openwrt/staging/mkresin.git
3 years agoipq806x: ubnt,unifi-ac-hd: reorder eth0 and eth1
Mark Mentovai [Mon, 12 Apr 2021 15:59:20 +0000 (11:59 -0400)]
ipq806x: ubnt,unifi-ac-hd: reorder eth0 and eth1

The Ubiquiti UniFi AC HD (UAP-AC-HD, UAP301) has two Ethernet ports,
labeled MAIN and SECONDARY, connected to gmac2 and gmac1, respectively.
The standard probe order results in gmac1/SECONDARY being eth0 and
gmac2/MAIN being eth1. This does not match the stock firmware, is
contrary to user expectation, causes the wrong (high) MAC address to be
used in a bridged configuration (the default for this device), and makes
the gmac2/MAIN port unusable in the preinit environment (such as for
failsafe). Until a recent patch, gmac1/SECONDARY (eth0) was not even
usable.

This reorders the ports so that gmac2/MAIN is eth0, and the now-working
gmac1/SECONDARY is eth1. eth0 has the low MAC address and eth1 has the
high; when bridged, the bridge takes on the correct low MAC address.
This matches the stock firmware. The MAIN port is usable for failsafe
during preinit.

This device does not have a switch on board, so there's no possibility
to remap ports via switch configuration. "ip link set $interface name"
is used instead, during preinit before networking is configured.

Signed-off-by: Mark Mentovai <mark@moxienet.com>
Build-tested: ipq806x/ubnt,unifi-ac-hd
Run-tested: ipq806x/ubnt,unifi-ac-hd

3 years agoipq806x: ubnt,unifi-ac-hd: use on-board PHYs
Mark Mentovai [Mon, 12 Apr 2021 15:59:19 +0000 (11:59 -0400)]
ipq806x: ubnt,unifi-ac-hd: use on-board PHYs

Unlike many ipq806x devices, Ubiquiti UniFi AC HD (UAP-AC-HD, UAP301)
has no switch on board. Its two Ethernet ports are connected to Atheros
AR8033 PHYs. It is not appropriate to use fixed-link in this
configuration. Instead, configure the correct PHYs in the device tree
configuration to allow the at803x driver to load.

Signed-off-by: Mark Mentovai <mark@moxienet.com>
Build-tested: ipq806x/ubnt,unifi-ac-hd
Run-tested: ipq806x/ubnt,unifi-ac-hd

3 years agoipq806x: dwmac: clear forced speed during probe
Mark Mentovai [Mon, 12 Apr 2021 15:59:18 +0000 (11:59 -0400)]
ipq806x: dwmac: clear forced speed during probe

On a Ubiquiti UniFi AC HD (ubnt,unifi-ac-hd, UAP-AC-HD, UAP301), a
forced speed on gmac1 is set in the QSGMII PCS_ALL_CH_CTL register,
presumably by the bootloader (4.3.28), preventing the interface from
being usable. The QSDK NSS GMAC driver takes care to clear the forced
speed in nss_gmac_qsgmii_dev_init
(https://source.codeaurora.org/quic/qsdk/oss/lklm/nss-gmac/tree/ipq806x/nss_gmac_init.c?h=nss
at d5bb14925861).

gmac1 is connected to the port on the device labeled SECONDARY, and is
currently eth0 but will be switched to eth1 by a subsequent patch. By
clearing the QSGMII PCS forced speed during dwmac initialization when
SGMII is in use, this port becomes usable.

This patch is upstreamable, and will be sent upstream after successful
testing in OpenWrt.

Signed-off-by: Mark Mentovai <mark@moxienet.com>
Build-tested: ipq806x/ubnt,unifi-ac-hd
Run-tested: ipq806x/ubnt,unifi-ac-hd

3 years agokernel: fix parsing fixed subpartitions
Rafał Miłecki [Thu, 6 May 2021 11:10:17 +0000 (13:10 +0200)]
kernel: fix parsing fixed subpartitions

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agotplink-safeloader: fix product_name of TP-Link AD7200
Alex Henrie [Tue, 4 May 2021 03:07:17 +0000 (21:07 -0600)]
tplink-safeloader: fix product_name of TP-Link AD7200

The stock firmware does not accept firmware with "Talon" in the name.

Tested on firmware version 1.0.10 Build 20160902 rel. 57400 which came
preinstalled, as well as latest firmware version 2.0.1 Build 20170103
rel.71053 flashed from
AD7200v1-up-ver2-0-1-P1[20170103-rel71053]_2017-01-04_10.08.28.bin.

Fixes: 1a775a4fd033 ("ipq806x: add support for TP-Link Talon AD7200")
Signed-off-by: Alex Henrie <alexhenrie24@gmail.com>
[added details about vendor firmware]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agoprocd: update to git HEAD
Daniel Golle [Wed, 5 May 2021 12:17:30 +0000 (13:17 +0100)]
procd: update to git HEAD

 021ece8 procd: Use /dev/console for serial console if exists

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobase-files: upgrade: take down loop and LVM before upgrade
Daniel Golle [Sat, 1 May 2021 19:21:54 +0000 (20:21 +0100)]
base-files: upgrade: take down loop and LVM before upgrade

Users of devices with large block storage may choose to have an LVM
partition on the same device which is used for booting OpenWrt.
The presents a problem during sysupgrade as the root device is then
still busy and changing partitions will not work as desired,
leading to data corruption in case the newly flashed image is larger
than the currently installed one.
Having loop devices setup causes similar havoc.
Make sure all volume groups are offline and all loop devices have been
released before sysupgrade.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agodnsmasq: Update to version 2.85
Alan Swanson [Tue, 27 Apr 2021 15:24:05 +0000 (16:24 +0100)]
dnsmasq: Update to version 2.85

Fixes issue with merged DNS requests in 2.83/2.84 not being
retried on the firsts failed request causing lookup failures.

Also fixes the following security problem in dnsmasq:
* CVE-2021-3448:
  If specifiying the source address or interface to be used
  when contacting upstream name servers such as:
  server=8.8.8.8@1.2.3.4, server=8.8.8.8@1.2.3.4#66 and
  server=8.8.8.8@eth0 then all would use the same socket
  bound to the explicitly configured port. Now only
  server=8.8.8.8@1.2.3.4#66 will use the explicitly
  configured port and the others random source ports.

Remove upstreamed patches and update remaining patch.

Signed-off-by: Alan Swanson <reiver@improbability.net>
[refreshed old runtime support patch]
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
3 years agoltq-dsl-base: Make package nonshared to fix image builder
Hauke Mehrtens [Sun, 2 May 2021 20:59:36 +0000 (22:59 +0200)]
ltq-dsl-base: Make package nonshared to fix image builder

This package depends on the lantiq target and is only build for that
target. A normal package would be build by the SDK builder probably
under a different target and then this package will not be selected.
Mark it as nonshared to build it when the lantiq target gets build.

Fixes: FS#3773, FS#3774
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agomac80211: Update to version 5.10.34-1
Hauke Mehrtens [Sun, 2 May 2021 16:45:40 +0000 (18:45 +0200)]
mac80211: Update to version 5.10.34-1

The removed patches were applied upstream and are not needed anymore.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agobusybox: backport fix for CVE-2021-28831
Hauke Mehrtens [Sun, 2 May 2021 15:42:19 +0000 (17:42 +0200)]
busybox: backport fix for CVE-2021-28831

This backports a fix for the low priority CVE-2021-28831:
  decompress_gunzip.c in BusyBox through 1.32.1 mishandles the error bit
  on the huft_build result pointer, with a resultant invalid free or
  segmentation fault, via malformed gzip data.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoath79: Deactivate ZyXEL NBG6716 by default
Hauke Mehrtens [Sun, 2 May 2021 13:00:40 +0000 (15:00 +0200)]
ath79: Deactivate ZyXEL NBG6716 by default

The kernel image is too big now and the build fails.

WARNING: Image file zyxel_nbg6716-kernel.bin is too big: 4205404 > 4194304

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoiwinfo: update to latest Git HEAD
David Bauer [Sat, 1 May 2021 22:22:56 +0000 (00:22 +0200)]
iwinfo: update to latest Git HEAD

c45f0b5 iwinfo: add 802.11ax HE rate information

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agouboot-mediatek: bpi-r64: add TFTP update options to eMMC bootmenu
Daniel Golle [Sat, 1 May 2021 21:32:54 +0000 (22:32 +0100)]
uboot-mediatek: bpi-r64: add TFTP update options to eMMC bootmenu

Just like on SPI-NAND, also allow updating the bootloader comfortably
on eMMC installations.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: bpi-r64: add eMMC bootloader artifacts
Daniel Golle [Sat, 1 May 2021 21:31:28 +0000 (22:31 +0100)]
mediatek: bpi-r64: add eMMC bootloader artifacts

In order to allow easily updating the bootloader on eMMC also provide
artifacts for that. Support for updating bootloader via TFTP will be
added to the loader CLI menu in a follow-up commit.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: add missing kernel config option
Hauke Mehrtens [Sat, 1 May 2021 19:11:22 +0000 (21:11 +0200)]
kernel: add missing kernel config option

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agomediatek: Add missing config options
Hauke Mehrtens [Sat, 1 May 2021 20:00:15 +0000 (22:00 +0200)]
mediatek: Add missing config options

It is possible to select CONFIG_MTD_PARSER_TRX on the mediatek now. Add
this option to the kernel configuration file.

Fixes: 58c5e2566482 ("mediatek: support non standard trx magic values")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agokernel: Reorder kernel configuration options
Hauke Mehrtens [Sat, 1 May 2021 19:58:26 +0000 (21:58 +0200)]
kernel: Reorder kernel configuration options

Reorder the options and remove double entries.
This was generated by running this:
$ ./scripts/kconfig.pl '+' target/linux/generic/config-5.10 /dev/null > target/linux/generic/config-5.10.new
$ mv target/linux/generic/config-5.10.new target/linux/generic/config-5.10
$ ./scripts/kconfig.pl '+' target/linux/generic/config-5.4 /dev/null > target/linux/generic/config-5.4.new
$ mv target/linux/generic/config-5.4.new target/linux/generic/config-5.4

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agokernel: partly revert: update mt7530 EEE patch from upstream
Hauke Mehrtens [Sat, 1 May 2021 19:49:23 +0000 (21:49 +0200)]
kernel: partly revert: update mt7530 EEE patch from upstream

This partly reverts commit 20a924d2ae378fe35b74f81be3b7f58366472e80.

This commit broke the build of the Mediatek target with kernel 5.4, for
kernel 5.10 it is fine.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agokernel: qlcnic: fix typo in module description
Chukun Pan [Sat, 1 May 2021 11:08:01 +0000 (19:08 +0800)]
kernel: qlcnic: fix typo in module description

Fixes: f88c64d28ccf ("kernel: netdev: add qlcnic")
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
3 years agoglibc: update to latest 2.33 HEAD (bug 27744)
Hans Dedecker [Wed, 28 Apr 2021 18:48:01 +0000 (20:48 +0200)]
glibc: update to latest 2.33 HEAD (bug 27744)

3f5080aedd nptl: Do not build nptl/tst-pthread-gdb-attach as PIE
36783141cf nptl: Check for compatible GDB in nptl/tst-pthread-gdb-attach
ea299b62e8 nptl_db: Support different libpthread/ld.so load orders (bug 27744)
162df872f0 x86: tst-cpu-features-supports.c: Update AMX check

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
3 years agomediatek: correct address of ethernet PHY
David Bauer [Tue, 20 Apr 2021 01:14:10 +0000 (03:14 +0200)]
mediatek: correct address of ethernet PHY

We still have no driver for the PHY, however we can fix it's
address.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoath79: drop cs-gpios property
David Bauer [Wed, 3 Mar 2021 09:23:00 +0000 (10:23 +0100)]
ath79: drop cs-gpios property

The spi-ath79 driver performs the chipselect by writing to dedicated
register in the SPI register block. So the GPIO numbers were not used.

Tested-on: Enterasys WS-AP3705i

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoath79: set number of chipselect lines
David Bauer [Thu, 29 Apr 2021 12:07:40 +0000 (14:07 +0200)]
ath79: set number of chipselect lines

All chipsets from AR7100 up to QCA9563 have three dedicated chipselect
lines for the integrated SPI controller. Remove the number of
chipselects from the platform data, as there is no need to manually set
this to a different value.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoath79: add SPI_MASTER_GPIO_SS flag
David Bauer [Thu, 29 Apr 2021 12:07:01 +0000 (14:07 +0200)]
ath79: add SPI_MASTER_GPIO_SS flag

Add the SPI_MASTER_GPIO_SS flag for the spi-ath79 driver. Otherwise,
the custom chipselect function is never called. This breaks hardware,
where the three dedicated chipselect lines are used instead of generic
GPIO pins.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoapm821xx: MBL: correct phy-mode delay settings
Christian Lamparter [Sat, 1 May 2021 09:36:10 +0000 (11:36 +0200)]
apm821xx: MBL: correct phy-mode delay settings

This came up in an upstream commit: "
b1dd9bf688b0 "net: phy: broadcom: Fix RGMII delays for BCM50160 and BCM50610M"

The PHY driver entry for BCM50160 and BCM50610M calls
bcm54xx_config_init() but does not call bcm54xx_config_clock_delay() in
order to configuration appropriate clock delays on the PHY, fix that."

So the "rgmii" phy-mode has always been wrong, but went unnoticed since
the broadcom phy driver didn't push the delay settings to the chip.

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
3 years agokernel: qlcnic: add dependency to kmod-hwmon-core
Hauke Mehrtens [Sat, 1 May 2021 09:58:47 +0000 (11:58 +0200)]
kernel: qlcnic: add dependency to kmod-hwmon-core

QLCNIC_HWMON was activated when hwmon was set, but the dependency was
missing. This broke the build bot builds. Fix this by explicitly
activating HWMON support and adding a dependency.

Fixes: f88c64d28ccf ("kernel: netdev: add qlcnic")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agomac80211/rtl: backport a rtl8192cu AP mode fix
Rui Salvaterra [Wed, 21 Apr 2021 11:09:17 +0000 (12:09 +0100)]
mac80211/rtl: backport a rtl8192cu AP mode fix

Running USB devices in AP mode is never a good idea. That said, fix the TIM
issue in rtl8192cu [1], allowing these devices to "work" in AP mode.

[1] https://patchwork.kernel.org/project/linux-wireless/patch/20210419065956.6085-1-pkshih@realtek.com/

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agotoolchain/binutils: add binutils 2.36.1
DENG Qingfang [Mon, 26 Apr 2021 04:13:51 +0000 (12:13 +0800)]
toolchain/binutils: add binutils 2.36.1

Add binutils version 2.36.1

Signed-off-by: DENG Qingfang <dqfext@gmail.com>
3 years agotoolchain/binutils: update to 2.35.2
DENG Qingfang [Mon, 26 Apr 2021 04:13:50 +0000 (12:13 +0800)]
toolchain/binutils: update to 2.35.2

Update binutils 2.35 to 2.35.2

Signed-off-by: DENG Qingfang <dqfext@gmail.com>
3 years agomvebu: LS421DE: improve pin configuration
Daniel González Cabanelas [Tue, 27 Apr 2021 08:58:15 +0000 (10:58 +0200)]
mvebu: LS421DE: improve pin configuration

The CLK125 output pin at the ethernet PHY is connected via capacitor to
GND and nowhere else. Disable it. Also tune the LED masks.

The MPP56 and MPP60 pins at the SoC are conected to the μPD720202 USB3.0
chip:
  - MPP56: wired to PCIe CLKREQ# (out)
  - MPP60: wired to PCIe RESET# (in)
Configure the pcie pinmux for these pins.

Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
3 years agonetfilter: remove no-op kconfig symbols
Rui Salvaterra [Wed, 28 Apr 2021 12:54:17 +0000 (13:54 +0100)]
netfilter: remove no-op kconfig symbols

These have long been obsolete. For reference, here's the Linux version where
each symbol has been dropped:

CONFIG_IP6_NF_QUEUE - 3.5
CONFIG_IP6_NF_TARGET_LOG - 3.4
CONFIG_IP_NF_MATCH_DSCP - 2.6.19
CONFIG_NF_CONNTRACK_IPV4 - 4.19
CONFIG_NF_CONNTRACK_IPV6 - 4.19
CONFIG_NF_CONNTRACK_RTCACHE - out-of-tree, superseded by flow offloading

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agofeeds: management: remove dead and out of project feed
Sven Roederer [Sat, 17 Apr 2021 14:18:43 +0000 (16:18 +0200)]
feeds: management: remove dead and out of project feed

Remove management feed which has not been active since some years.

Signed-off-by: Sven Roederer <devel-sven@geroedel.de>
3 years agokernel: netdev: add qlcnic
Vieno Hakkerinen [Tue, 20 Apr 2021 13:37:44 +0000 (15:37 +0200)]
kernel: netdev: add qlcnic

Add driver for QLogic QLE8240 and QLE8242 Converged Ethernet devices.

Signed-off-by: Vieno Hakkerinen <vieno@hakkerinen.eu>
3 years agokernel: add kmod-ipvlan support
Chen Minqiang [Wed, 21 Apr 2021 05:27:42 +0000 (13:27 +0800)]
kernel: add kmod-ipvlan support

This kmod is similar to macvlan with the difference being that the
endpoints have the same mac address.

It is useful on cloud where only one mac address allowed on port,
where macvlan not works but ipvlan would.

One use case is where multiple IPs and gateways assign on one net port

Signed-off-by: Chen Minqiang <ptpt52@gmail.com>
3 years agouboot-envtools: add support for Buffalo WZR-HP-G300NH
Mauri Sandberg [Thu, 22 Apr 2021 07:52:41 +0000 (10:52 +0300)]
uboot-envtools: add support for Buffalo WZR-HP-G300NH

This adds an entries for wzr-hp-g300nh-rb and wzr-hp-g300nh-s.

Signed-off-by: Mauri Sandberg <sandberg@mailfence.com>
3 years agokernel: Activate FORTIFY_SOURCE for MIPS kernel 5.4
Hauke Mehrtens [Sat, 3 Apr 2021 18:02:54 +0000 (20:02 +0200)]
kernel: Activate FORTIFY_SOURCE for MIPS kernel 5.4

CONFIG_FORTIFY_SOURCE=y is already set in the generic kernel
configuration, but it is not working for MIPS on kernel 5.4, support for
MIPS was only added with kernel 5.5, other architectures like aarch64
support FORTIFY_SOURCE already since some time.

This patch adds support for FORTIFY_SOURCE to MIPS with kernel 5.4,
kernel 5.10 already supports this and needs no changes.

This backports one patch from kernel 5.5 and one fix from 5.8 to make
fortify source also work on our kernel 5.4.

The changes are not compatible with the
306-mips_mem_functions_performance.patch patch which was also removed
with kernel 5.10, probably because of the same problems. I think it is
not needed anyway as the compiler should automatically optimize the
calls to memset(), memcpy() and memmove() even when not explicitly
telling the compiler to use the build in variant.

This increases the size of an uncompressed kernel by less than 1 KB.

Acked-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agokernel: update mt7530 EEE patch from upstream
DENG Qingfang [Thu, 22 Apr 2021 05:08:11 +0000 (13:08 +0800)]
kernel: update mt7530 EEE patch from upstream

The new EEE patch is accepted upstream, so backport it and replace the
current one.

Cc: René van Dorst <opensource@vdorst.com>
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
3 years agokernel: backport mtk_soc_eth fixes from v5.13
Ilya Lipnitskiy [Sat, 24 Apr 2021 00:19:22 +0000 (17:19 -0700)]
kernel: backport mtk_soc_eth fixes from v5.13

Fixes logic that leads to this error when booting mt7621 and other
devices that use the mediatek ethernet driver:
  [   23.144378] mtk_soc_eth 1e100000.ethernet: PPE table busy

The rest are mostly moved from pending-5.10 to backport-5.10 with a
couple of cleanups and improvements from upstream.

Refresh patches.

Links:
    https://git.kernel.org/netdev/net-next/c/c5d66587b890
    https://git.kernel.org/netdev/net-next/c/3f57d8c40fea
    https://git.kernel.org/netdev/net-next/c/5196c4178549
    https://git.kernel.org/netdev/net-next/c/787082ab9f7b
    https://git.kernel.org/netdev/net-next/c/c30c4a827390
    https://git.kernel.org/netdev/net-next/c/3630d519d7c3
    https://git.kernel.org/netdev/net-next/c/16ef670789b2
    https://git.kernel.org/netdev/net-next/c/59555a8d0dd3
    https://git.kernel.org/netdev/net-next/c/6b4423b258b9
    https://git.kernel.org/netdev/net-next/c/e9229ffd550b
    https://git.kernel.org/netdev/net-next/c/4e6bf609569c
    https://git.kernel.org/netdev/net-next/c/816ac3e6e67b
    https://git.kernel.org/netdev/net-next/c/16769a8923fa
    https://git.kernel.org/netdev/net-next/c/db2c7b353db3
    https://git.kernel.org/netdev/net-next/c/fa817272c37e
    https://git.kernel.org/netdev/net-next/c/3bc8e0aff23b

Fixes: f07fe36f22fc ("kernel: update flow offload patches to upstream version")
Cc: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agokernel: Move an upstreamed patch under backports
Mauri Sandberg [Fri, 16 Apr 2021 14:52:59 +0000 (17:52 +0300)]
kernel: Move an upstreamed patch under backports

This CFI patch was accepted upstream for 5.13. Move it away from under
ath79 and place under backports to be removed in due time.

Signed-off-by: Mauri Sandberg <sandberg@mailfence.com>
3 years agokernel: bump 5.4 to 5.4.115
John Audia [Wed, 28 Apr 2021 14:51:49 +0000 (10:51 -0400)]
kernel: bump 5.4 to 5.4.115

All patches automatically rebased.

Build system: x86_64
Build-tested: ipq806x/R7800
Run-tested: ipq806x/R7800

No dmesg regressions, everything functional

Signed-off-by: John Audia <graysky@archlinux.us>
3 years agokernel: bump 5.4 to 5.4.114
John Audia [Tue, 27 Apr 2021 22:18:36 +0000 (18:18 -0400)]
kernel: bump 5.4 to 5.4.114

Manually rebased*
  generic/backport-5.4/700-v5.5-net-core-allow-fast-GRO-for-skbs-with-Ethernet-heade.patch

Added new backport*
  generic/backport-5.4/050-gro-fix-napi_gro_frags-Fast-GRO-breakage-due-to-IP-a.patch

All others updated automatically.

The new backport was included based on this[1] upstream commit that will be
mainlined soon.  This change is needed because Eric Dumazet's check for
NET_IP_ALIGN (landed in 5.4.114) causes huge slowdowns on drivers which use
napi_gro_frags().

Build system: x86_64
Build-tested: ipq806x/R7800
Run-tested: ipq806x/R7800

No dmesg regressions, everything functional

*Credit to Alexander Lobakin
1. https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=7ad18ff6449cbd6beb26b53128ddf56d2685aa93

Signed-off-by: John Audia <graysky@archlinux.us>
3 years agoath79: mikrotik: swap RB922UAGS-5HPaCD eth0/1 MACs
Roger Pueyo Centelles [Thu, 29 Apr 2021 09:38:16 +0000 (11:38 +0200)]
ath79: mikrotik: swap RB922UAGS-5HPaCD eth0/1 MACs

Since support for SFP on the MikroTik RouterBOARD 922UAGS-5HPacD was
added by 4387fe00cb, the MAC addresses for eth0 (Ethernet) and eth1
(SFP) were swapped. This patch fixes the 02_network script to assign MAC
addresses correctly, so they match the label and the vendor's OS.

Tested on a RouterBOARD 922UAGS-5HPacD board.

Signed-off-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
3 years agokernel: bump 5.10 to 5.10.33
Rui Salvaterra [Wed, 28 Apr 2021 14:08:11 +0000 (15:08 +0100)]
kernel: bump 5.10 to 5.10.33

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agokernel: backport mtk_ppe busy-wait loop fix
Ilya Lipnitskiy [Sat, 17 Apr 2021 22:33:39 +0000 (15:33 -0700)]
kernel: backport mtk_ppe busy-wait loop fix

Fixes logic that leads to this error when booting mt7621 and other
devices that use the mediatek ethernet driver:
  [   23.144378] mtk_soc_eth 1e100000.ethernet: PPE table busy

Link: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=c5d66587b8900201e1530b7c18d41e87bd5812f4
Fixes: f07fe36f22fc ("kernel: update flow offload patches to upstream version")
Cc: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoipq40xx: add support for MikroTik SXTsq 5 ac
Roger Pueyo Centelles [Fri, 23 Oct 2020 14:05:28 +0000 (16:05 +0200)]
ipq40xx: add support for MikroTik SXTsq 5 ac

This commit adds support for the MikroTik SXTsq 5 ac (RBSXTsqG-5acD),
an outdoor 802.11ac wireless CPE with one 10/100/1000 Mbps Ethernet
port.

Specifications:
 - SoC: Qualcomm Atheros IPQ4018
 - RAM: 256 MB
 - Storage: 16 MB NOR
 - Wireless: IPQ4018 (SoC) 802.11a/n/ac 2x2:2, 16 dBi antennae
 - Ethernet: IPQ4018 (SoC) 1x 10/100/1000 port, 10-28 Vdc PoE in
 - 1x Ethernet LED (green)
 - 7x user-controllable LEDs
  · 1x power (blue)
  · 1x user (green)
  · 5x rssi (green)

Note:
 Serial UART is probably available on the board, but it has not been
 tested.

Flashing:
 Boot via TFTP the initramfs image. Then, upload a sysupgrade image
 via SSH and flash it normally. More info at the "Common procedures
 for MikroTik products" page https://openwrt.org/toh/mikrotik/common.

Signed-off-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
3 years agouboot-mediatek: unifi6lr: mtd erase before write to flash
Chen Minqiang [Tue, 27 Apr 2021 03:54:34 +0000 (11:54 +0800)]
uboot-mediatek: unifi6lr: mtd erase before write to flash

Erase firmware ereas before writing to recovery or production partition
when updating them via the bootloader menu.

Signed-off-by: Chen Minqiang <ptpt52@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoarm-trusted-firmware-mediatek: fix boot hang on unifi-6-lr
Daniel Golle [Wed, 28 Apr 2021 19:46:46 +0000 (20:46 +0100)]
arm-trusted-firmware-mediatek: fix boot hang on unifi-6-lr

The Ubiquiti Networks UniFi 6 LR access point comes with a total of
512 MB RAM provided by 2x 8-bit DDR3 SDRAM. This combination lead to
problems with the DDR calibration on boot resulting in occasional hang
on boot. Use updated calibration binary provided by MediaTek to make
boot on that device more reliable.
The binary has also been tested on the BananaPi BPi-R64 board and that
also works just fine with the new binary.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomac80211: minstrel_ht: fix issue in calculating success probability
Felix Fietkau [Wed, 28 Apr 2021 19:06:36 +0000 (21:06 +0200)]
mac80211: minstrel_ht: fix issue in calculating success probability

Missing braces in a macro were leading to badly working rates sometimes
getting a success probabilty of 1.0

Signed-off-by: Felix Fietkau <nbd@nbd.name>
3 years agoiwinfo: update to latest Git HEAD
David Bauer [Mon, 12 Apr 2021 22:13:33 +0000 (00:13 +0200)]
iwinfo: update to latest Git HEAD

50b64a6 iwinfo: add basic IEEE 802.11ax support
70d2136 iwinfo: nl80211: perform split wiphy dump
cd23727 iwinfo: cli: fix hwmode formatting

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoath79: add missing kernel config symbol
Stijn Tintel [Sat, 24 Apr 2021 12:54:24 +0000 (15:54 +0300)]
ath79: add missing kernel config symbol

Commit f724a583dcf7 updated the nand-rb4xx driver to the 5.10 testing
kernel, but forgot to add the new kernel config symbol it introduces to
the 5.10 config.

Fixes: f724a583dcf7 ("ath79: mikrotik: update nand-rb4xx driver")
Reported-by: Russell Senior <russell@personaltelco.net>
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
3 years agodnsmasq: add ignore hosts dir to dnsmasq init script
João Henriques [Sun, 18 Apr 2021 22:41:30 +0000 (23:41 +0100)]
dnsmasq: add ignore hosts dir to dnsmasq init script

When running multiple instances of dnsmasq, for example one being for the lan
and another for a guest network, it might not be desirable to have the same dns names
configured in both networks

Signed-off-by: João Henriques <joaoh88@gmail.com>
3 years agonat46: update to git HEAD
Hans Dedecker [Sat, 24 Apr 2021 19:03:02 +0000 (21:03 +0200)]
nat46: update to git HEAD

9df230a Zero checksum handling  (#25)
7688102 skip generating a UDP4 csum if the UDP6 csum is 0 and zero_csum_pass is set
4a9e4b1 zero csum: avoid calling the checksum adjustment function if udp csum is zero
2de14c5 add a module argument to skip checksum adjust for UDP if checksum is zero

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
3 years agobpftools: update to v5.11.16, simplify make
Tony Ambardar [Sun, 11 Apr 2021 00:40:21 +0000 (17:40 -0700)]
bpftools: update to v5.11.16, simplify make

Update to the latest stable upstream version.

Drop unneeded make variables to remove redundant assignments seen during
invocation of package Makefile.

Also remove the following patch now included upstream:

  * 200-fix-install-param-order-on-macos.patch

Compile and run-tested on malta/mips32be, using bpftool directly and also
libbpf (linked with tc) to inspect and load simple eBPF programs.

Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
3 years agobpftools: fix feature override for masking clang
Tony Ambardar [Mon, 12 Apr 2021 23:46:22 +0000 (16:46 -0700)]
bpftools: fix feature override for masking clang

Rename feature variable clang-bpf-global-var following upstream changes.
This restores the HAVE_CLANG feature override and should avoid rare build
errors where a recent host clang and BTF-enabled host kernel are present.

Fixes: 23be333401f0 ("bpftools: update to 5.10.10")
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
3 years agoimage: use portable filename in 'append-image-stage'
Daniel Golle [Sat, 24 Apr 2021 11:12:09 +0000 (12:12 +0100)]
image: use portable filename in 'append-image-stage'

The 'append-image-stage' command doesn't work when setting the
EXTRA_IMAGE_NAME option of the ImageBuilder as in that case
DEVICE_IMG_PREFIX is modified and no longer matches the value it had in
buildroot. Choose a filename independent of DEVICE_IMG_PREFIX for
images staged using 'append-image-stage' to fix that.

Fixes: de4b29dab9 ("image: introduce 'append-image-stage' build command")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: bpi-r64: fix sdcard.img.gz in ImageBuilder
Daniel Golle [Fri, 23 Apr 2021 22:10:08 +0000 (23:10 +0100)]
mediatek: bpi-r64: fix sdcard.img.gz in ImageBuilder

Generating the sdcard.img.gz file requires the initramfs/recovery
image to be present. Use the newly introduced 'append-image-stage'
build command to fix the ImageBuilder for the BPi-R64.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoimage: introduce 'append-image-stage' build command
Daniel Golle [Fri, 23 Apr 2021 22:05:54 +0000 (23:05 +0100)]
image: introduce 'append-image-stage' build command

Similar to 'append-image' this new command appends an existing binary.
'append-image-stage' also makes a copy of that binary and keeps it in
$(STAGING_DIR_IMAGE). When called from within the ImageBuilder, this
copy is used instead of expecting the binary to be present.
This is useful for artifacts which include the initramfs/recovery image
which is usually not included in the ImageBuilder.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: build AHCI driver as module
Daniel Golle [Fri, 23 Apr 2021 22:12:09 +0000 (23:12 +0100)]
mediatek: mt7622: build AHCI driver as module

Build the driver for the in-SoC AHCI SATA host as module, just like for
the other subtargets. No board requires booting off SATA, so we don't
need to have it built-in. All boards with SATA support already select
kmod-ata-ahci-mtk which provides the module.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoccache: update to 4.2.1
DENG Qingfang [Thu, 22 Apr 2021 04:28:39 +0000 (12:28 +0800)]
ccache: update to 4.2.1

Update ccache to 4.2.1

Signed-off-by: DENG Qingfang <dqfext@gmail.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [refresh patch]
3 years agokernel: bump 5.10 to 5.10.32
Rui Salvaterra [Wed, 21 Apr 2021 18:37:49 +0000 (19:37 +0100)]
kernel: bump 5.10 to 5.10.32

Deleted (reverse-appliable):
mediatek/patches-5.10/360-mtd-rawnand-mtk-Fix-WAITRDY-break-condition-and-time.patch

Automatically refreshed:
ath79/patches-5.10/0036-GPIO-add-named-gpio-exports.patch
generic/backport-5.10/600-v5.12-net-extract-napi-poll-functionality-to-__napi_poll.patch
generic/backport-5.10/601-v5.12-net-implement-threaded-able-napi-poll-loop-support.patch
generic/backport-5.10/602-v5.12-net-add-sysfs-attribute-to-control-napi-threaded-mod.patch
generic/backport-5.10/603-v5.12-net-fix-race-between-napi-kthread-mode-and-busy-poll.patch
generic/pending-5.10/600-netfilter_conntrack_flush.patch
generic/pending-5.10/613-netfilter_optional_tcp_window_check.patch
generic/pending-5.10/680-NET-skip-GRO-for-foreign-MAC-addresses.patch
generic/pending-5.10/768-net-dsa-mv88e6xxx-Request-assisted-learning-on-CPU-port.patch
lantiq/patches-5.10/0030-GPIO-add-named-gpio-exports.patch
ramips/patches-5.10/800-GPIO-add-named-gpio-exports.patch

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agoiproute2: avoid unneeded compiles to speed building
Tony Ambardar [Sun, 18 Apr 2021 10:09:44 +0000 (03:09 -0700)]
iproute2: avoid unneeded compiles to speed building

Skip building Makefile targets that aren't packaged: tipc, dcb, ifstat,
rtacct, lnstat, and man. Also, only compile targets needed for the current
build variant i.e. don't compile 'tc' when building an 'ip' variant and
vice versa.

These changes reduce typical build times by over 30%:
  $ make package/iproute2/clean && time make -j8 package/iproute2/compile
  (old)
  ...
  real    2m24.985s
  user    3m12.537s
  sys     0m26.677s

  (new)
  ...
  real    1m36.945s
  user    2m8.734s
  sys     0m20.046s

Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
3 years agokernel: drop the conntrack rtcache patch
Rui Salvaterra [Thu, 8 Apr 2021 22:57:48 +0000 (23:57 +0100)]
kernel: drop the conntrack rtcache patch

It's in backports-5.4, but it wasn't ever merged. Upstream followed another
approach, with flow offloading, which has much better performance. Drop this
obsolete patch and refresh the kernel patches.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
Acked-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Felix Fietkau <nbd@nbd.name>
3 years agotreewide: switch the timer frequency to 100 Hz
Rui Salvaterra [Mon, 5 Apr 2021 18:36:37 +0000 (19:36 +0100)]
treewide: switch the timer frequency to 100 Hz

Some targets select HZ=100, others HZ=250. There's no reason to select a higher
timer frequency (and 100 Hz are available in every architecture), so change all
targets to 100 Hz.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agotreewide: remove redundant ubifs kconfig symbols
Rui Salvaterra [Mon, 5 Apr 2021 18:04:40 +0000 (19:04 +0100)]
treewide: remove redundant ubifs kconfig symbols

For the targets which enable ubifs, these symbols are already part of the
generic kconfigs. Drop them from the target kconfigs.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agokernel: enable lzo, zlib and zstd in ubifs
Rui Salvaterra [Mon, 5 Apr 2021 18:01:13 +0000 (19:01 +0100)]
kernel: enable lzo, zlib and zstd in ubifs

Make sure all compression algorithms are enabled, for the time being, in order
not to break sysupgrade. In the future, we'd like to disable all except zstd, as
it's the best all-around performer.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agokernel: default to zstd compression in ubifs
Rui Salvaterra [Mon, 5 Apr 2021 15:45:47 +0000 (16:45 +0100)]
kernel: default to zstd compression in ubifs

Compared to both lzo and zlib, zstd is the best all-around performer. Backport
a kernel patch [1] (to both 5.4 and 5.10), in order to set it as default for
ubifs, if available.

[1] https://lore.kernel.org/linux-mtd/CAFLxGvwhtRY-6kT-sN=AgjvyssHb5qdTS6WQHkhKu3YrPuTkdw@mail.gmail.com/

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agobcm53xx: fix Luxul XWR-3150 LAN ports numbering
Rafał Miłecki [Wed, 21 Apr 2021 05:34:30 +0000 (07:34 +0200)]
bcm53xx: fix Luxul XWR-3150 LAN ports numbering

Port numbers were reverted.

Fixes: c459a6bf482f ("bcm53xx: add support for Luxul FullMAC WiFi devices")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agouboot-rockchip: update to v2021.04
David Bauer [Mon, 12 Apr 2021 22:28:12 +0000 (00:28 +0200)]
uboot-rockchip: update to v2021.04

Update the uboot-rockchip to the latest upstream release.
Remove upstreamed patches.

Tested-on: FriendlyElec NanoPi R2S

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agokernel: move patch to backport after upstreaming
Ilya Lipnitskiy [Tue, 20 Apr 2021 01:25:27 +0000 (18:25 -0700)]
kernel: move patch to backport after upstreaming

The patch is now in net-next so keep it in backport-5.10 to keep things
more organized.

Link: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=6ecaf81d4ac6365f9284f9d68d74f7c209e74f98
Cc: Daniel Golle <daniel@makrotopia.org>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agokernel/generic: add missing config symbol
Stijn Tintel [Mon, 19 Apr 2021 16:34:38 +0000 (19:34 +0300)]
kernel/generic: add missing config symbol

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
3 years agogemini: Bump to kernel v5.10
Linus Walleij [Sat, 17 Apr 2021 22:37:37 +0000 (00:37 +0200)]
gemini: Bump to kernel v5.10

Only two patches against mainline remains. Switch to v5.10
which works very nicely with all Gemini devices.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
[backported don't disable option CONFIG_BPF_SYSCAL]
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
3 years agoocteon: add 5.10 as testing kernel
Stijn Tintel [Sun, 18 Apr 2021 11:53:41 +0000 (14:53 +0300)]
octeon: add 5.10 as testing kernel

Remove upstreamed parts from 100-ubnt_edgerouter2_support.patch.
Compile and runtime tested on Ubiquiti EdgeRouter Lite.

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
3 years agoramips: fix at803x patch
David Bauer [Sun, 18 Apr 2021 12:38:40 +0000 (14:38 +0200)]
ramips: fix at803x patch

This patch broke while refreshing on a kernel bump. It adds the
config_aneg function twice to the AR8035 PHY now.

Restore the old behavior to restore the patches original intent.

Fixes commit 57e32650efa4 ("kernel: bump 5.10 to 5.10.31")
Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agorockchip: use alternative CID path
David Bauer [Sat, 17 Apr 2021 18:32:57 +0000 (20:32 +0200)]
rockchip: use alternative CID path

Use an alternative path to access the CID of the SD card in MMC0, used
for the generation of MAC addresses. With Kernel 5.10, the device name
of the MMC controller changed, breaking MAC address generation.

The new path is compatible with Kernel 5.4 as well as Kernel 5.10.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agorockchip: add missing config symbols
David Bauer [Fri, 16 Apr 2021 20:45:17 +0000 (22:45 +0200)]
rockchip: add missing config symbols

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agorockchip: fix kernel 5.10 built for RockPi 4
David Bauer [Fri, 16 Apr 2021 23:10:43 +0000 (01:10 +0200)]
rockchip: fix kernel 5.10 built for RockPi 4

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agogeneric: add missing config symbols
David Bauer [Fri, 16 Apr 2021 20:44:53 +0000 (22:44 +0200)]
generic: add missing config symbols

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agokernel: select kmod-backlight as dependency for kmod-drm
David Bauer [Sat, 17 Apr 2021 18:04:48 +0000 (20:04 +0200)]
kernel: select kmod-backlight as dependency for kmod-drm

Select kmod-backlight as dependency for kmod-drm in case kmod-backlight
is compiled.

This fixes kernel 5.10 build issues on x86-geode with all kmods
enabled:

Package kmod-drm is missing dependencies for the following libraries:
backlight.ko

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agokernel: netdev: make kmod-of-mdio conflict with x86
David Bauer [Fri, 16 Apr 2021 23:27:09 +0000 (01:27 +0200)]
kernel: netdev: make kmod-of-mdio conflict with x86

CONFIG_MDIO_OF depends on CONFIG_OF which is not enabled for x86.

Fixes builds of x86 with Kernel 5.10 due to unresolved dependencies.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agox86: amd-xgbe: add missing dependencies on 5.10
Chuanhong Guo [Sun, 18 Apr 2021 10:05:08 +0000 (18:05 +0800)]
x86: amd-xgbe: add missing dependencies on 5.10

fix the following build error when building with test kernel:

Package kmod-amd-xgbe is missing dependencies for the following libraries:
mdio_devres.ko

Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
3 years agouboot-envtools: mvebu: add Buffalo LS421DE
Daniel González Cabanelas [Sat, 10 Apr 2021 16:42:42 +0000 (18:42 +0200)]
uboot-envtools: mvebu: add Buffalo LS421DE

The Buffalo Linkstation LS421DE NAS lacks an uboot env config file.

Create it via scripts.

Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
3 years agox86/64: Add support for AMD SOC Ethernet controller
Zheng Bao [Tue, 13 Apr 2021 04:46:17 +0000 (12:46 +0800)]
x86/64: Add support for AMD SOC Ethernet controller

This feature is already supported in kernel. Add definition
and configuration in Makefile of OpenWRT.
It is tested on board Dibbler with processor V1605B.
Dmesg is as below.
--------------------------------------------
[    0.317122] smpboot: CPU0: AMD Ryzen Embedded V1605B with Radeon
Vega Gfx (family: 0x17, model: 0x11, stepping: 0x0)

[    8.227270] libphy: amd-xgbe-mii: probed
[    8.228189] amd-xgbe 0000:02:00.1 eth0: net device enabled
[    8.228999] libphy: amd-xgbe-mii: probed
[    8.229921] amd-xgbe 0000:02:00.2 eth1: net device enabled
.....
[   11.760869] br-lan: port 1(eth0) entered blocking state
[   11.761365] br-lan: port 1(eth0) entered disabled state
[   11.762019] device eth0 entered promiscuous mode
[   11.785285] 8021q: adding VLAN 0 to HW filter on device eth1
[   16.881431] amd-xgbe 0000:02:00.1 eth0: Link is Up - 1Gbps/Full -
flow control off
[   16.882108] br-lan: port 1(eth0) entered blocking state
[   16.882583] br-lan: port 1(eth0) entered forwarding state
[   16.883249] IPv6: ADDRCONF(NETDEV_CHANGE): br-lan: link becomes ready
-----------------------------------------------

Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
3 years agoath79: mikrotik: update nand-rb4xx driver
Roger Pueyo Centelles [Fri, 9 Apr 2021 15:12:31 +0000 (17:12 +0200)]
ath79: mikrotik: update nand-rb4xx driver

This updates the NAND driver for MikroTik RB4XX series to work with
kernel 5.10, similarly to the ar934x-nand driver (fb64e2c3).

Support for kernel 5.10 was added to all ath79 subtargets except for the
mikrotik one by commit d6b785d, since patch 920-mikrotik-rb4xx.patch
needed to be reworked. Later, commit f8512661 enabled kernel 5.10 for
the mikrotik subtarget with the nand-rb4xx driver still pending, which
is updated and added back by this patch.

Compile-tested only.

Signed-off-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
3 years agomvebu: Fix sysupgrade for GL.iNet GL-MV1000
Sven Eckelmann [Thu, 25 Mar 2021 12:26:15 +0000 (13:26 +0100)]
mvebu: Fix sysupgrade for GL.iNet GL-MV1000

The GL.iNet GL-MV1000 is booting from eMMC and the images for it are in
theory sysupgrade compatible. But the platform upgrade scripts were not
adjusted to select the mmcblock device as upgrade target. This resulted in
a failed sysupgrade because the mtd device (NOR flash) was instead tried to
be modified by the sysupgrade script.

Fixes: 050c24f05c85 ("mvebu: add support for GL.iNet GL-MV1000")
Signed-off-by: Sven Eckelmann <sven@narfation.org>
3 years agomvebu: armada 370: dts: fix the crypto engine
Daniel González Cabanelas [Sun, 4 Apr 2021 21:06:46 +0000 (23:06 +0200)]
mvebu: armada 370: dts: fix the crypto engine

The crypto engine in Armada 370 SoCs is currently broken. It can be
checked installing the required packages for testing openssl with hw
acceleration:

  opkg install openssl-util
  opkg install kmod-cryptodev
  opkg install libopenssl-devcrypto

After configuring /etc/ssl/openssl.cnf to let openssl use the crypto
engine for digest operations, and performing some checksums..

  md5sum 10M-file.bin
  openssl md5 10M-file.bin

...we can see they don't match.

There might be an alignment or size constraint issue caused by the
idle-sram area.

Use the whole crypto sram and disable the idle-sram area to fix it. Also
disable the idle support by adding the broken-idle property to prevent
accessing the disabled idle-sram.

We don't care about disabling the idle support since it is already broken
in Armada 370 causing a huge performance loss because it disables
permanently the L2 cache. This was reported in the Openwrt forum and
elsewhere by Debian users with different board models.

Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
3 years agoramips: replace mt7621s hack with upstream patch
Ilya Lipnitskiy [Wed, 14 Apr 2021 04:27:01 +0000 (21:27 -0700)]
ramips: replace mt7621s hack with upstream patch

Refresh patches.

Tested on a dual-core MT7621 device (Ubiquiti ER-X) and a single-core
MT7621 device (Netgear R6220).

This change will make future kernel upgrades easier (avoids conflicts
with upstream).

Link: https://lore.kernel.org/lkml/20210407200738.149207-1-ilya.lipnitskiy@gmail.com/
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: Chuanhong Guo <gch981213@gmail.com>
3 years agoramips: reduce spi-max-frequency for Xiaomi MI Router 4AG
David Bentham [Sun, 4 Apr 2021 12:08:43 +0000 (13:08 +0100)]
ramips: reduce spi-max-frequency for Xiaomi MI Router 4AG

Reduce spi-max-frequency for Xiaomi MI Router 4AG model

Xiaomi MI Router 4AG MTD uses two flash chips (no specific on router versions when produced from factory) - GD25Q128C and W25Q128BV.

These flash chips are capable of high frequency, but due to poor board design or manufacture process.

We are seeing the following errors in the linux kernel bootup:

`spi-nor spi0.0: unrecognized JEDEC id bytes: cc 60 1c cc 60 1c
 spi-nor: probe of spi0.0 failed with error -2`

This causes the partitions not to be detected

`VFS: Cannot open root device "(null)" or unknown-block(0,0): error -6`

Then creates a bootloop and a bricked router.

The solution to limit this race condition is to reduce the frequency from 80 mhz to 50 mhz.

Signed-off-by: David Bentham <db260179@gmail.com>
3 years agoarchs38: set device vendor and model variables
Moritz Warning [Mon, 8 Mar 2021 15:34:25 +0000 (16:34 +0100)]
archs38: set device vendor and model variables

This fixes the profiles.json output.

Signed-off-by: Moritz Warning <moritzwarning@web.de>
3 years agoarc770: set device vendor and model variables
Moritz Warning [Mon, 8 Mar 2021 13:12:53 +0000 (14:12 +0100)]
arc770: set device vendor and model variables

This fixes the profiles.json output.

Signed-off-by: Moritz Warning <moritzwarning@web.de>
3 years agotoolchain/gdb: fix compilation with ccache
Rosen Penev [Mon, 5 Apr 2021 01:32:30 +0000 (18:32 -0700)]
toolchain/gdb: fix compilation with ccache

For some reason, one of the configure checks results in some infinite
loop and ends up spawning endless gcc processes, causing OOM. Just pass
a configure var to avoid it.

Same fix as 6e23813c1ed7494a9da3b5e59cfcafd71097c420

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agokernel/modules: move act_gact into kmod-sched-core
DENG Qingfang [Fri, 9 Apr 2021 04:25:08 +0000 (12:25 +0800)]
kernel/modules: move act_gact into kmod-sched-core

As the name suggests, act_gact has the generic actions such as dropping
and accepting packets, so move it into kmod-sched-core.

Signed-off-by: DENG Qingfang <dqfext@gmail.com>
3 years agolinux-firmware: ath10k: add support for Qualcomm Atheros QCA9377
Josef Schlehofer [Sat, 10 Apr 2021 17:07:40 +0000 (19:07 +0200)]
linux-firmware: ath10k: add support for Qualcomm Atheros QCA9377

Add firmware and board file for Qualcomm Atheros QCA9377 802.11ac Wireless
Network Adapter (rev 31) recognized as [168c:0042].

This card supports standard 1x1 802.11ac Wave2, BT5, and MU-MIMO.

Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
3 years agoltq-vdsl-app: extent dsl metrics with state_num and power_state_num
Florian Eckert [Tue, 13 Apr 2021 09:30:33 +0000 (11:30 +0200)]
ltq-vdsl-app: extent dsl metrics with state_num and power_state_num

With the old ubus dsl API, the numbers for the individual line_states and
power_states were also returned. These were not ported to the new DSL
C-API. This commit adds the missing information.

For this the internal values are mapped to numbers.

* additional JSON output for state_num:
"state_num": <map_state_number>

Since not all values are meaningful only the following values are
implemented, this can be extended if the future.

* LSTATE_MAP_NOT_INITIALIZED
* LSTATE_MAP_EXCEPTION
* LSTATE_MAP_IDLE
* LSTATE_MAP_SILENT
* LSTATE_MAP_HANDSHAKE
* LSTATE_MAP_FULL_INIT
* LSTATE_MAP_SHOWTIME_NO_SYNC
* LSTATE_MAP_SHOWTIME_TC_SYNC
* LSTATE_MAP_RESYNC

* additinal JSON output for power_level:
"power_state_num": <map_power_satte_number>,

Since there are not so many here, all are mapped.

* PSTATE_MAP_NA,
* PSTATE_MAP_L0,
* PSTATE_MAP_L1,
* PSTATE_MAP_L2,
* PSTATE_MAP_L3,

Signed-off-by: Florian Eckert <fe@dev.tdt.de>
v6:
Add state LSTATE_MAP_NOT_INITILIZED at the beginning of the list
Start the list LSTATE_MAP with -1
Reviewed-by: Andre Heider <a.heider@gmail.com>
3 years agokernel: Disable IXP4xx physmap by default
Linus Walleij [Wed, 14 Apr 2021 08:36:21 +0000 (10:36 +0200)]
kernel: Disable IXP4xx physmap by default

This makes no sense on anything but the IXP4xx platform
that we do not even support anymore. If we bring it back,
it can be selectively enabled for that platform only.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agokernel: Disable CPU_THERMAL by default
Linus Walleij [Wed, 14 Apr 2021 08:45:06 +0000 (10:45 +0200)]
kernel: Disable CPU_THERMAL by default

This config is for platforms that have:
A) CONFIG_THERMAL set (we disable it by default)
B) Has thermal zones that need to control cpufreq

This is not suitable for default config, it needs to be enabled
on a per-target basis in individual configs.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agokernel: Disable GPIO_CDEV_V1 by default
Linus Walleij [Wed, 14 Apr 2021 08:53:11 +0000 (10:53 +0200)]
kernel: Disable GPIO_CDEV_V1 by default

We do not enable GPIO_CDEV so do not enable CDEV_V1
either. As we rebuild the whole userspace and kernel
in OpenWRT it doesn't make any sense to support the
old ABI anyway: the new one should be used by default.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agokernel: Disable CONFIG_DMA_PERNUMA_CMA by default
Linus Walleij [Wed, 14 Apr 2021 23:11:09 +0000 (01:11 +0200)]
kernel: Disable CONFIG_DMA_PERNUMA_CMA by default

This CMA memory allocation option only applies to NUMA
(Non-Uniform Memory Access) systems which are seldom
the kind of systems that OpenWRT address.

It is safe to assume that any system that need this
option would turn it on locally.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agokernel: Remove dead scheduler options
Linus Walleij [Wed, 14 Apr 2021 23:48:47 +0000 (01:48 +0200)]
kernel: Remove dead scheduler options

The non-MQ CFQ, deadline and noop IO schedulers are
deleted from the Linux kernel
since commit f382fb0bcef4c37dc049e9f6963e3baf204d815c
"block: remove legacy IO schedulers".

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agokernel: bump 5.4 to 5.4.113
John Audia [Fri, 16 Apr 2021 13:46:53 +0000 (09:46 -0400)]
kernel: bump 5.4 to 5.4.113

Ran update_kernel.sh in a fresh clone without any existing toolchains.  No
manual intervention needed.

Build system: x86_64
Build-tested: ipq806x/R7800
Run-tested: ipq806x/R7800

No dmesg regressions, everything functional

Signed-off-by: John Audia <graysky@archlinux.us>