6f74e269c8 kernel: bump 4.14 to 4.14.87
b2f243394e kernel: bump 4.9 to 4.9.144
2213b20aee kernel: bump 4.14 to 4.14.86
bcd7644007 kernel: bump 4.9 to 4.9.143
d5afaa4114 openvpn: re-add option comp_lzo
629073e86d rpcd: update to latest Git head
1cd945ea22 ramips: fix leds on GL.iNet GL-MT300N-V2
fbadfecbb5 ipq40xx: fix openmesh sysupgrade with tar content out of order
06a20afb34 rules.mk: fix syntax error
d40e90986c rules.mk: add INSTALL_SUID macro
d40de11d1b base-files: fix prerm return value, align with postinst code
cc8e875039 sdk: find kernel modules when KDIR is a symlink
a8b292afe6 uhttpd: update to latest Git head
7a8b75375c uhttpd: support multiple Lua prefixes
fede6df09e uhttpd: update to latest Git head
9671a2d2c8 apm821xx: MBL: load kernel/dtb from SATA 0:1 first
26ebcc88e3 apm821xx: wndr4700: restore sd-card media detection
5337319bdf uclient: update to latest Git head
Compile-tested: ar71xx, ipq40xx, x86-64
Runtime-tested: ar71xx
ath10k-ct uses a rather high number of buffers to communicate with the
QCA/Candelatech firmware. Especially the HTC (host-target-communication)
and HTT (host-target-transport) can take up a lot of memory when data is
transferred over a ath10k wifi link.
Even a 256MB device with three radios can go OOM (while sending to three
devices using UDP unicast/multicast packets) with the default buffer
limits. Unfortunately, this also reduced the maximum 5GHz throughput on an
IPQ40xx device from 494/432 Mbit/s for TCP transfers (download/upload) to
438/343 Mbit/s. These throughput values should still be good enough for a
Freifunk installation. And these changes are already used by ath10k in
OpenWrt 18.06.
bc7abe8729f3 include/kernel-build.mk: fix kernel rebuild on backport patch changes
c99c70e05303 kernel: backport page fragment API changes from 4.10+ to 4.9
ath10k has to switch the regulatory domain when switching the default
country when ATH_USER_REGD is enabled. This is for example important when
switching from country US (FCC) to DE (ETSI). The ath implementation will
use ath_reg_dyn_country_user_allow to check if this is allowed.
Unfortunately, this function always seems to return false to
ath_reg_notifier_apply even when ATH_USER_REGD is enabled. But it must
actually always accept the requests from the user (when ATH_USER_REGD is
enabled) to correctly set the conformance test limits in the ath10k binary
blob.
Not doing it will sometimes allow too high transmit powers on edge channels
for correctly calibrated devices outside their default regulatory domain.
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
The assignment of the gpio_count and reg_base in ath79_gpio_output_select
takes care that only supported platforms can use ath79_gpio_output_select.
But the rebase of the patches after the last OpenWrt base update
accidentally added an extra check for the SoC which was missing support
soc_is_qca955x().
Boots on QCA955X devices which were using ath79_gpio_output_select stopped
with
Kernel bug detected[#1]:
[...]
Call Trace:
[<803e6174>] ath79_gpio_output_select+0x30/0xe8
[<803ed590>] om5p_acv2_setup+0x24/0x200
[<803e4dc0>] ath79_setup+0x28/0x38
[<80069ac8>] do_one_initcall+0x148/0x1ec
[<803e2cc4>] kernel_init_freeable+0x150/0x208
[<80065218>] kernel_init+0x10/0x114
[<80060878>] ret_from_kernel_thread+0x14/0x1c
Fixes: a22c1d5670 ("Update OpenWrt base")
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Some (possibly broken) bootloaders incorrectly initialize the at8033 phy.
This breaks auto-negotation on these SGMII devices. The mode has therefore
to be set back in to a valid configuration by Linux.
Fixes#911
Using HTTPS breaks the download on Debian Wheezy. The switch was an
unintended side effect of the backport from LEDE.
Also, fix the commit message of the backport patch (mwlwifi is not updated
anymore since the last OpenWrt base upgrade).
Fixes#919
Generate flashable images for the Archer C7 v2 with current stock firmware
again.
To set the region code, the GLUON_REGION variable must be set to "us" or
"eu" in site.mk or as a make argument.
Fixes#860
This patchset enables the RX LNA for the CPE210/510, improving RX by about
20dB. The profiles for CPE210 and CPE510 is split into two images.
The problematic patch switching the CPE510 to the secondary ART is left
out.
The new MR1750v2 device support is only available in LEDE master. The
relevant patches have to backported to add support for them in Gluon
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
The new OM2P-HSv3 device support is only available in LEDE master. The
relevant patches have to backported to add support for them in Gluon
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
This patchset enables the RX LNA for the CPE210/510, improving RX by about
20dB. The profiles for CPE210 and CPE510 is split into two images, so the
CPE510 can use the correct ART offset, improving the TX power by 10dB.
Fixes#796
Most doubles that are delivered via respondd have limited input
precision, but are converted with up to 17 digits of precision. That can
cause ugly blowups like 0.2800000000000001 in the output, which is
avoided by specifying better format strings (like "%.2f" in most cases).
The new ath9k/ath10k based devices are only available in OpenWrt trunk. The
relevant patches have to backported to add support for them in Gluon
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
The workaround to generate sysupgrade images for OpenMesh devices in gluon
is replaced in LEDE/OpenWrt by a special patch. It is therefore better to
drop the workaround and use the upstream version.
Reported-by: Matthias Schiffer <mschiffer@universe-factory.net>
Signed-off-by: Sven Eckelmann <sven.eckelmann@open-mesh.com>
Original commit message:
MIPS: ath79: make bootconsole wait for both THRE and TEMT
This makes the ath79 bootconsole behave the same way as the generic 8250
bootconsole.
Also waiting for TEMT (transmit buffer is empty) instead of just THRE
(transmit buffer is not full) ensures that all characters have been
transmitted before the real serial driver starts reconfiguring the serial
controller (which would sometimes result in garbage being transmitted.)
This change does not cause a visible performance loss.
In addition, this seems to fix a hang observed in certain configurations on
many AR7xxx/AR9xxx SoCs during autoconfig of the real serial driver.
A more complete follow-up patch will disable 8250 autoconfig for ath79
altogether (the serial controller is detected as a 16550A, which is not
fully compatible with the ath79 serial, and the autoconfig may lead to
undefined behavior on ath79.)
