openwrt/staging/robimarko.git
3 years agogeneric: fix debloat dma-buf 5.10 patch
Álvaro Fernández Rojas [Sat, 21 Aug 2021 08:50:21 +0000 (10:50 +0200)]
generic: fix debloat dma-buf 5.10 patch

The current patch produces the following error when CONFIG_DMABUF_HEAPS is
enabled:
drivers/built-in.a: member drivers/dma-buf/heaps in archive is not an object

Fixes: b10d6044599d ("kernel: add linux 5.10 support")
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agogeneric: add missing 5.10 config symbols
Álvaro Fernández Rojas [Sat, 21 Aug 2021 08:48:07 +0000 (10:48 +0200)]
generic: add missing 5.10 config symbols

These symbols are needed for bcm27xx 5.10 kernel support.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobcm27xx-gpu-fw: update to latest version
Álvaro Fernández Rojas [Sat, 21 Aug 2021 17:05:14 +0000 (19:05 +0200)]
bcm27xx-gpu-fw: update to latest version

Multiple firmware fixes needed for kernel updates.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agobcm27xx-userland: update to latest version
Álvaro Fernández Rojas [Sat, 21 Aug 2021 17:03:24 +0000 (19:03 +0200)]
bcm27xx-userland: update to latest version

Properly recognise all BCM2711 variants

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
3 years agomediatek: mt7623: backport musb, improve HDMI console
Daniel Golle [Sat, 21 Aug 2021 14:32:13 +0000 (15:32 +0100)]
mediatek: mt7623: backport musb, improve HDMI console

Backport support for dual-role USB 2.0 as that's what is actually
built-into MT7623.
Improve HDMI console by enabling VT and setting up tty1..tty6.
Re-add accidentally removed CONFIG_ARM_ARCH_TIMER.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoprereq-build: require python3-distutils
Andre Heider [Tue, 10 Aug 2021 10:22:17 +0000 (12:22 +0200)]
prereq-build: require python3-distutils

Debian and Ubuntu ship a python3-minimal package which does not include
the distutils module. This is not supported by upstream and can be
considered a broken python distribution.

In practice, many scripts depend on said module, and this is a reoccuring
pain point for building various OpenWrt packages.

Require and check for said module, enough time has been wasted on this.

A list of just the most recent issues:
https://github.com/openwrt/packages/pull/16304
https://github.com/openwrt/packages/pull/16027
https://github.com/openwrt/packages/pull/15443
https://github.com/openwrt/packages/pull/14394
https://github.com/openwrt/packages/pull/12909
https://github.com/openwrt/packages/issues/12443
https://github.com/openwrt/packages/pull/11035
https://github.com/openwrt/packages/issues/10993

Signed-off-by: Andre Heider <a.heider@gmail.com>
3 years agoramips: fix ethernet MAC address on Omega2
Michael Siegenthaler [Thu, 19 Aug 2021 00:58:56 +0000 (17:58 -0700)]
ramips: fix ethernet MAC address on Omega2

According to https://docs.onion.io/omega2-docs/mac-address.html, 0x28 is
the correct location to read the address on Onion Omega 2(+) devices.

This fixes a regression introduced by commit 77e850fe76e3 ("ramips: tidy up
MAC address setup for Linkit Smart and Omega2"), which was a cleanup that
intended to preserve existing behavior. In my testing with v19.07.7,
however, the MAC address determined from the device tree takes precedence
over the one set by 02_network, so the aforementioned commit actually
changed the behavior.

Signed-off-by: Michael Siegenthaler <msiegen@google.com>
[Adapt patch to nvmem usage]
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoltq-vdsl/ltq-vdsl-app: clean up configure args
Jan Hoffmann [Thu, 19 Aug 2021 16:51:45 +0000 (18:51 +0200)]
ltq-vdsl/ltq-vdsl-app: clean up configure args

The userspace application now uses the model=full option to match the
configuration of the kernel module. The source no longer contains SOAP
support, which was the primary reason to build only typical instead
of full before.

This makes several CLI commands, which were already supported in the
kernel module, available in the userspace application. For example, this
includes bbsg which allows to get information about VDSL2 bands.

Some previously applied build options were redundant. Disabling ADSL MIB
support is unnecessary, as it only applies to Danube. ADSL LED support
is no longer included in the source. ReTx counters are already included
with model type full.

This increases the size of the userspace application by approximately
15 kB (uncompressed). The kernel module does not change at all.

Signed-off-by: Jan Hoffmann <jan@3e8.eu>
3 years agofirmware-utils: add support for ZyXEL ZLD firmware
Vincent Wiemann [Mon, 2 Aug 2021 22:21:46 +0000 (00:21 +0200)]
firmware-utils: add support for ZyXEL ZLD firmware

This commit adds a tool for extracting and creating firmware archives for:
- ZyXEL NWA5121-N
- ZyXEL NWA5121-NI
- ZyXEL NWA5123-NI
- ZyXEL NWA5123-AC
- ZyXEL NWA5301
- ZyXEL NWA3560N v2
- ZyXEL WAC6100
- ZyXEL WAC6500

And maybe others...

Signed-off-by: Vincent Wiemann <vincent.wiemann@ironai.com>
3 years agodropbear: allow to use with xinetd
Fritz D. Ansel [Mon, 2 Aug 2021 18:38:31 +0000 (20:38 +0200)]
dropbear: allow to use with xinetd

with xinetd allowed+blocked (ipv6) hosts could be set
what is not possible with stock dropbear package

The file size increased 12 Bytes, so this "opimisation" did not really helped.
Within a compressed storage format it is 0..

ipk: 111.171 -> 111.361 = 190 bytes
bin: 215.128 -> 215.140 =  12 bytes

Signed-off-by: Fritz D. Ansel <fdansel@yandex.ru>
3 years agokernel: bump 5.4 to 5.4.142
John Audia [Wed, 18 Aug 2021 12:06:08 +0000 (08:06 -0400)]
kernel: bump 5.4 to 5.4.142

Removed upstreamed:
  hack-5.4/991-platform-x86-pcengines-apuv2-revert-simswitch.patch

All other 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 5.4.141
John Audia [Sun, 15 Aug 2021 13:13:34 +0000 (09:13 -0400)]
kernel: bump 5.4 to 5.4.141

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 5.10 to 5.10.60
Rui Salvaterra [Wed, 18 Aug 2021 23:19:24 +0000 (00:19 +0100)]
kernel: bump 5.10 to 5.10.60

No deleted or manually refreshed patches.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agogdb: update to 10.2
Rosen Penev [Sat, 28 Mar 2020 23:55:45 +0000 (16:55 -0700)]
gdb: update to 10.2

Replace some OpenWrt patches with openembedded ones for easier
maintainability. Remove several outdated ones as well.

Replace PKG_RELEASE with AUTORELEASE to avoid manual bumps.

Remove !arc dependency as it is supported upstream now.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agomediatek: mt7623: enable more hardware features
Daniel Golle [Sat, 21 Aug 2021 03:34:15 +0000 (04:34 +0100)]
mediatek: mt7623: enable more hardware features

In order to make HDMI console available on the BananaPi BPi-R2 select
various Kconfig symbols which are useful for systems with graphics.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agouboot-mediatek: pass console=tty1 for BPi-R2
Daniel Golle [Sat, 21 Aug 2021 03:31:57 +0000 (04:31 +0100)]
uboot-mediatek: pass console=tty1 for BPi-R2

BananaPi BPi-R2 comes with HDMI and MIPI-DSI. Use dislpay facility in
Linux by add "console=tty1" boot argument.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: Do not deactivate staging drivers
Hauke Mehrtens [Sun, 15 Aug 2021 15:39:40 +0000 (17:39 +0200)]
kernel: Do not deactivate staging drivers

Do not deactivate the kernel configuration symbol CONFIG_STAGING in the
target configurations any more. This prevented the build of the exfat.ko
for example.

Fixes: FS#3979
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agolayerscape: Do not activate ext2 and ext3 driver options
Hauke Mehrtens [Sat, 14 Aug 2021 12:04:02 +0000 (14:04 +0200)]
layerscape: Do not activate ext2 and ext3 driver options

The ext3 driver was already removed, the kernel config options are only
there for backwards compatibility. The eth4 driver takes care of ext3
file systems. The ext4 driver also handled ext2 file systems.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoarchs38: Do not activate ext3 driver options
Hauke Mehrtens [Sat, 14 Aug 2021 12:02:41 +0000 (14:02 +0200)]
archs38: Do not activate ext3 driver options

The ext3 driver was already removed, the kernel config options are only
there for backwards compatibility. The eth4 driver takes care of ext3
file systems.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoat91bootstrap: Fix compile with binutils 2.36
Hauke Mehrtens [Fri, 20 Aug 2021 22:08:42 +0000 (00:08 +0200)]
at91bootstrap: Fix compile with binutils 2.36

This fixes a build problem seen after binutils 2.36 is used by default.

Fixes: 3f41153b1c28 ("toolchain/binutils: switch to version 2.36.1 by default")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agouboot-at91: fix dtc compilation on host gcc 10
Hauke Mehrtens [Fri, 20 Aug 2021 22:07:54 +0000 (00:07 +0200)]
uboot-at91: fix dtc compilation on host gcc 10

Backport a patch from upstream U-Boot to fix the compile with host GCC 10.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agowolfssl: fix build with GCC 10 on 32 x86 targets
Stijn Tintel [Fri, 20 Aug 2021 20:45:20 +0000 (23:45 +0300)]
wolfssl: fix build with GCC 10 on 32 x86 targets

Backport upstream patch to fix build with GCC 10 on 32 x86 targets.

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
3 years agomediatek: convert mtd-mac-address to nvmem for MT7629 RFB
Adrian Schmutzler [Fri, 20 Aug 2021 12:42:56 +0000 (14:42 +0200)]
mediatek: convert mtd-mac-address to nvmem for MT7629 RFB

This conversion appears to have been overlooked since it's in a
kernel patch.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agolantiq: convert mtd-mac-address to nvmem for Fritz!Box 736x
Adrian Schmutzler [Fri, 20 Aug 2021 12:27:13 +0000 (14:27 +0200)]
lantiq: convert mtd-mac-address to nvmem for Fritz!Box 736x

Convert this series by moving the definitions to the individual
devices.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: convert mtd-mac-address to nvmem for UniElec devices
Adrian Schmutzler [Thu, 19 Aug 2021 22:18:12 +0000 (00:18 +0200)]
ramips: convert mtd-mac-address to nvmem for UniElec devices

Convert this series by moving the definitions to the individual
devices.

Now all devices on ramips are converted.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoath79: remove redundant nvmem definitions
Adrian Schmutzler [Thu, 19 Aug 2021 22:09:58 +0000 (00:09 +0200)]
ath79: remove redundant nvmem definitions

Due to use of a script when migrating from mtd-mac-address, a few
of the definitions are redundant in DTSI and DTS files. Remove
those and consolidate the definitions in parent DTSI files in a
few cases.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: remove redundant nvmem definitions
Adrian Schmutzler [Wed, 18 Aug 2021 23:50:27 +0000 (01:50 +0200)]
ramips: remove redundant nvmem definitions

Due to use of a script when migrating from mtd-mac-address, a few
of the definitions are redundant in DTSI and DTS files. Remove
those.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: fix indent in DTS for GL.iNet VIXMINI/microuter-N300
Adrian Schmutzler [Wed, 18 Aug 2021 23:32:56 +0000 (01:32 +0200)]
ramips: fix indent in DTS for GL.iNet VIXMINI/microuter-N300

Use correct indent.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: convert mtd-mac-address to nvmem for D-Link DIR-8xx
Adrian Schmutzler [Wed, 18 Aug 2021 23:16:39 +0000 (01:16 +0200)]
ramips: convert mtd-mac-address to nvmem for D-Link DIR-8xx

Convert this series by moving the definitions to the individual
devices.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoselinux-policy: update to version 0.9
Daniel Golle [Fri, 20 Aug 2021 00:46:14 +0000 (01:46 +0100)]
selinux-policy: update to version 0.9

 592ac0f add a note
 4bacd14 sslcertfile: list /etc/ssl
 7bdefa4 example: indicate that skip is an option
 d1e9a85 wifi: sys pipe usage
 eb903e1 README: add note about policycoreutils-setfiles weak dependency
 762e011 ttyd: signull all subjects
 fbfc079 acme: add basic support for acme_cleanup.sh and acme_setup.sh
 9ac7592 acme: transition to sys.subj on generic initscript execution
 f3dd1ba acme: missing rules related to sys.subj trans on file.initscriptfile
 ae273fa odhcp6c/netifd: support drop-in directories
 5fa9b41 subj: do not encourage misconfiguration
 44722b6 blockd, logd, odhcpc6, ubiutil, mtdstordev
 a775d93 21.02 related
 a473691 rcboot runs rcuhttpd which creates /tmp/etc for /tmp/etc/uhttpd
 290e9fb rcuhttpd: related to rcboot and uci-defaults
 3fc0d8b rcuhttpd: lists /etc/uci-defaults
 1f5ef48 removes ubvol.lock policy and adds move mtd/ubi partitions

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoimage: fix build with SELinux
Daniel Golle [Thu, 19 Aug 2021 00:23:41 +0000 (01:23 +0100)]
image: fix build with SELinux

The option '-xattr' for mksquashfs4 should be '-xattrs' which lead to
build failure with SELinux enabled. Add the missing 's'.

Fixes: 4baf47b9a8 ("images: squashfs: xattrs should not depend on buld host")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoath79: convert remaining mtd-mac-address cases to nvmem
Adrian Schmutzler [Wed, 18 Aug 2021 18:06:29 +0000 (20:06 +0200)]
ath79: convert remaining mtd-mac-address cases to nvmem

Since the nvmem-based approach for retrieving MAC addresses
appears to depend on the addresses being set up after the
partitions, it is no longer possible to keep the MAC address
setup in shared DTSI files while the partitions itself are
set up in DTS files for the individual devices.

In ath79 the firmware partition is typically located somewhere
"in the middle" of the partition table. Thus, it's not trivial
to share the partitions containing MAC address information in
a common DTSI (like we did in some cases on ramips).

In this commit, MAC address setup is thus moved to the relevant
partitions, and in most cases needs to be duplicated. While
the duplication is not really nice, it eventually provides a
cleaner and more tidy setup, making the DTS(I) file
fragmentation a bit more logical. This should also help
with adding new devices, as information is distributed across
less locations.

For consistency, this commit also moves the mtd-cal-data property
"down" together with the MAC address setup, so it's not based
on a partition before the latter is defined either. (This is
only done for those files touched due to nvmem conversion.)

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agoramips: convert most mtd-mac-address cases in DTSI to nvmem
Adrian Schmutzler [Tue, 17 Aug 2021 18:53:27 +0000 (20:53 +0200)]
ramips: convert most mtd-mac-address cases in DTSI to nvmem

Convert most of the cases from mtd-mac-address to nvmem where
MAC addresses are set in the DTSI, but the partitions are only
located in the device DTS. This posed some problems earlier, since
in these cases we are using partitions before they are defined,
and the nvmem system did not seem to like that.

There have been a few different resolution approaches, based on
the different tradeoffs of deduplication vs. maintainability:

 1. In many cases, the partition tables were identical except for
    the firmware partition size, and the firmware partition was
    the last in the table.
    In these cases, the partition table has been moved to the
    DTSI, and only the firmware partition's "reg" property has
    been kept in the DTS files. So, the updated nvmem definition
    could stay in the DTSI files as well.

 2. For all other cases, splitting up the partition table would
    have introduced additional complexity. Thus, the nodes to be
    converted to nvmem have been moved to the DTS files where the
    partitioning was defined.

 3. For Netgear EX2700 and WN3000RP v3, the remaining DTSI file
    was completely dissolved, as it was quite small and the name
    was not really nice either.

 4. The D-Link DIR-853 A3 was converted to nvmem as well, though
    it is just a plain DTS file not taken care of in the first
    wave.

In addition, some minor rearrangements have been made for tidyness.

Not covered (yet) by this patch are:

 * Various unielec devices
 * The D-Link DIR-8xx family

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agobcm27xx-userland: factor out a -dev package
Michael Heimpold [Sat, 27 Mar 2021 19:48:30 +0000 (20:48 +0100)]
bcm27xx-userland: factor out a -dev package

Installing headers and static libraries to the target system seems
to be not required for most use cases, so let's factor them
out into a dedicated -dev package.

This cuts down to disk usage to around 50% of the original
package to ~ 2MB - not that disk space is an issue normally,
but when using inside an initramfs only project, it counts.

Signed-off-by: Michael Heimpold <mhei@heimpold.de>
3 years agotoolchain/gcc: switch to version 10 by default
Stijn Tintel [Thu, 13 May 2021 16:45:48 +0000 (19:45 +0300)]
toolchain/gcc: switch to version 10 by default

Runtime-tested on:
* ath79
* bcm27xx/bcm2708
* bcm27xx/bcm2709
* bcm27xx/bcm2711
* mvebu/cortexa53
* octeon
* realtek
* x86/64

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agotoolchain/binutils: switch to version 2.36.1 by default
Stijn Tintel [Thu, 13 May 2021 16:54:34 +0000 (19:54 +0300)]
toolchain/binutils: switch to version 2.36.1 by default

Runtime-tested on:
* ath79
* bcm27xx/bcm2708
* bcm27xx/bcm2709
* bcm27xx/bcm2711
* mvebu/cortexa53
* octeon
* realtek
* x86/64

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoumbim: bump to git HEAD
Stijn Tintel [Wed, 18 Aug 2021 17:45:14 +0000 (20:45 +0300)]
umbim: bump to git HEAD

 de56231 umbim: fix compilation with GCC 10

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agobusybox: fix compilation with GCC 10
Stijn Tintel [Thu, 13 May 2021 13:23:26 +0000 (16:23 +0300)]
busybox: fix compilation with GCC 10

When compiling busybox with GCC 10 and CONFIG_PKG_ASLR_PIE_ALL=y, there
are hundreds of errors like:

relocation R_MIPS16_26 against `xzalloc' cannot be used when making a
shared object; recompile with -fPIC

Simply solve this by no longer disabling PKG_ASLR_PIE, so that $(FPIC)
is properly added to the CFLAGS and LDFLAGS.

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agoipq40xx: fix Edgecore ECW5211 boot
Stijn Tintel [Wed, 18 Aug 2021 13:11:11 +0000 (16:11 +0300)]
ipq40xx: fix Edgecore ECW5211 boot

The bootloader will look for a configuration section named ap.dk01.1-c2
in the FIT image. If this doesn't exist, the device won't boot.

Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
3 years agoramips: add WPS button for newifi d1
David Yang [Sun, 25 Jul 2021 13:21:02 +0000 (21:21 +0800)]
ramips: add WPS button for newifi d1

This device has a WPS button under WiFi antenna cover, add it to dts.

Signed-off-by: David Yang <mmyangfl@gmail.com>
3 years agoramips: mt7628: move mtd-eeprom out of root DTSI
Adrian Schmutzler [Tue, 17 Aug 2021 12:07:57 +0000 (14:07 +0200)]
ramips: mt7628: move mtd-eeprom out of root DTSI

The mt76x8 subtarget is the only one in ramips that stores the
mediatek,mtd-eeprom property directly in the "root" mt7628an.dtsi.

This is not optimal for a few different reasons:

 * If you don't really know it or are used to other (sub)targets,
   the property will be set somewhat magically.
 * The property is set based on &factory partition before (if at all)
   this partition is defined.
 * There are several devices that have different offset or even
   different partitions to read from, which will then be overwritten
   in the DTS files. Thus, definitions are scattered between root
   DTSI and individual files.

Based on these circumstances, the "root" definition is removed and
the property is added to the device-based DTS(I) files where needed
and applicable. This should be easier to grasp for unexperienced
developers and will move the property closer to the partition
definitions.

Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
3 years agokernel: bump 5.10 to 5.10.59
Rui Salvaterra [Sun, 15 Aug 2021 22:54:28 +0000 (23:54 +0100)]
kernel: bump 5.10 to 5.10.59

No deleted or manually refreshed patches.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agobase-files: rename 'sdcard' to 'legacy-sdcard'
Daniel Golle [Sat, 7 Aug 2021 13:30:53 +0000 (14:30 +0100)]
base-files: rename 'sdcard' to 'legacy-sdcard'

While an image layout based on MBR and 'bootfs' partition may be easy
to understand for users who are very used to the IBM PC and always have
the option to access the SD card outside of the device (and hence don't
really depend on other recovery methods or dual-boot), in my opinion
it's a dead end for many desirable features on embedded systems,
especially when managed remotely (and hence without an easy option to
access the SD card using another device in case things go wrong, for
example).

Let me explain:

* using a MSDOS/VFAT filesystem to store kernel(s) is problematic, as a
  single corruption of the bootfs can render the system into a state
  that it no longer boots at all. This makes dual-boot useless, or at
  least very tedious to setup with then 2 independent boot partitions
  to avoid the single point of failure on a "hot" block (the FAT index
  of the boot partition, written every time a file is changed in
  bootfs). And well: most targets even store the bootloader environment
  in a file in that very same FAT filesystem, hence it cannot be used
  to script a reliable dual-boot method (as loading the environment
  itself will already fail if the filesystem is corrupted).

* loading the kernel uImage from bootfs and using rootfs inside an
  additional partition means the bootloader can only validate the
  kernel -- if rootfs is broken or corrupted, this can lead to a reboot
  loop, which is often a quite costly thing to happen in terms of
  hardware lifetime.

* imitating MBR-boot behavior with a FAT-formatted bootfs partition
  (like IBM PC in the 80s and 90s) is just one of many choices on
  embedded targets. There are much better options with modern U-Boot
  (which is what we use and build from source for all targets booting
  off SD cards), see examples in mediatek/mt7622 and mediatek/mt7623.

Hence rename the 'sdcard' feature to 'legacy-sdcard', and prefix
functions with 'legacy_sdcard_' instead of 'sdcard_'.

Tested-by: Stijn Tintel <stijn@linux-ipv6.be>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoprocd: add missing dependency and fix empty mount triggers
Daniel Golle [Sun, 15 Aug 2021 16:41:59 +0000 (17:41 +0100)]
procd: add missing dependency and fix empty mount triggers

procd.sh:
 Instead of triggering on every mount.add event, there should be no
 mount trigger at all in case none of the directories passed to
 procd_add_*_mount_trigger() are located on a mountpoint configured in
 /etc/config/fstab.

uxc:
 add missing dependency on rpcd.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agoprocd: update to git HEAD
Daniel Golle [Sun, 15 Aug 2021 14:41:50 +0000 (15:41 +0100)]
procd: update to git HEAD

 040fecc system: fix issues reported by Coverity
 48f481b service: make sure string read is null terminated
 16dbc2a uxc: fix a bunch of issues discovered by Coverity
 ff9002f uxc: fix help output
 104b49d uxc: support config in uvol

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agofirewall: update to git HEAD
Kevin Darbyshire-Bryant [Sat, 14 Aug 2021 18:10:32 +0000 (19:10 +0100)]
firewall: update to git HEAD

40e5f6a ipsets: permit default timeout of 0

Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
3 years agouboot-rockchip: update to v2021.07
David Bauer [Mon, 9 Aug 2021 23:05:50 +0000 (01:05 +0200)]
uboot-rockchip: update to v2021.07

Tested on NanoPi R2S

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agokernel: bump 5.4 to 5.4.140
John Audia [Thu, 12 Aug 2021 13:28:18 +0000 (09:28 -0400)]
kernel: bump 5.4 to 5.4.140

Removed upstreamed bcm27xx/patches-5.4:
    950-0977-USB-gadget-f_hid-avoid-crashes-and-log-spam.patch
    950-0980-SQUASH-USB-gadget-f_hid-remove-more-spam.patch

All other 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 5.4.139
John Audia [Mon, 9 Aug 2021 14:47:18 +0000 (10:47 -0400)]
kernel: bump 5.4 to 5.4.139

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 agoprocd: update to git HEAD
Daniel Golle [Sat, 14 Aug 2021 18:01:37 +0000 (19:01 +0100)]
procd: update to git HEAD

 48638ad hotplug-dispatch: yet another rare memory leak disovered by Coverity
 459b3e8 jail: fix several issues discovered by Coverity
 2562e2b ujail-console: add missing error handling discovered by coverity

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agofstools: update to git HEAD
Daniel Golle [Sat, 14 Aug 2021 17:56:12 +0000 (18:56 +0100)]
fstools: update to git HEAD

 629726d blockd: fix resource leak discovered by coverity scan
 68ae639 libubi: fix several issues discovered by Coverity
 a77c4fa ubi: fix resource leak in legacy codepath
 2e3aca2 block: fix two resources leaks discovered by Coverity

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agokernel: bump 5.10 to 5.10.58
Rui Salvaterra [Thu, 12 Aug 2021 23:16:29 +0000 (00:16 +0100)]
kernel: bump 5.10 to 5.10.58

No deleted or manually refreshed patches.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agokernel: bump 5.10 to 5.10.57
Rui Salvaterra [Mon, 9 Aug 2021 21:05:05 +0000 (22:05 +0100)]
kernel: bump 5.10 to 5.10.57

No deleted or manually refreshed patches.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agoipq40xx: ar40xx: reset port status register
David Bauer [Fri, 13 Aug 2021 21:28:44 +0000 (23:28 +0200)]
ipq40xx: ar40xx: reset port status register

This resolves incosnsitencies of the configured RX / TX flow control
modes between different boards or bootloaders.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoipq40xx: ar40xx: use FIELD_GET macro
David Bauer [Fri, 13 Aug 2021 22:22:28 +0000 (00:22 +0200)]
ipq40xx: ar40xx: use FIELD_GET macro

This improves code readability.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoipq40xx: increase EX6150v2 SPI frequency
David Bauer [Sat, 14 Aug 2021 00:20:40 +0000 (02:20 +0200)]
ipq40xx: increase EX6150v2 SPI frequency

The chip supports clock speeds up to 50 MHz, however it won't even read
the chip-id correctly at this frequency.

45 MHz however works reliable.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agoprocd: update to git HEAD
Daniel Golle [Fri, 13 Aug 2021 03:45:17 +0000 (04:45 +0100)]
procd: update to git HEAD

 9f233f5 system: make rootfs type accessible through board call

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
3 years agobcm4908: add missing config symbol
David Bauer [Thu, 12 Aug 2021 22:50:49 +0000 (00:50 +0200)]
bcm4908: add missing config symbol

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agomediatek: add missing config symbols
David Bauer [Thu, 12 Aug 2021 21:40:32 +0000 (23:40 +0200)]
mediatek: add missing config symbols

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agotegra: add missing Kconfig symbol
David Bauer [Thu, 12 Aug 2021 21:22:08 +0000 (23:22 +0200)]
tegra: add missing Kconfig symbol

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agorockchip: add missing Kconfig symbols
David Bauer [Thu, 12 Aug 2021 20:19:45 +0000 (22:19 +0200)]
rockchip: add missing Kconfig symbols

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years ago6in4: delete tunnel on interface teardown
Ansuel Smith [Mon, 2 Aug 2021 11:41:21 +0000 (13:41 +0200)]
6in4: delete tunnel on interface teardown

Delete tunnel on 6in4 interface teardown.
Should solve problem related to tunnel stuck on restart loop
with "Unknown Command" on tunnel restart due to wan connection drop.
Fixes: FS#3690
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
3 years agohostapd: fix broken check in radar detection notification
Felix Fietkau [Wed, 11 Aug 2021 17:00:22 +0000 (19:00 +0200)]
hostapd: fix broken check in radar detection notification

This check was accidentally left in after reworking the code,
causing a segfault

Signed-off-by: Felix Fietkau <nbd@nbd.name>
3 years agogeneric: add missing config symbol
David Bauer [Wed, 11 Aug 2021 00:11:17 +0000 (02:11 +0200)]
generic: add missing config symbol

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agomediatek: add missing config symbol
David Bauer [Wed, 11 Aug 2021 00:09:25 +0000 (02:09 +0200)]
mediatek: add missing config symbol

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agotreewide: add various missing config symbols
David Bauer [Tue, 10 Aug 2021 18:22:18 +0000 (20:22 +0200)]
treewide: add various missing config symbols

Fixes commit 91a52f22a13d ("treewide: backport support for nvmem on non platform devices")

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agogeneric: Kconfig: exit on unset symbol
David Bauer [Thu, 22 Jul 2021 23:32:38 +0000 (01:32 +0200)]
generic: Kconfig: exit on unset symbol

When a target configuration has unser Kconfig symbols, the build will
fail when OpenWrt is compiled with V=s and stdin is connected to a tty.

In case OpenWrt is compiled without either of these preconditions, the
build will uscceed with the symbols in question being unset.

Modify the kernel configuration in a way it fails on unset symbols
regardless of the aformentioned preconditions.

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agogeneric: fix kernel panic on existing mac-address node
David Bauer [Mon, 9 Aug 2021 22:28:01 +0000 (00:28 +0200)]
generic: fix kernel panic on existing mac-address node

Calling free for the OF property can result in a kernel panic, as the
buffer in question might be referenced elsewhere. Also, it is not
removed from the tree.

Always allocate a new property and updating the tree with it fixes both
issues.

Fixes commit 91a52f22a13d ("treewide: backport support for nvmem on non platform devices")

Signed-off-by: David Bauer <mail@david-bauer.net>
3 years agofirmware-utils: nand_ecc: replace GPL boilerplate with SPDX
Rafał Miłecki [Fri, 6 Aug 2021 10:59:23 +0000 (12:59 +0200)]
firmware-utils: nand_ecc: replace GPL boilerplate with SPDX

Cc: yajin <yajin@vm-kernel.org>
Cc: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agofirmware-utils: trx: replace GPL-2.0-or-later boilerplate with SPDX
Rafał Miłecki [Fri, 6 Aug 2021 10:59:22 +0000 (12:59 +0200)]
firmware-utils: trx: replace GPL-2.0-or-later boilerplate with SPDX

This was missed because scancode license scanner was confused by
comments about crc32buf().

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agofirmware-utils: osbridge-crc: replace GPL-2.0-only boilerplate with SPDX
Rafał Miłecki [Fri, 6 Aug 2021 10:59:21 +0000 (12:59 +0200)]
firmware-utils: osbridge-crc: replace GPL-2.0-only boilerplate with SPDX

This was missed because scancode license scanner was confused by
comments about crc32buf().

Cc: Gabor Juhos <juhosg@openwrt.org>
Cc: Gabor Juhos <juhosg@freemail.hu>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agofirmware-utils: mkmylofw: replace GPL-2.0-or-later boilerplate with SPDX
Rafał Miłecki [Fri, 6 Aug 2021 10:59:20 +0000 (12:59 +0200)]
firmware-utils: mkmylofw: replace GPL-2.0-or-later boilerplate with SPDX

This was missed because scancode license scanner was confused by a
comment about (no) copyrights in the init_crc_table().

Cc: Gabor Juhos <juhosg@openwrt.org>
Cc: Gabor Juhos <juhosg@freemail.hu>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
3 years agofirmware-utils: mkmerakifw-old: replace GPL-2.0-only boilerplate with SPDX
Rafał Miłecki [Fri, 6 Aug 2021 10:59:19 +0000 (12:59 +0200)]
firmware-utils: mkmerakifw-old: replace GPL-2.0-only boilerplate with SPDX

This was missed because scancode license scanner was confused by a
comment about Cisco's GPL code github repository.

Cc: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Acked-by: Christian Lamparter <chunkeey@gmail.com>
3 years agoomap: Remove EXT2 and EXT3 driver config
Hauke Mehrtens [Sun, 8 Aug 2021 23:13:05 +0000 (01:13 +0200)]
omap: Remove EXT2 and EXT3 driver config

The EXT4 driver also takes care of EXT2 and EXT3 file systems.
Activating the EXT2 driver kernel config options unlocked some other
ext2 driver related options which OpenWrt did not take care of.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agouboot-layerscape: fix dtc compilation on host gcc 10
Hauke Mehrtens [Sun, 8 Aug 2021 22:03:43 +0000 (00:03 +0200)]
uboot-layerscape: fix dtc compilation on host gcc 10

Backport a patch from upstream U-Boot to fix the compile with host GCC 10.

Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agolayerscape: Fix multiple bugs in of_get_mac_address() changes
Hauke Mehrtens [Sun, 8 Aug 2021 20:50:27 +0000 (22:50 +0200)]
layerscape: Fix multiple bugs in of_get_mac_address() changes

The change which backported the of_get_mac_address() change broke some
patches in the layerscape target so the patches did not apply any more.

This commit makes them apply again and also fixes some other problems
related to this change.

Fixes commit 91a52f22a13d ("treewide: backport support for nvmem on non platform devices")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agokernel: Apply change to of_get_mac_address() to ks8851 too
Hauke Mehrtens [Sun, 8 Aug 2021 20:38:12 +0000 (22:38 +0200)]
kernel: Apply change to of_get_mac_address() to ks8851 too

The code from ks8851.c was moved to ks8851_common.c, so it was not
backported. This broke the compile of the omap target which uses this
driver.

Fixes commit 91a52f22a13d ("treewide: backport support for nvmem on non platform devices")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
3 years agomvebu: armada-37xx: add patch to forbid cpufreq for 1.2 GHz
Josef Schlehofer [Tue, 27 Jul 2021 11:20:43 +0000 (13:20 +0200)]
mvebu: armada-37xx: add patch to forbid cpufreq for 1.2 GHz

This patch is backported from linux-arm-kernel [1] to improve situation, when
it was reported that 1.2 GHz variant is unstable with DFS.
It waits to be accepted upstream, however, it waits for Marvell people to respond.

[1] https://patchwork.kernel.org/project/linux-arm-kernel/patch/20210630225601.6372-1-kabel@kernel.org/

Fixes: 7b868fe04a89 ("Revert "mvebu: 5.4 fix DVFS caused random boot crashes"")
Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
3 years agoRevert "mvebu: 5.10 fix DVFS caused random boot crashes"
Josef Schlehofer [Tue, 27 Jul 2021 11:20:42 +0000 (13:20 +0200)]
Revert "mvebu: 5.10 fix DVFS caused random boot crashes"

Based on the discussion on the mailing list [1], the patch which was
reverted, it reverts only one patch without the subsequent ones.

This leads to the SoC scaling issue not using a CPU parent clock, but
it uses DDR clock. This is done for all variants, and it's wrong because
commits (hacks) that were using the DDR clock are no longer in the mainline kernel.

If someone has stability issues on 1.2 GHz, it should not affect all
routers (1 GHz, 800 MHz) and it should be rather consulted with guys, who are trying to
improve the situation in the kernel and not making the situation worse.

There are two solutions in cases of instability:
a) disable cpufreq
b) underclock it up to 1 GHz

This reverts commit 080a0b74e39d159eecf69c468debec42f28bf4d8.

[1] https://lists.openwrt.org/pipermail/openwrt-devel/2021-June/035702.html

Fixes: d37947681772 ("mvebu: armada-37xx: add patch to forbid cpufreq for 1.2 GHz")
CC: Pali Rohár <pali@kernel.org>
Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
3 years agopcre: update to 8.45
Rosen Penev [Tue, 13 Jul 2021 20:32:30 +0000 (13:32 -0700)]
pcre: update to 8.45

Switch to AUTORELEASE to avoid manual increments.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agonettle: update to 3.7.3
Rosen Penev [Tue, 13 Jul 2021 20:30:24 +0000 (13:30 -0700)]
nettle: update to 3.7.3

Switch to AUTORELEASE to avoid manual increments.

Refreshed patches.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agombedtls: update to 2.16.11
Rosen Penev [Tue, 13 Jul 2021 20:27:09 +0000 (13:27 -0700)]
mbedtls: update to 2.16.11

Switched to AUTORELEASE to avoid manual increments.

Release notes:
https://github.com/ARMmbed/mbedtls/releases/tag/v2.16.11

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agolibpcap: update to 1.10.1
Rosen Penev [Tue, 13 Jul 2021 20:24:54 +0000 (13:24 -0700)]
libpcap: update to 1.10.1

Switch to AUTORELEASE to avoid manual increments.

Refreshed patches.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agolibnftnl: update to 1.2.0
Rosen Penev [Tue, 13 Jul 2021 20:19:08 +0000 (13:19 -0700)]
libnftnl: update to 1.2.0

Switch to AUTORELEASE to avoid manual increments.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agolibcap: update to 2.51
Rosen Penev [Tue, 13 Jul 2021 20:16:50 +0000 (13:16 -0700)]
libcap: update to 2.51

Switched to AUTORELEASE to avoid manual increments.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agoelfutils: update to 0.182
Rosen Penev [Sun, 3 Jan 2021 23:18:33 +0000 (15:18 -0800)]
elfutils: update to 0.182

Add --disable-libdebuginfod with remove libcurl dependency.

Remove totally unused host elfutils.

Refreshed and rebased patches.

Also happens to fix compilation with GCC11.

Newer versions of elfutils seem to have some kind of dependency on
obstack.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agogettext-full: disable parallel compilation
Rosen Penev [Mon, 12 Jul 2021 01:00:55 +0000 (18:00 -0700)]
gettext-full: disable parallel compilation

Fails fairly reliably with make -j 12 on a Ryzen 3600.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agoutil-linux: update to 2.37
Rosen Penev [Tue, 13 Jul 2021 20:49:14 +0000 (13:49 -0700)]
util-linux: update to 2.37

Switched to AUTORELEASE to avoid manual increments.

Changed PKG_LICENSE to SPDX format.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
3 years agohostapd: respect fixed channel BW in HE20 mode
Jesus Fernandez Manzano [Tue, 20 Jul 2021 10:13:49 +0000 (12:13 +0200)]
hostapd: respect fixed channel BW in HE20 mode

When using htmode 'HE20' with a radio mode that uses wpa-supplicant
(like mesh or sta), it will default to 40 MHz bw if disable_ht40 is not
set. This commit fixes this behaviour.

Signed-off-by: Jesus Fernandez Manzano <jesus.manzano@galgus.net>
3 years agolibpcap: add rpcapd as package
Stephan Schmidtmer [Sun, 25 Jul 2021 11:50:53 +0000 (11:50 +0000)]
libpcap: add rpcapd as package

This enables building of rpcapd and adds it as a package.

It is a daemon that allows remote packet capturing from another machine.
E.g. Wireshark can talk to it using the Remote Capture Protocol (RPCAP).
https://www.tcpdump.org/manpages/rpcapd.8.html

Compile and run tested: OpenWrt SNAPSHOT r17190-2801fe6132 on x86/64

Signed-off-by: Stephan Schmidtmer <hurz@gmx.org>
3 years agofritz-tools: fix returning wrong values due to strncmp usage
Daniel Kestrel [Sun, 25 Jul 2021 18:32:05 +0000 (20:32 +0200)]
fritz-tools: fix returning wrong values due to strncmp usage

When having two keys that start with the same characters and the second
key just has one character more nand_tffs_read and tffs_read return the
wrong value for the longer key. This is due to the usage of strncmp in
combination with the length of the shorter key which is usually first in
the list before the longer key and when strncmp matches, the search is
stopped. The problem only occurs when the length of the two keys is
different, not if just the last character is different. The fix is to
use strcmp and as such it will only return the value if the key (name)
and the key to look for (namefilter) have the same value and length. A
sample case returning wrong values is when keys macwlan and macwlan2 are
defined and querying macwlan2 returns the value for macwlan.

Signed-off-by: Daniel Kestrel <kestrel1974@t-online.de>
3 years agoiproute2: bump to 5.13
Ansuel Smith [Tue, 27 Jul 2021 18:13:40 +0000 (20:13 +0200)]
iproute2: bump to 5.13

Drop patch 185-libbpf-add-limits-h merged upstream
Update patch 170-ip_tiny
Update patch 130-no_netem_tpic_dcb...

Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
3 years agoiw: enable print wpa ie for scan
Jianhui Zhao [Mon, 2 Aug 2021 07:08:46 +0000 (07:08 +0000)]
iw: enable print wpa ie for scan

This increases the size of the iw_5.9-8fab0c9e-3_mips_24kc.ipk from
41166 to 41942 bytes by 776 bytes.

Signed-off-by: Jianhui Zhao <zhaojh329@gmail.com>
3 years agotoolchain/gcc: bump gcc 11 to 11.2
Rui Salvaterra [Wed, 28 Jul 2021 11:33:09 +0000 (12:33 +0100)]
toolchain/gcc: bump gcc 11 to 11.2

Patches automatically refreshed.

Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
3 years agolayerscape: armv8_64b: do not disable CONFIG_IP_ADVANCED_ROUTER
Martin Schiller [Tue, 27 Jul 2021 07:10:44 +0000 (09:10 +0200)]
layerscape: armv8_64b: do not disable CONFIG_IP_ADVANCED_ROUTER

This option is needed e.g. to use strongswan for IPSec.

BTW: This was the only target where this option was disabled.

Signed-off-by: Martin Schiller <ms@dev.tdt.de>
3 years agogemini: Add hdparm setting
Linus Walleij [Fri, 23 Jul 2021 10:29:27 +0000 (12:29 +0200)]
gemini: Add hdparm setting

This uses "hdparm" (if present) to get the harddisk into low
power mode on NAS set-ups.

Cc: Adrian Schmutzler <mail@adrianschmutzler.de>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
3 years agomvebu: armada-37xx: add patch to forbid cpufreq for 1.2 GHz
Josef Schlehofer [Tue, 27 Jul 2021 11:22:31 +0000 (13:22 +0200)]
mvebu: armada-37xx: add patch to forbid cpufreq for 1.2 GHz

This patch is backported from linux-arm-kernel [1] to improve situation, when
it was reported that 1.2 GHz variant is unstable with DFS.
It waits to be accepted upstream, however, it waits for Marvell people to respond.

[1] https://patchwork.kernel.org/project/linux-arm-kernel/patch/20210630225601.6372-1-kabel@kernel.org/

Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
3 years agoRevert "mvebu: 5.4 fix DVFS caused random boot crashes"
Josef Schlehofer [Tue, 27 Jul 2021 11:22:30 +0000 (13:22 +0200)]
Revert "mvebu: 5.4 fix DVFS caused random boot crashes"

Based on the discussion on the mailing list [1], the patch which was
reverted, it reverts only one patch without the subsequent ones.

This leads to the SoC scaling issue not using a CPU parent clock, but
it uses DDR clock. This is done for all variants, and it's wrong because
commits (hacks) that were using the DDR clock are no longer in the mainline kernel.

If someone has stability issues on 1.2 GHz, it should not affect all
routers (1 GHz, 800 MHz) and it should be rather consulted with guys, who are trying to
improve the situation in the kernel and not making the situation worse.

There are two solutions in cases of instability:
a) disable cpufreq
b) underclock it up to 1 GHz

This reverts commit 080a0b74e39d159eecf69c468debec42f28bf4d8.

[1] https://lists.openwrt.org/pipermail/openwrt-devel/2021-June/035702.html

CC: Pali Rohár <pali@kernel.org>
Signed-off-by: Josef Schlehofer <pepe.schlehofer@gmail.com>
3 years agomvebu: Turris Omnia: use SFP module, if present
Klaus Kudielka [Sat, 27 Feb 2021 11:56:10 +0000 (12:56 +0100)]
mvebu: Turris Omnia: use SFP module, if present

Follow the recommendations stated in the Turris Omnia DTS for eth2:

"In case SFP module is present, U-Boot has to enable the sfp node above,
remove phy-handle property, and add managed = "in-band-status" property."

The boot script is written in a way, that it works for all U-Boot
versions deployed by the vendor so far (2015.10-rc2, 2019.07).

Signed-off-by: Klaus Kudielka <klaus.kudielka@gmail.com>
3 years agomvebu: backport Turris Omnia DTS changes to 5.4
Klaus Kudielka [Sat, 17 Jul 2021 17:32:57 +0000 (19:32 +0200)]
mvebu: backport Turris Omnia DTS changes to 5.4

Kernel 5.4 receives a reduced set, just to make the SFP cage work.
While we are at it, move the patches accepted upstream to the 0xx series.

Signed-off-by: Klaus Kudielka <klaus.kudielka@gmail.com>