wifi 3x3 in HT40 mode was used.
Cables from the AP to the wifi station were used to reduce interference
problems during the tests.

The wlan interface was put in one bridge interface called br-wlan. This
bridge usually contains some more wlan interfaces. The eth0 was put in a
second bridge called br-lan. This usually contains some other privileged
wlan or mesh interfaces. Routing was added between br-lan and br-wlan.

Three kernels were tested:

 * (default) OpenWrt kernel for this device
 * (brfilter-global) OpenWrt kernel with CONFIG_BRIDGE_NETFILTER=y
 * (brfilter-local)  OpenWrt kernel with CONFIG_BRIDGE_NETFILTER=y and
    without 644-bridge_optimize_netfilter_hooks.patch

The changes to the the netfilter settings of the bridge were done via:

 * (brfilter-global) /sbin/sysctl -w net.bridge.bridge-nf-call-iptables=1
 * (brfilter-lobal) echo 1 > /sys/class/net/br-lan/bridge/nf_call_iptables
   and/or echo 1 > /sys/class/net/br-wan/bridge/nf_call_iptables

A station connected to the wlan0 (AP) interface was used to send traffic to
a PC connected via ethernet. iperf with 3 concurrent transmissions was used
to generate the traffic.

| kernel          | br-nf-* global | nf-call* iface | download | upload   |
|-----------------|----------------|----------------|----------|----------|
| default         | 0              | -              |      209 |      268 |
| brfilter-global | 0              | -              |      185 |      243 |
| brfilter-local  | 0              | -              |      187 |      243 |
| brfilter-local  | 0              | br-lan         |      157 |      226 |
| brfilter-local  | 0              | br-lan br-wlan |      139 |      161 |
| brfilter-global | 1              | -              |      136 |      162 |

Download/upload results in Mibit/s

It can be seen that the patch doesn't eliminate the overhead. It can also
be seen that the throughput of brfilter-global and brfilter-local with
disabled filtering is the roughly the same. Also the throughput for
brfilter-global and brfilter-local for enabled filtering on all bridges is
roughly the same.

But also the brfilter-local throughput is higher when only br-lan requires
the filtering. This setting would not be possible with
644-bridge_optimize_netfilter_hooks.patch applied and thus can only be
compared with brfilter-global and filtering enabled for all interfaces.

Signed-off-by: Sven Eckelmann <sven@open-mesh.com>
Backport of r46835

SVN-Revision: 47002

kernel: bridge, multicast-to-unicast: assign src after pskb_may_pull()

A call to pskb_may_pull() might reallocate skb->data. Therefore we
should only assign the src-pointer after any potential reallocations.

Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46721

SVN-Revision: 47001

Revert "kernel: disable multicast-to-unicast translation for ipv6 neighbor solicitation (#17625)"

This reverts commit a080e8e1943156168913d0353a2e99d1151102aa.

It did not fix the problem but just hid some symptom. The real issue was
that IGMP/MLD report suppression was not considered for the
multicast-to-unicast feature. A recent netifd which isolates IGMP/MLD
reports between STAs by utilizing AP-isolation and bridge-hairpinning
should have fixed this.

It is perfectly fine to apply multicast-to-unicast to IPv6 Neighbor
Solicitations, too (once that feature is configured correctly).

Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46720

SVN-Revision: 47000

netifd: update to the latest version (as of trunk r46841)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 46999

procd: update to latest git revision

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46962

SVN-Revision: 46996

kernel: solos-pci: Increase headroom on received packets

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46993

SVN-Revision: 46995

kernel: ipv6: ip6_fragment: fix headroom tests and skb leak

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46992

SVN-Revision: 46994

CC: firewall: depend on kmod-ipt-conntrack (#20542)
Our ruleset requires kernel support for conntrack state matching, therfore
depend on the require kmod. Fixes #20542.

Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
Backport of r46990

SVN-Revision: 46991

x86/geode: enable DMI and Geos support (fixes #20512)

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Backport of r46986

SVN-Revision: 46988

x86: fix Geode LX timekeeping (#20531)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46985

SVN-Revision: 46987

ppp: backport use a more reliable way to set script environment

Signed-off-by: Steven Barth <steven@midlink.org>
SVN-Revision: 46943

base-files: replace the opkg key with the one used for the final image

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 46910

ramips: fix chunked-io support for big flash chips (#20504)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46904

SVN-Revision: 46905

bcm53xx: add NAND flash chip description for Asus RT-AC87U

This isn't that important due to different NAND driver but keeps DTS and
backports consistent.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Inspired by r46892 (trunk).

SVN-Revision: 46902

ath10k: reduce dma burst size on newer firmware to fix crashes

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46900

SVN-Revision: 46901

kernel: drop some left XHCI patch for 4.0

It was accidentally added in r46094.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 46849

CC: opkg: ensure image provided feeds are available after sysupgrade
Split the opkg.conf into three files, to make it easier to support custom
feeds and configs:

* /etc/opkg.conf -> base opkg configuration
* /etc/opkg/distfeeds.conf -> default Openwrt package feeds
* /etc/opkg/customfeeds.conf -> custom package feeds

Of these three, only the base opkg.conf and the customfeeds.conf is marked
as to be kept, so that the distfeeds.conf from the image is always used.

To ease migration, a script is added that moves any feeds from /etc/opkg.conf
to /etc/opkg/customfeeds.conf on first boot.

Also ensure that any keys used for verification are also kept in upgrade.

Backport of r46491.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46848

CC: kernel: update 3.18 to 3.18.21
Changelog:
 * https://www.kernel.org/pub/linux/kernel/v3.x/ChangeLog-3.18.21

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46847

ipq806x: delete from CC branch; it is broken and unsupported

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
SVN-Revision: 46845

kernel: restore 640-bridge_no_eap_forward.patch to its original form

It was corrupted in r38528. The most obvious symptom is repeated messages like this:

Tue Sep  8 08:25:18 2015 kern.warn kernel: [77141.972226] br-lan: received packet on wlan0 with own address as source address

Signed-off-by: Dmitry Ivanov <dima@ubnt.com>
Backport of r46821

SVN-Revision: 46822

brcm47xx: apply serial flash size trick to Netgear WNR1000 V3

It also uses different block size just like WGR614 V10.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Backport of r46810

SVN-Revision: 46812

brcm47xx: add Netgear WNR1000 V3 support in the Linux arch code

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Backport of r46805

SVN-Revision: 46811

uhttpd: update to the latest version, fixes deferred cgi script processing (#20458)

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Backport of r46807

SVN-Revision: 46808

ramips: fix devicetree corruption with some boot loaders if the caches are not ready at boot

Signed-off-by: John Crispin <blogic@openwrt.org>
Backport of r46766

SVN-Revision: 46767

CC: hostapd: properly enable 802.11w support
backport of r46737

Add CONFIG_IEEE80211W variable to DRIVER_MAKEOPTS so that 802.11w
support is properly compiled in full variant.

This fixes #20179

Signed-off-by: Janusz Dziemidowicz <rraptorr@nails.eu.org>
SVN-Revision: 46751

CC: broadcom-wl: add interface after bridge is up
backport of r46749

Without this patch adding the wifi device to the bridge may fail
because the bridge is not already configured when the wifi device gets
configured. This patch makes broadcom-wl wait till the bridge is ready.

This fixes #17262

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46750

CC: broadcom-wl: fix bash comparison error
Backport of r46746

This fixes #17495

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46748

CC: broadcom-wl: fix section mismatches in broadcom-wl driver
Backport of r46745

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46747

brcm47xx: fix Linksys E1200 V2 image filename

Drop unwanted suffix (copy & paste mistake).

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Backport of r46742

SVN-Revision: 46743

mac80211: add pending brcmfmac patch fixing warm reboots

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Backport of r46739

SVN-Revision: 46740

mac80211: update brcmfmac to the latest version

This also adds one patchset of pending changes that fix support for
multiple interfaces. It was somehow broken.
Driver couldn't correctly match firmware and system interfaces resulting
in not working APs and WARNINGs.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 46736

mac80211: group last brcmfmac changes into (half) patchset

This prepares some place for backporting extra patches.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 46735

bcm53xx: add support for Netgear R7000

Backport of r46363 r46730 r46731 r46732.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
SVN-Revision: 46733

CC: brcm63xx: fix WAP-5813n default network config
/etc/uci-defaults/02_network had a typo, making it generate the wrong
network config.

Closes #20407.

Backport of r46727.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46728

CC: mvebu: kirkwood: fix ehci-orion probe if generic-phy isn't enabled
Properly treat -ENOSYS as no PHY, else ehci-orion won't work without
generic phy support.

Backport of r46711.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46712

CC: mac80211: disable ipw2x00 for big endian targets
The ipw2x00 drivers assume that the system they are running is little
endian, and access everything in native byte order. When run on a big
endian system, everything breaks apart.

Since fixing this is non trivial on a first glance, disable them for
big endian targets.

Backport of r46708.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46710

CC: brcm63xx: ensure dummy byte is set for mapped spi flash with fast read
Some CFEs seem to misconfigure the mapped memory flash access with
fast read but without a dummy byte, causing all accesses to be prefixed
with 0xff.
This of course breaks reading out the nvram, so do not just move back to
single i/o accessors, but also ensure that the dummy byte is correctly
set.

Backport of r46707.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
SVN-Revision: 46709

mvebu: fix cobra chainmask (backport of r46697)

Thanks Sebastian for spotting it.

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46698

mwlwifi: downgrade to 10.3.0.3-20150804

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46695

mwlwifi: fix commit hash for 10.3.0.8

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46692

mvebu: add support for the Linksys boot counter (backport of r46690)

The u-boot boot counter was never reset after a successful boot,
which sometimes could make some variables become out of sync.
This patch adds support for the boot counter and enables
auto_recovery unconditionally.

Signed-off-by: Jonas Gorski <jogo@openwrt.org>
Signed-off-by: Rob Mosher <nyt-openwrt@countercultured.net>
Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46691

uhttpd: we don't know where the device is located, so reflect that in the cert

backport of r46688

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46689

mwlwifi: upgrade to 10.3.0.8-20150818

Signed-off-by: Imre Kaloz <kaloz@openwrt.org>
SVN-Revision: 46686

xz: update to version 5.2.1

Backport of 46603.

This adds support for symbol versioning which is needed on
CentOs/RedHat 7 to fix #20134.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46681

CC: kernel: update 3.18 to 3.18.20
Changelog:
 * https://www.kernel.org/pub/linux/kernel/v3.x/ChangeLog-3.18.20

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46680

procd: fix seccomp build on some architectures

Backport of r46612.

fix generating syscall-names.h

Sometimes the syscall number is not defined with a number but with an
offset to an other syscall and then make_syscall_h.sh created some
broken header file.

For example the bit/syscall.h from musl for i386 has this:

  #define __NR_timer_create     259
  #define __NR_timer_settime    (__NR_timer_create+1)

With this patch the resulting array looks like this:

 [259] = "timer_create",
 [(__NR_timer_create+1)] = "timer_settime",

This closes #20195.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
SVN-Revision: 46679

CC: uhttpd: pass X-HTTP-Method-Override header to cgi scripts
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
Backport of r46677

SVN-Revision: 46678