openwrt/staging/hauke.git
3 years agobmips: improve CPU frequency patch
Álvaro Fernández Rojas [Mon, 8 Mar 2021 16:45:33 +0000 (17:45 +0100)]
bmips: improve CPU frequency patch

Fixes BCM6358 address and calculations.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobcm63xx: AD1018-nor: add NAND flash
Daniel González Cabanelas [Sat, 6 Mar 2021 18:59:19 +0000 (19:59 +0100)]
bcm63xx: AD1018-nor: add NAND flash

The Sercomm AD1018 has a NAND flash. We recently added support for NANDs
in this target.

Use the internal NAND as additional storage.

Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
3 years agogitignore: add .vscode for VS Code users
Álvaro Fernández Rojas [Mon, 8 Mar 2021 09:46:06 +0000 (10:46 +0100)]
gitignore: add .vscode for VS Code users

For Visual Studio Code users, .vscode is created inside the workspace.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobmips: dts: comestic changes
Álvaro Fernández Rojas [Mon, 8 Mar 2021 09:42:25 +0000 (10:42 +0100)]
bmips: dts: comestic changes

Improve and refactor device tree files.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobmips: remove unneeded console from bootargs
Álvaro Fernández Rojas [Mon, 8 Mar 2021 09:12:42 +0000 (10:12 +0100)]
bmips: remove unneeded console from bootargs

console parameter is no longer needed after latest procd update.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobmips: automatically detect CPU frequency
Álvaro Fernández Rojas [Fri, 5 Mar 2021 16:51:39 +0000 (17:51 +0100)]
bmips: automatically detect CPU frequency

Some BCM63xx SoCs support multiple CPU frequencies depending on HW
configuration.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agoprocd: update to git HEAD
Daniel Golle [Mon, 8 Mar 2021 00:10:51 +0000 (00:10 +0000)]
procd: update to git HEAD

 945d0d7 utils: fix C style in header file
 2cfc26f inittab: detect active console from kernel if no console= specified

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobpftools: fix param order for install on macOS
Georgi Valkov [Tue, 2 Mar 2021 10:41:10 +0000 (12:41 +0200)]
bpftools: fix param order for install on macOS

Fix: bpftools 5.11.2 does not compile on macOS, because the -m option
was placed between src and dst. Corrected by moving -m 644 before src.

Signed-off-by: Georgi Valkov <gvalkov@abv.bg>
3 years agobpftools: fix libbpf pkgconfig file
Tony Ambardar [Mon, 1 Mar 2021 09:07:45 +0000 (01:07 -0800)]
bpftools: fix libbpf pkgconfig file

The pkgconfig file hardcodes a host library directory which cannot be
overridden by OpenWrt during builds. Use SED to fix this and potential
include directory problems, as is done with several other packages.

This fixes a strange issue intermittently seen building iproute2 on the
oxnas target:

iptables modules directory: /usr/lib/iptables
libc has setns: yes
SELinux support: no
libbpf support: no
libbpf version 0.3.0 is too low, please update it to at least 0.1.0
LIBBPF_FORCE=on set, but couldn't find a usable libbpf

Fixes: 2f0d672088c3 ("bpftools: add utility and library packages
supporting eBPF usage")
Reported-by: Russell Senior <russell@personaltelco.net>
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
3 years agomediatek: remove no longer needed sysupgrade hack
Daniel Golle [Sun, 7 Mar 2021 18:15:50 +0000 (18:15 +0000)]
mediatek: remove no longer needed sysupgrade hack

Keeping configuration is now handled in fstools like for other types
of flash as well.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: remove execute bit and shebang from 01_leds
Adrian Schmutzler [Sat, 6 Mar 2021 19:22:15 +0000 (20:22 +0100)]
mediatek: mt7622: remove execute bit and shebang from 01_leds

This was added recently and thus overlooked in 85b1f4d8ca2d
("treewide: remove execute bit and shebang from board.d files").

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoipq806x: work-around borked QCA SDK bootloader
Daniel Golle [Sat, 6 Mar 2021 19:04:15 +0000 (19:04 +0000)]
ipq806x: work-around borked QCA SDK bootloader

The bootloader of many ipq806x boards seems to require the config node
of the FIT image to be 'config@1' (or a secific different value).
This requirement used to be implicitely satisfied because OpenWrt used
to also call the configuration node inside a FIT image 'config@1'.
However, as recent U-Boot now prohibits the use of the '@' symbol as
part of node names, this was changed by
commit 5ec60cbe9d ("scripts: mkits.sh: replace @ with - in nodes")
Explicitely restore the default name of the configuration node to
'config@1' on ipq806x.
(ipq807x is unaffected as DEVICE_DTS_CONFIG default is set
"config@hk01" in target/linux/ipq807x/image/Makefile)

Reported-by: Chen Minqiang <ptpt52@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoipq40xx: work-around borked QCA SDK bootloader
Daniel Golle [Sat, 6 Mar 2021 18:59:16 +0000 (18:59 +0000)]
ipq40xx: work-around borked QCA SDK bootloader

The bootloader of many ipq40xx boards seems to require the config node
of the FIT image to be 'config@1' (or a secific different value).
This requirement used to be implicitely satisfied because OpenWrt used
to also call the configuration node inside a FIT image 'config@1'.
However, as recent U-Boot now prohibits the use of the '@' symbol as
part of node names, this was changed by
commit 5ec60cbe9d ("scripts: mkits.sh: replace @ with - in nodes")
Explicitely restore the default name of the configuration node to
'config@1' on ipq40xx.

Reported-by: Chen Minqiang <ptpt52@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: disable RTC on Bananapi R64 and refresh patches
Daniel Golle [Sat, 6 Mar 2021 18:38:30 +0000 (18:38 +0000)]
mediatek: disable RTC on Bananapi R64 and refresh patches

The in-SoC RTC of the Bananapi R64 is more disruptive than useful
without a battery connected. Disable it to not have Linux use the
RTC provided time 2000-01-01 00:00:00 after power-loss.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agotreewide: make AddDepends/usb-serial selective
Adrian Schmutzler [Thu, 25 Feb 2021 10:54:43 +0000 (11:54 +0100)]
treewide: make AddDepends/usb-serial selective

Make packages depending on usb-serial selective, so we do not have
to add kmod-usb-serial manually for every device.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agotreewide: remove execute bit and shebang from board.d files
Adrian Schmutzler [Fri, 26 Feb 2021 13:43:50 +0000 (14:43 +0100)]
treewide: remove execute bit and shebang from board.d files

So far, board.d files were having execute bit set and contained a
shebang. However, they are just sourced in board_detect, with an
apparantly unnecessary check for execute permission beforehand.

Replace this check by one for existance and make the board.d files
"normal" files, as would be expected in /etc anyway.

Note:

This removes an apparantly unused '#!/bin/sh /etc/rc.common' in
target/linux/bcm47xx/base-files/etc/board.d/01_network

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: 5.10: refresh configs
Ilya Lipnitskiy [Sun, 28 Feb 2021 07:17:47 +0000 (23:17 -0800)]
ramips: 5.10: refresh configs

Run-tested on Ubiquiti EdgeRouter X.

Compile tested on all other subtargets.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: 5.10: port and refresh patches, ralink drv
Ilya Lipnitskiy [Sat, 20 Feb 2021 03:11:14 +0000 (19:11 -0800)]
ramips: 5.10: port and refresh patches, ralink drv

Enable testing kernel.

Fix compile errors by using new kernel APIs.

Fix fuzz by manually editing patches to ensure the code goes in the
right place.

For 721-NET-no-auto-carrier-off-support.patch, revert upstream commit
a307593a6 to keep the OpenWrt ralink driver operational.

Add mt7621-pci-phy patch to select REGMAP_MMIO as discussed in PR #3693
and #3952.

Run automatic quilt refresh on the rest.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: 5.10: rename patches to follow guide
Ilya Lipnitskiy [Sun, 28 Feb 2021 00:00:24 +0000 (16:00 -0800)]
ramips: 5.10: rename patches to follow guide

Rename patches to follow the 3-digit classification from the OpenWrt
Developer Guide.

https://openwrt.org/docs/guide-developer/build-system/use-patches-with-buildsystem#naming_patches

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: 5.10: delete upstreamed patches
Ilya Lipnitskiy [Sat, 20 Feb 2021 03:04:38 +0000 (19:04 -0800)]
ramips: 5.10: delete upstreamed patches

0098-disable_cm.patch is not needed because upstream fixed CM handling.

The rest are straightforward removals of upstreamed patches.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: 5.10: copy patches from 5.4
Ilya Lipnitskiy [Sat, 20 Feb 2021 02:59:29 +0000 (18:59 -0800)]
ramips: 5.10: copy patches from 5.4

Strict copy, no changes made.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: copy config-5.4 to config-5.10
Ilya Lipnitskiy [Sat, 27 Feb 2021 23:12:21 +0000 (15:12 -0800)]
ramips: copy config-5.4 to config-5.10

Strict copy, no changes made.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoRevert "ramips: add support for kernel 5.10"
Adrian Schmutzler [Sat, 6 Mar 2021 10:23:38 +0000 (11:23 +0100)]
Revert "ramips: add support for kernel 5.10"

This reverts commit b4aad29a1d7ad77d67073c1c54b28c429c64ed9b.

This was accidentally folded into a single commit. Remove it and
apply it properly again.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agokernel: fix kmod-usb3 on platforms without PCI
Daniel Golle [Sat, 6 Mar 2021 04:18:10 +0000 (04:18 +0000)]
kernel: fix kmod-usb3 on platforms without PCI

Partially restore the wild-card matching for kmod-usb3 modules to fix
build on platforms without PCI which otherwise file, as seen on
buildbot:

ERROR: module '[...]/linux-5.4.102/drivers/usb/host/xhci-pci.ko' is missing.
modules/usb.mk:1675: recipe for target '[...]/kmod-usb3_5.4.102-1_mips_mips32.ipk' failed

Fixes: 7bda2e9aba ("kernel: fix kmod-usb3 dependencies")
Fixes: be23f9818a ("apm821xx: add support for kernel 5.10")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoapm821xx, mt7621, oxnas, rockchip, zynq: remove PLUGIN_HOSTCC
Ilya Lipnitskiy [Sat, 6 Mar 2021 02:28:43 +0000 (18:28 -0800)]
apm821xx, mt7621, oxnas, rockchip, zynq: remove PLUGIN_HOSTCC

This config symbol should not be defined in target configs.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agokernel: generic: filter out CONFIG_PLUGIN_HOSTCC
Ilya Lipnitskiy [Sat, 6 Mar 2021 02:28:42 +0000 (18:28 -0800)]
kernel: generic: filter out CONFIG_PLUGIN_HOSTCC

It appears to be an automatic Kconfig symbol that varies depending on
the host platform. There is no need to define it in target configs, so
filter it out.

Also sort config-filter entries alphabetically.

Cc: Adrian Schmutzler <mail@adrianschmutzler.de>
Cc: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoarm-trusted-firmware-mediatek: prune now uneeded declarations
Daniel Golle [Sat, 6 Mar 2021 03:28:52 +0000 (03:28 +0000)]
arm-trusted-firmware-mediatek: prune now uneeded declarations

Remove unneeded delcarations form package Makefile now that everything
comes from github.com/mtk-openwrt upstream.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: fix kmod-usb3 dependencies
Ilya Lipnitskiy [Sat, 6 Mar 2021 02:06:02 +0000 (18:06 -0800)]
kernel: fix kmod-usb3 dependencies

Adding CONFIG_USB_XHCI_PCI_RENESAS to KCONFIG made it so it was set
for all targets that included kmod-usb3, not just the intended
apm821xx/nand. xhci-pci has a dependency on xhci-pci-renesas if
CONFIG_USB_XHCI_PCI_RENESAS is set, breaking kmod-usb3 builds on
Linux 5.10 on all other targets.

Rework the dependencies by breaking out xhci-hcd, xhci-mtk, and
xhci-pci-renesas into new hidden kernel modules and setting kmod-usb3
dependencies properly.

Tested by building mt7621 and apm821xx/nand with kmod-usb3 on Linux 5.10

Fixes: be23f981 ("apm821xx: add support for kernel 5.10")
Cc: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoarm-trusted-firmware-mediatek: don't select DDR3_FLYBY for 1ddr
Daniel Golle [Fri, 5 Mar 2021 23:53:51 +0000 (23:53 +0000)]
arm-trusted-firmware-mediatek: don't select DDR3_FLYBY for 1ddr

DDR3_FLYBY has accidentally been set also for the 1-chip variant which
lead to broken, unbootable images. Fix that.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoramips: add support for kernel 5.10
Ilya Lipnitskiy [Sat, 27 Feb 2021 23:12:21 +0000 (15:12 -0800)]
ramips: add support for kernel 5.10

Enable testing kernel.

Delete upstreamed patches:
 0098-disable_cm.patch can be dropped, upstream fixed CM handling.

Fix compile errors by using new kernel APIs.

Fix fuzz by manually editing patches to ensure the code goes in the
right place.

For 721-NET-no-auto-carrier-off-support.patch, revert upstream commit
a307593a6 to keep the OpenWrt ralink driver operational.

Add mt7621-pci-phy patch to select REGMAP_MMIO as discussed in PR #3693
and #3952.

Rename patches to follow the 3-digit classification from the OpenWrt
Developer Guide.

Run automatic quilt refresh.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoramips: 5.4: refresh configs
Ilya Lipnitskiy [Fri, 5 Mar 2021 20:07:40 +0000 (12:07 -0800)]
ramips: 5.4: refresh configs

Automatic refresh by running make kernel_oldconfig on each target.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agoapm821xx: add support for kernel 5.10 for testing
Christian Lamparter [Sat, 20 Feb 2021 16:44:12 +0000 (17:44 +0100)]
apm821xx: add support for kernel 5.10 for testing

This patch copies over refreshed patches from 5.4.

 - dropped crypto patches (they got upstreamed)

 - dropped renesas USB 3 firmware loader (they got upstreamed)

 - NAND now needs extra device-properties for ECC settings.

Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
3 years agomediatek: mt7622: bpi-r64: allow MAC addresses to be inherited
Daniel Golle [Fri, 5 Mar 2021 19:09:10 +0000 (19:09 +0000)]
mediatek: mt7622: bpi-r64: allow MAC addresses to be inherited

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agouboot-mediatek: improve BPi-R64 support
Daniel Golle [Fri, 5 Mar 2021 19:07:54 +0000 (19:07 +0000)]
uboot-mediatek: improve BPi-R64 support

 * allow MAC address from U-Boot env to be inhertied
 * allow eMMC installation to succeed also without recovery present
   on the SD Card.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: bump 5.10 to 5.10.20
Rui Salvaterra [Thu, 4 Mar 2021 19:51:39 +0000 (19:51 +0000)]
kernel: bump 5.10 to 5.10.20

Also add a new kconfig symbol (CONFIG_KCMP) to the generic config,
disabling the SYS_kcmp syscall (it was split from
CONFIG_CHECKPOINT_RESTORE, which is disabled by default, so the
previous behaviour is kept).

Removed (upstreamed) patches:
  070-net-icmp-pass-zeroed-opts-from-icmp-v6-_ndo_send-bef.patch
  081-wireguard-device-do-not-generate-ICMP-for-non-IP-pac.patch
  082-wireguard-queueing-get-rid-of-per-peer-ring-buffers.patch
  083-wireguard-kconfig-use-arm-chacha-even-with-no-neon.patch
  830-v5.12-0002-usb-serial-option-update-interface-mapping-for-ZTE-P685M.patch

Manually rebased patches:
  313-helios4-dts-status-led-alias.patch
  104-powerpc-mpc85xx-change-P2020RDB-dts-file-for-OpenWRT.patch

Run tested:
  ath79 (TL-WDR3600)
  mvebu (Turris Omnia)

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agoarm-trusted-firmware-mediatek: update to ATF 2.4 (2021-02-25)
Daniel Golle [Fri, 5 Mar 2021 17:14:02 +0000 (17:14 +0000)]
arm-trusted-firmware-mediatek: update to ATF 2.4 (2021-02-25)

All necessary blobs are now contained in the upstream repository, no
more wild replacing of blobs needed.
This new version also contains new storage drivers for (SPI-)NAND which
already comes with support for FM35Q1GA, so that patch can be dropped
as well.
Tested on:
 * Bananapi BPi-R64
   - sdmmc-2ddr
   - emmc-2ddr
 * Linksys E8450
   - snand-1ddr

All works fine (booting Bananapi BPi-R64 from SD Card does NOT require
a signed image, so patch arm-trusted-firmware-mediatek to allow doing
that).

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobuild: use config-1 instead of config@1 as default
Robert Marko [Thu, 4 Mar 2021 11:37:22 +0000 (12:37 +0100)]
build: use config-1 instead of config@1 as default

U-boot will reject all nodes with @ since commit:
https://gitlab.denx.de/u-boot/u-boot/-/commit/79af75f7776fc20b0d7eb6afe1e27c00fdb4b9b4

This will cause the OpenWrt images to fail booting,
to rectify use the config-1 as default.

Signed-off-by: Robert Marko <robert.marko@sartura.hr>
3 years agoscripts: mkits.sh: replace @ with - in nodes
Robert Marko [Thu, 4 Mar 2021 11:37:21 +0000 (12:37 +0100)]
scripts: mkits.sh: replace @ with - in nodes

U-boot will reject the nodes with @ for the address since
commit:
https://gitlab.denx.de/u-boot/u-boot/-/commit/79af75f7776fc20b0d7eb6afe1e27c00fdb4b9b4

This in turn will cause the failure to boot with OpenWrt
generated images.

So, to rectify that simply replace @ with -.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
(cover also newly added rootfs@1 and initrd@1 nodes)
Signed-off-by: Robert Marko <robert.marko@sartura.hr>
3 years agobcm4908: prepare to support TP-Link Archer C2300 V1
Rafał Miłecki [Fri, 5 Mar 2021 13:38:41 +0000 (14:38 +0100)]
bcm4908: prepare to support TP-Link Archer C2300 V1

It's a BCM4906 based device (2 CPU cores). It has 512 MiB of RAM, 4 LAN
ports, 1 WAN port, 2 USB ports, NAND flash. WiFi unknown at this point.

Flashing is possible using CFE only, proper image will be worked on
later.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agobcm63xx-cfe: update to the latest master
Rafał Miłecki [Fri, 5 Mar 2021 13:37:01 +0000 (14:37 +0100)]
bcm63xx-cfe: update to the latest master

d035016 tp-link: rename to tplink to match DT vendor prefix

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agokernel/modules: fix CDC NCM dependencies
Rui Salvaterra [Wed, 3 Mar 2021 14:24:17 +0000 (14:24 +0000)]
kernel/modules: fix CDC NCM dependencies

CDC NCM support only depends on CDC Ethernet with Linux 5.10. Fix thusly.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agogeneric: config-5.10: add missing symbols
Ilya Lipnitskiy [Wed, 3 Mar 2021 16:58:22 +0000 (08:58 -0800)]
generic: config-5.10: add missing symbols

Add symbols missing after latest crypto.mk and WireGuard changes.

Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agokernel: backport GCC 10 usbip build fix for 5.4
Rui Salvaterra [Fri, 5 Mar 2021 12:09:31 +0000 (12:09 +0000)]
kernel: backport GCC 10 usbip build fix for 5.4

From the original commit message:

"With GCC 10, building usbip triggers error for multiple definition
of 'udev_context', in:
- libsrc/vhci_driver.c:18 and
- libsrc/usbip_host_common.c:27.

Declare as extern the definition in libsrc/usbip_host_common.c."

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agokernel: remove symbols from 5.10 no longer present on upstream
Luis Araneda [Thu, 4 Mar 2021 01:03:52 +0000 (22:03 -0300)]
kernel: remove symbols from 5.10 no longer present on upstream

REFCOUNT_FULL was removed for linux 5.5:
commit fb041bb7c0a9 (locking/refcount: Consolidate implementations of refcount_t)

COMMON_CLK_VERSATILE was removed on linux 5.8:
commit 5f55f1fb187d (clk: versatile: Fix kconfig dependency on COMMON_CLK_VERSATILE)

Signed-off-by: Luis Araneda <luaraneda@gmail.com>
3 years agobcm63xx: rename upstreamed patch
Sieng Piaw Liew [Wed, 24 Feb 2021 01:23:43 +0000 (09:23 +0800)]
bcm63xx: rename upstreamed patch

Patch to fix kernel panic was recently accepted upstream so rename patch
and add acked lines to reflect that.

Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
(add the same patch for v5.10)
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agokernel: b53: update the BCM5365 UID in the fixup
Daniel González Cabanelas [Sat, 27 Feb 2021 10:58:33 +0000 (11:58 +0100)]
kernel: b53: update the BCM5365 UID in the fixup

The BCM5365 UID was updated in the driver, but we should also update it in the
fixup.

Fixes: cbcac4fde8ba ("kernel: b53: update the BCM5365 UID")
Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
(Ammend commit description, add Fixes tag)
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobcm27xx: bcm2711: disable HW_RANDOM_BCM2835
Álvaro Fernández Rojas [Fri, 5 Mar 2021 11:13:45 +0000 (12:13 +0100)]
bcm27xx: bcm2711: disable HW_RANDOM_BCM2835

This driver is only present on BCM2708, BCM2709 and BCM2710.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agorssileds: use new application led trigger backend
Florian Eckert [Wed, 5 Feb 2020 08:20:39 +0000 (09:20 +0100)]
rssileds: use new application led trigger backend

This will use the new application led trigger backend. For now this is
the only package that uses leds trigger in user space to configure the
kernel led triggers.

The callback script only emmits a message for now, so that this LED is now
managed by the rssileds service. Until now a generic warning was emitted that
this LED trigger is not supported. But that is not true.

-> Skipping trigger 'rssileds' for led '<name>' due to missing kernel module

I think this callback should be changed in the future to restart the
service.

Signed-off-by: Florian Eckert <fe@dev.tdt.de>
3 years agobase-files: add new application led trigger backend
Florian Eckert [Wed, 5 Feb 2020 08:23:27 +0000 (09:23 +0100)]
base-files: add new application led trigger backend

For now we have only kernel LED trigger support. With this change it is now
possible to use application triggers.

If we configure a LED with a non kernel trigger, then we check on every
restart and boot of the LED service if we have this trigger as an application
in "/usr/libexec/led-trigger". If this file with the name is found, then we
execute this to init the LED.

Possible use cases are:

- Start/Stop/Restart an application led trigger service for this led
- Init a LED that is configured by a hotplug script (VPN tunnel established)

Signed-off-by: Florian Eckert <fe@dev.tdt.de>
3 years agoramips: rename mtk-hsdma to hsdma-mt7621
Ilya Lipnitskiy [Fri, 5 Mar 2021 03:53:21 +0000 (19:53 -0800)]
ramips: rename mtk-hsdma to hsdma-mt7621

Follows upstream rename:
https://lore.kernel.org/driverdev-devel/20210130034507.2115280-1-ilya.lipnitskiy@gmail.com/

Fixes ramips builds on 5.4.102

Cc: John Audia <graysky@archlinux.us>
Cc: David Bauer <mail@david-bauer.net>
Cc: Petr Štetiar <ynezz@true.cz>
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
3 years agobcm63xx-cfe: update to the latest master
Rafał Miłecki [Fri, 5 Mar 2021 09:14:28 +0000 (10:14 +0100)]
bcm63xx-cfe: update to the latest master

3fb6f1c tp-link: c2300-v1: add cferam file
79f9578 sercomm: vox-2.5: add cferam file

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agofstools: fix build with glibc
Daniel Golle [Fri, 5 Mar 2021 02:03:58 +0000 (02:03 +0000)]
fstools: fix build with glibc

stropts.h is unavailable under glibc (and unneeded when building
against glibc). Include it only if not building against glibc.

Reported-by: @DazzyWalkman
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoipq40xx: net: phy: qca807x: fix GPIO driver
Robert Marko [Thu, 4 Mar 2021 11:38:31 +0000 (12:38 +0100)]
ipq40xx: net: phy: qca807x: fix GPIO driver

While rebasing into setting bits instead of magic values,
I accidentally forgot to actually set the force bit.

Without it using the pins as GPIO-s did not actually work.

Fixes: b5c93ed ("ipq40xx: add Qualcomm QCA807x driver")
Signed-off-by: Robert Marko <robert.marko@sartura.hr>
3 years agoath79: enable AT803X PHY for kernel 5.10
David Bauer [Wed, 3 Mar 2021 22:50:41 +0000 (23:50 +0100)]
ath79: enable AT803X PHY for kernel 5.10

The config symbol for the AT803x device driver was lost when adding
kernel 5.10.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoath79: replace chipselect workaround with a proper patch
David Bauer [Wed, 3 Mar 2021 16:23:16 +0000 (17:23 +0100)]
ath79: replace chipselect workaround with a proper patch

Debugging the SPI CS issue with kernel 5.10 resulted in a better
understanding for the root cause and a proper patch with a better
explanation.

Exchange the old hack patch with a more efficient (and upstreamable)
solution.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoarm-trusted-firmware-mediatek: don't try to install absent files
Daniel Golle [Thu, 4 Mar 2021 21:56:56 +0000 (21:56 +0000)]
arm-trusted-firmware-mediatek: don't try to install absent files

Don't try to install files which no longer exist
Since {e,sd}mmc are now produced by ptgen they have been removed.

Fixes: 5a3562cd1d ("arm-trusted-firmware-mediatek: remove {e,sd}mmc headers")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel-5.4: bump to 5.4.102 and refresh patches
Jason A. Donenfeld [Thu, 4 Mar 2021 20:37:13 +0000 (13:37 -0700)]
kernel-5.4: bump to 5.4.102 and refresh patches

5.4.102 backported a lot of stuff that our WireGuard backport already
did, in addition to other patches we had, so those patches were
removed from that part of the series. In the process other patches were
refreshed or reworked to account for upstream changes.

This commit involved `update_kernel.sh -v -u 5.4`.

Cc: John Audia <graysky@archlinux.us>
Cc: David Bauer <mail@david-bauer.net>
Cc: Petr Štetiar <ynezz@true.cz>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
3 years agomediatek: mt7622: check firmware metadata
Daniel Golle [Thu, 4 Mar 2021 02:49:38 +0000 (02:49 +0000)]
mediatek: mt7622: check firmware metadata

All mt7622 devices except for the UBI-variant of the mt7622-rfb1 carry
metadata appended to the sysupgrade image.
Add it for the mt7622-rfb1-ubi as well and check it on sysupgrade to
avoid accidentally flashing firmware for the wrong device (or variant
or future DEVICE_COMPAT_VERSION).

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agofeeds.conf.default: remove freifunk feed
Perry Melange [Sun, 21 Feb 2021 13:00:56 +0000 (14:00 +0100)]
feeds.conf.default: remove freifunk feed

The freifunk feed is being removed becasue
a) it is an external project and the OpenWrt team does not have access to it.
b) upon original addition of the feed, there was only a very weak tendency for
the addition.
c) there is a general lack of interest in the freifunk repo to review and/or
merge pull requests.
d) as far as can be found, all projects which use the freifunk feed have their
own make system and self-maintained feeds list.  They do not use the
feeds.conf.default from the openwrt repo.

more information can be read at the following links:

http://lists.openwrt.org/pipermail/openwrt-devel/2021-February/033807.html
https://github.com/freifunk/openwrt-packages/issues/37

Signed-off-by: Perry Melange <isprotejesvalkata@gmail.com>
3 years agobuild: add which command to build requirements
Petr Štetiar [Wed, 3 Mar 2021 18:48:55 +0000 (19:48 +0100)]
build: add which command to build requirements

`which` utility is not shipped by default for example on recent Arch
Linux and then any steps relying on its presence fails, like for example
following Python3 prereq build check:

 $ python3 --version
 Python 3.9.1

 $ make
 /bin/sh: line 1: which: command not found
 ...
 Checking 'python3'... failed.

So make `which` utility host build requirement.

References: PR#3820 FS#3525
Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agoRevert "build: replace which with Bash command built-in"
Clemens Fruhwirth [Fri, 29 Jan 2021 18:57:37 +0000 (19:57 +0100)]
Revert "build: replace which with Bash command built-in"

This reverts commit c7aec47e5e3a3ff7b5fdaa11cd1e62cae6746acb.

The original commit replaces 'which' with 'command'. Sadly most of
them are not equivalent and for 'which -a', there is no easy
replacements that would not reimplement PATH parsing logic. Hence
revert. Keeping a dependency on which is absolutely fine.

Signed-off-by: Clemens Fruhwirth <clemens@endorphin.org>
3 years agobuild: fix checks for GCC11
Petr Štetiar [Mon, 1 Mar 2021 12:29:19 +0000 (13:29 +0100)]
build: fix checks for GCC11

Fedora 34 already uses GCC11.

Reported-by: Marcin Juszkiewicz <marcin-openwrt@juszkiewicz.com.pl>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agobuild: make GCC version 6+ minimal host build requirement
Petr Štetiar [Mon, 1 Mar 2021 12:27:17 +0000 (13:27 +0100)]
build: make GCC version 6+ minimal host build requirement

Currently minimal GNU supported GCC version is 7 (from May 2, 2017),
buildbots are using default GCC version 6 on Debian 9 (old stable),
current Debian stable has GCC version 8.3.0.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agokernel: generic: add further missing symbols
Petr Štetiar [Fri, 26 Feb 2021 16:12:15 +0000 (17:12 +0100)]
kernel: generic: add further missing symbols

Compile testing i.mx6 with ALL_KMODS=y, PACKAGE_perf=y and bunch of
tracing/probing symbols has unveiled bunch of missing config options so
add them.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agoimx6: add 5.10 as testing kernel
Petr Štetiar [Fri, 26 Feb 2021 16:15:34 +0000 (17:15 +0100)]
imx6: add 5.10 as testing kernel

So anyone interested can help with testing.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agoimx6: refresh kernel config with 5.10 symbols
Petr Štetiar [Fri, 26 Feb 2021 15:39:05 +0000 (16:39 +0100)]
imx6: refresh kernel config with 5.10 symbols

Just by running `make kernel_oldconfig` and unsetting following options
manually as those cores are cortex-a7 based and thus irrelevant for the
currently default cortex-a9 used cores.

 CONFIG_CLK_IMX6SL is not set
 CONFIG_CLK_IMX6SX is not set
 CONFIG_CLK_IMX6UL is not set

Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agoimx6: copy kernel 5.4 config and patches to 5.10
Petr Štetiar [Fri, 26 Feb 2021 15:23:39 +0000 (16:23 +0100)]
imx6: copy kernel 5.4 config and patches to 5.10

So the changes are visible in the next step. Removed upstreamed patches.

Signed-off-by: Petr Štetiar <ynezz@true.cz>
3 years agobuild: use numeric uid/gid on cpio calls
Felix Fietkau [Wed, 3 Mar 2021 15:09:04 +0000 (16:09 +0100)]
build: use numeric uid/gid on cpio calls

There are systems that don't have the 'root' group, so don't rely on host
specific user/group names

Signed-off-by: Felix Fietkau <nbd@nbd.name>
3 years agoarm-trusted-firmware-mediatek: remove {e,sd}mmc headers
Daniel Golle [Wed, 3 Mar 2021 00:56:31 +0000 (00:56 +0000)]
arm-trusted-firmware-mediatek: remove {e,sd}mmc headers

Turned out those are simply MBR with active boot partition. And not
needed at all on emmc. Remove them as ptgen can now generate hybrid
MBR sufficient to boot MT7622 from SD Card.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: change image generation
Oskari Lemmela [Tue, 2 Mar 2021 19:42:11 +0000 (21:42 +0200)]
mediatek: mt7622: change image generation

- set only one EFI system partition
- use shorter path for DEVICE_DTS file

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agomediatek: mt7622: use ptgen generated MBR header
Oskari Lemmela [Tue, 2 Mar 2021 19:42:10 +0000 (21:42 +0200)]
mediatek: mt7622: use ptgen generated MBR header

mt7622 uses MBR partition for booting from SD card.
Add hybrid MBR entry with boot flag after PMBR entry.

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agofirmware-utils/ptgen: change GPT to LBA addressing
Oskari Lemmela [Tue, 2 Mar 2021 19:42:09 +0000 (21:42 +0200)]
firmware-utils/ptgen: change GPT to LBA addressing

Do not align partitions with sectors. Only kb align
for GPT is supported.

Use 254 heads and 63 sectors for PMBR.

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agofirmware-utils/ptgen: add support for hybrid MBR
Oskari Lemmela [Tue, 2 Mar 2021 19:42:08 +0000 (21:42 +0200)]
firmware-utils/ptgen: add support for hybrid MBR

Adding -H option copies partition to MBR after pmbr entry.
Max 3 partitions can be copied to MBR.

Hybrid MBR is needed only in special cases.
For example mt7622 SD card boot needs MBR entry with boot
flag enabled.

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agofirmware-utils/ptgen: fix partition guid and name
Oskari Lemmela [Tue, 2 Mar 2021 19:42:07 +0000 (21:42 +0200)]
firmware-utils/ptgen: fix partition guid and name

guid and name function should be called before storing
partition info.

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agofirmware-utils/ptgen: fix compile warnings
Oskari Lemmela [Tue, 2 Mar 2021 19:42:06 +0000 (21:42 +0200)]
firmware-utils/ptgen: fix compile warnings

remove extra arguments from printf message

Signed-off-by: Oskari Lemmela <oskari@lemmela.net>
3 years agofstools: update to git HEAD
Daniel Golle [Tue, 2 Mar 2021 19:15:28 +0000 (19:15 +0000)]
fstools: update to git HEAD

 19d7d93 libfstools: partname: several fixes

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoprocd: update to git HEAD
Daniel Golle [Tue, 2 Mar 2021 19:12:25 +0000 (19:12 +0000)]
procd: update to git HEAD

 64e9f3a procd: fix compilation with newer musl

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agolibubox: update to git HEAD
Daniel Golle [Tue, 2 Mar 2021 19:09:11 +0000 (19:09 +0000)]
libubox: update to git HEAD

 2e52c7e libubox: fix BLOBMSG_CAST_INT64 (do not override BLOBMSG_TYPE_DOUBLE)

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobmips: switch to upstream brcmnand patch
Álvaro Fernández Rojas [Tue, 2 Mar 2021 17:05:32 +0000 (18:05 +0100)]
bmips: switch to upstream brcmnand patch

This patch has been accepted upstream for v5.13.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agoRevert "toolchain: add autoconf-lean"
Daniel Golle [Tue, 2 Mar 2021 11:34:24 +0000 (11:34 +0000)]
Revert "toolchain: add autoconf-lean"

autoconf-lean was not ready yet. Revert it until things get sorted.

This reverts commit 32c664ff02910bf39a3fbd5a5a4a8bff3191dd03.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoRevert "build: use config.site generated by autoconf-lean, drop hardcoded sitefiles"
Daniel Golle [Tue, 2 Mar 2021 11:34:22 +0000 (11:34 +0000)]
Revert "build: use config.site generated by autoconf-lean, drop hardcoded sitefiles"

This reverts commit f439e291304a93b982e912dc91b80ca950a594f3.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoRevert "toolchain/autoconf-lean: add check for ssize_t"
Daniel Golle [Tue, 2 Mar 2021 11:34:18 +0000 (11:34 +0000)]
Revert "toolchain/autoconf-lean: add check for ssize_t"

This reverts commit 6b7b064492ddf339d0213779fb84d50ee3dc5eb7.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel-5.10: backport chacha non block size optimizations
Jason A. Donenfeld [Tue, 2 Mar 2021 08:51:42 +0000 (09:51 +0100)]
kernel-5.10: backport chacha non block size optimizations

These make a big difference when doing WireGuard with small armv7
routers, and the 5.4 backport already has it.

Suggested-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: David Bauer <mail@david-bauer.net>
Cc: Petr Štetiar <ynezz@true.cz>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
3 years agokernel-5.4: backport fd16931a2f51 for chacha neon
Jason A. Donenfeld [Tue, 2 Mar 2021 08:24:45 +0000 (09:24 +0100)]
kernel-5.4: backport fd16931a2f51 for chacha neon

Without this patch, the chacha block counter is not incremented on neon
rounds, resulting in incorrect calculations and corrupt packets.

This also switches to using `--no-numbered --zero-commit` so that future
diffs are smaller.

Reported-by: Hans Geiblinger <cybrnook2002@yahoo.com>
Reviewed-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Cc: David Bauer <mail@david-bauer.net>
Cc: Petr Štetiar <ynezz@true.cz>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
3 years agobcm53xx: add support for reading NVRAM based on DT mapping
Rafał Miłecki [Tue, 2 Mar 2021 08:04:48 +0000 (09:04 +0100)]
bcm53xx: add support for reading NVRAM based on DT mapping

This is required for devices that use NVRAM data for detecting currently
used firmware partition (e.g. Linksys devices).

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agouboot-mediatek: bpi-r64: make sure eMMC installation runs only once
Daniel Golle [Tue, 2 Mar 2021 01:42:34 +0000 (01:42 +0000)]
uboot-mediatek: bpi-r64: make sure eMMC installation runs only once

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: make sure image generation can run in parallel
Daniel Golle [Tue, 2 Mar 2021 01:41:00 +0000 (01:41 +0000)]
mediatek: mt7622: make sure image generation can run in parallel

The previous approach of referencing artifacts in follow-up artifacts
can't work with parallel builds in the current way image.mk is built.
Refactor things so this is not needed.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: blue LED should be active high as well
Daniel Golle [Mon, 1 Mar 2021 21:46:38 +0000 (21:46 +0000)]
mediatek: mt7622: blue LED should be active high as well

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agouboot-mediatek: bpi-r64: fix eMMC installation menu label
Daniel Golle [Mon, 1 Mar 2021 20:09:31 +0000 (20:09 +0000)]
uboot-mediatek: bpi-r64: fix eMMC installation menu label

Change boot menu label for eMMC installation to tell what it does now.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: bpi-r64: fix sysupgrade on empty disk
Daniel Golle [Mon, 1 Mar 2021 19:31:51 +0000 (19:31 +0000)]
mediatek: mt7622: bpi-r64: fix sysupgrade on empty disk

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: bpi-r64: simplify eMMC install procedure
Daniel Golle [Mon, 1 Mar 2021 14:07:33 +0000 (14:07 +0000)]
mediatek: mt7622: bpi-r64: simplify eMMC install procedure

Write everything needed for eMMC install into the gaps between
partitions on SD card. In that way, installation to eMMC only needs
the SD card, no additional files need to be loaded via TFTP any more.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: add the latest mtd patch extending ofpart parser
Rafał Miłecki [Mon, 1 Mar 2021 17:12:48 +0000 (18:12 +0100)]
kernel: add the latest mtd patch extending ofpart parser

This adds the latest version of ofpart commit. It hopefully
1. Doesn't break compilation
2. Doesn't break partitioning
(this time).

It's required to implement fixed partitioning with some quirks. It's
required by bcm53xx, bcm4908, kirkwood, lantiq and mvebu.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agotools/xxd: fix source URL
Daniel Golle [Mon, 1 Mar 2021 16:38:46 +0000 (16:38 +0000)]
tools/xxd: fix source URL

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobmips: switch to upstream SMP CPU mappings patch
Álvaro Fernández Rojas [Mon, 1 Mar 2021 16:26:00 +0000 (17:26 +0100)]
bmips: switch to upstream SMP CPU mappings patch

The patch has been accepted upstream with some minor modifications.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobuild: prevent dupes in autotools.mk
Sebastian Kemper [Mon, 1 Mar 2021 16:13:07 +0000 (17:13 +0100)]
build: prevent dupes in autotools.mk

autotools.mk does not have any protection currently that would prevent
it from being sourced multiple times. Note that both package.mk and
host-build.mk source autotools.mk. So any package Makefile that includes
both will cause hooks to be added twice (at least twice).

This is fixed by declaring a new variable, __autotools_inc, and only
continuing if this variable doesn't equal 1. The same is done by
rules.mk already.

Also, this commit does away with an ifneq that checks PKG_FIXUP (instead
of HOST_FIXUP) for patch-libtool before adding to the host pre-configure
hook. This does not make sense.

The second ifneq is amended. The current one manually does what the
define patch_libtool_host is already doing. It can just use the define.

Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
3 years agotoolchain/autoconf-lean: add check for ssize_t
Sebastian Kemper [Mon, 1 Mar 2021 16:02:23 +0000 (17:02 +0100)]
toolchain/autoconf-lean: add check for ssize_t

This was provided by the old static config.site files and is required by
some software, i.e. freeswitch.

Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
3 years agotools: xxd: use more convenient source tarball
Daniel Golle [Mon, 1 Mar 2021 14:06:33 +0000 (14:06 +0000)]
tools: xxd: use more convenient source tarball

Don't download all of vim just to build xxd. Use a tight tarball
containing only xxd sources instead.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: enable CONFIG_SPI_MTK_NOR
Daniel Golle [Mon, 1 Mar 2021 11:56:23 +0000 (11:56 +0000)]
mediatek: mt7622: enable CONFIG_SPI_MTK_NOR

Somehow this got disabled in the transition to 5.10.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agomediatek: mt7622: rename mt7622-ubi to mt7622-rfb1-ubi
Daniel Golle [Mon, 1 Mar 2021 11:42:27 +0000 (11:42 +0000)]
mediatek: mt7622: rename mt7622-ubi to mt7622-rfb1-ubi

This profile is meant to be used on MT7622 rfb1 AP, indicate that in
the name to make things less confusing.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobcm4908: backport Ethernet driver fixes from the 5.12
Rafał Miłecki [Mon, 1 Mar 2021 07:14:49 +0000 (08:14 +0100)]
bcm4908: backport Ethernet driver fixes from the 5.12

The most noticeable one is fix for RX stopping on high traffic.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>