Now that omnia-eeprom is marked nonshared building the cortex-a9 mvebu
subtarget will fail with:
ERROR: unable to select packages:
omnia-eeprom (no such package):
required by: world[omnia-eeprom]
This is because omnia-eeprom depends on TARGET_mvebu_cortexa9_DEVICE_cznic_turris-omnia
which will not be satisfied in buildbots since CONFIG_TARGET_ALL_PROFILES
and CONFIG_TARGET_PER_DEVICE_ROOTFS are set in which case
CONFIG_TARGET_mvebu_cortexa9_DEVICE_cznic_turris-omnia is not set.
So, lets simply depend on the mvebu/cortex-a9 subtarget.
Fixes: 371e7bef40 ("omnia-eeprom: Mark it nonshared")
Link: https://github.com/openwrt/openwrt/pull/17007
Signed-off-by: Robert Marko <robimarko@gmail.com>
Add ubihealthd to the nand-utils package, auto-create UCI config for
each UBI device and launch the daemon on boot.
The default time interval between scrubbing a random PED is 120 seconds
which means that a fully used 128 MiB flash chip gets scrubbed in about
a day and a half. The interval can be adjusted in UCI using the
'interval' option.
Suggested-by: Rodrigo Balerdi <lanchon@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://github.com/openwrt/openwrt/pull/16973
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Sync patch with upstream version and tag them.
Minor changes done to Pinctrl patch to support older kernel.
Patch automatically refreshed with make target/linux/refresh.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
DEFAULT_PACKAGES handling was moved in commit 40be892a02
("imagebuilder: move handling of DEFAULT_PACKAGES into shareable place")
to `include/default-packages.mk`, but they weren't removed from
ImageBuilder's Makefile, so lets remove it now.
Once removed, I've noticed, that it stopped working as there is
target.mk included later in that file, overriding the DEFAULT_PACKAGES
again, so moved it after this target.mk include.
Fixes: 40be892a02 ("imagebuilder: move handling of DEFAULT_PACKAGES into shareable place")
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Robert reported, that in firmware images generated by ASU, there is
`apk` package manager missing after the commit 44598c233d ("build:
remove broken dependency of metadata on toplevel .config variables").
That is happening, because apk got removed from `default_packages` list in
`profiles.json`, which is being generated by `json_overview_image_info` Make
target, which uses `scripts/json_overview_image_info.py` helper script,
which gets the information from `DEFAULT_PACKAGES` Make variable.
So lets fix it by providing `DEFAULT_PACKAGES` variable when its needed.
The reason why we didn't added those packages as a dependency to
base-files like any other packages, was to allow disabling them (in
order to save space).
Fixes: #16969
Fixes: openwrt/asu/issues/1084
Fixes: 44598c233d ("build: remove broken dependency of metadata on toplevel .config variables")
Reported-by: Robert Marko <robimarko@gmail.com>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Link: https://github.com/openwrt/openwrt/pull/16986
Signed-off-by: Robert Marko <robimarko@gmail.com>
It seems, that handling of DEFAULT_PACKAGES is needed in more places, so
lets move it into dedicated include file so it can be easily shared.
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Link: https://github.com/openwrt/openwrt/pull/16986
Signed-off-by: Robert Marko <robimarko@gmail.com>
Remove the remaining special handling of procd-ujail in a same way as
the rest of the packages was handled in the commit 4c65359af4 ("build:
fix including busybox, procd and apk/opkg in imagebuilder").
Fixes: 44598c233d ("build: remove broken dependency of metadata on toplevel .config variables")
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Link: https://github.com/openwrt/openwrt/pull/16986
Signed-off-by: Robert Marko <robimarko@gmail.com>
CONFIG_FB_INTEL is now visible on x86 since i915 driver is packaged as kmod
now and it stops compilation, so add it to the generic config.
Signed-off-by: Robert Marko <robimarko@gmail.com>
Escape special char for package description for APK mkpkg as the
description is passed as an args to mkpkg with --info option and can
easily escape from the "".
Currently escaped char `, $, ", \.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
The config_get_bool also works with on/off, yes/no, true/false.
Add 'main' section name. This will make it easier to change settings from uci.
Add a link to documentation.
Signed-off-by: Sergey Ponomarev <stokito@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/15579
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
These get dynamically set based on compiler version. Not relevant for
targets.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16770
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This symbol is no longer present.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16770
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Move settings CONFIG_INTEL_MEI_GSC_PROXY, CONFIG_INTEL_MEI_HDCP
and CONFIG_INTEL_MEI_PXP to target generic.
Signed-off-by: Mieczyslaw Nalewaj <namiltd@yahoo.com>
Link: https://github.com/openwrt/openwrt/pull/16971
Signed-off-by: Robert Marko <robimarko@gmail.com>
This tool was build in the phase 2 build, there the
TARGET_mvebu_cortexa9_DEVICE_cznic_turris-omnia dependecy was probably
not meat. Mark it as non shared to build it together with the target
where this option is set.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Link: https://github.com/openwrt/openwrt/pull/16975
Signed-off-by: Robert Marko <robimarko@gmail.com>
This tool was build in the phase 2 build, there the TARGET dependencies
are probably not meat. Mark it as non shared to build it together with
the targets where this option is set.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Link: https://github.com/openwrt/openwrt/pull/16975
Signed-off-by: Robert Marko <robimarko@gmail.com>
Disable drm-i915 module for target x86/geode.
Fixes: 77cfe8f ("x86: make i915 as a kmod with required firmware")
Signed-off-by: Mieczyslaw Nalewaj <namiltd@yahoo.com>
Link: https://github.com/openwrt/openwrt/pull/16977
Signed-off-by: Robert Marko <robimarko@gmail.com>
All ipq807x devices that were using the legacy 'mmc_do_upgrade' eMMC
sysupgrade code were ported to the replacement 'emmc_do_upgrade' code.
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16505
Signed-off-by: Robert Marko <robimarko@gmail.com>
Alphabetically sort devices in platform.sh
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16505
Signed-off-by: Robert Marko <robimarko@gmail.com>
Note that the old ad-hoc method did not explicitly align backup data
to 64 KiB boundaries.
Also note that the qnap 301w has a 'rootfs_data' partition in the eMMC
that is being ignored by fstools during boot, presumably due to a bug.
This is why the partition is also ignored in the sysupgrade code and
there is no definition of CI_DATAPART="rootfs_data".
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16505
Signed-off-by: Robert Marko <robimarko@gmail.com>
Note that the old ad-hoc method did not explicitly align backup data
to 64 KiB boundaries.
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16505
Signed-off-by: Robert Marko <robimarko@gmail.com>
Currently the build fails due to the following:
mv: cannot stat 'linux-mediatek_filogic/target-dir-7872e783/etc/apk/repositories': No such file or directory
as the changes done in the commit e031dab93e ("base-files: move apk
distfeeds into directory") forget to adapt image generation part.
While looking into this, I've realized, that we don't need this explicit
handling in the image generating code since the feeds are solely
configured by `base-files` and `apk` packages, so those should always
provide the correct feeds content, so lets simply drop this unnecessary
code.
Moving away /etc/opkg is done to prevent opkg from picking up the remote
feeds defined from base-files and only use the local feeds, but for apk
we explicitly pass --repositories-file which disables parsing of
/etc/apk/repositories and /etc/apk/repositories.d, so we do not need to
backup anything.
Fixes: #16981
Fixes: e031dab93e ("base-files: move apk distfeeds into directory")
Reported-by: Chen Minqiang <ptpt52@gmail.com>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
[jonas.gorski: add an explicit explanation where the original mv comes
from and why we don't need it for apk].
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
On aarch64 musl gcc 14.x compiler, trying compiling elfutils 0.192 with
lto option enabled will cause null-dereference error.
Example error message:
...
elf_compress.c: In function 'elf_compress':
elf_compress.c:675:26: error: potential null pointer dereference [-Werror=null-dereference]
675 | shdr->sh_flags |= SHF_COMPRESSED;
| ^
elf_compress_gnu.c: In function 'elf_compress_gnu':
elf_compress_gnu.c:127:25: error: potential null pointer dereference [-Werror=null-dereference]
127 | shdr->sh_size = new_size;
| ^ ^
...
This is a false postive warning but will abort compilation if gcc has
`-Werror` flag. This commit add a patch for this, see the bugzilla
report below.
This commit backports a series of patches to fix some errors.
Add patch:
- 007-add-libeu-symbols-to-libelf.patch
- 008-fix-autoconf-ENABLE_IMA_VERIFICATION.patch
- 009-fix-null-dereference-with-lto.patch
Link: https://sourceware.org/bugzilla/show_bug.cgi?id=32311
Signed-off-by: Ryan Keane <the.ra2.ifv@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16886
Signed-off-by: Robert Marko <robimarko@gmail.com>
Since the start of the Realtek target OpenWrt works with RTL83XX as the
target architecture. Upstream is using MACH_REALTEK_RTL instead. To
simplify further development align that.
Signed-off-by: Markus Stockhausen <markus.stockhausen@gmx.de>
Link: https://github.com/openwrt/openwrt/pull/16963
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Analog to how we handle custom feeds in opkg, add a customfeeds.list for
custom package feeds and mark it as a configuration file.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16940
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Analog to how we handle distfeeds with opkg, have the base feeds defined
in a file in repositories.d instead of the base /etc/apk/repositories.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16940
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Currently, libreadline only installs
```
/usr/lib/libhistory.so.8 -> libhistory.so.8.2
/usr/lib/libhistory.so.8.2
/usr/lib/libreadline.so.8 -> libreadline.so.8.2
/usr/lib/libreadline.so.8.2
```
But there is no `libreadline.so` or `libhistory.so` available.
So this happens:
```
root@OpenWRT:~# cat a.c
int main() {
}
root@OpenWRT:~# gcc a.c -lreadline
/usr/bin/ld: cannot find -lreadline: No such file or directory
collect2: error: ld returned 1 exit status
```
Unless, of course, one uses `-l:libreadline.so.8`... But that
doesn't help with binaries that try to dynamically open
`libreadline.so`. I have one of those here (the STklos Scheme
compiler -- I didn't make a PR for it because it's far from
being ready, but one issue is that it does use dlopen to use
readline...)
With the symlink, it works:
```
root@OpenWRT:~# ln -s /usr/lib/libreadline.so.8 /usr/lib/libreadline.so
root@OpenWRT:~#
root@OpenWRT:~# gcc a.c -lreadline
root@OpenWRT:~#
```
Another example: when trying to package rlwrap, the build failed
complaining it could not find readline (using `-lreadline`).
It would then be necessary to change rlwrap's `configure.ac`
(and also in all packages that use readline), but it seems
simpler to add the symlinks...
This PR changes the Makefile so it will include the links.
Signed-off-by: Jeronimo Pellegrini <j_p@aleph0.info>
Link: https://github.com/openwrt/openwrt/pull/16445
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
This commit fixes and closes#16313.
Switch the x86 kernel's timer to tickless operation which is
more power efficient since it is not woken up by periodic timer
interrupts when idle. Also add several other options for CPU
idle governors particularly the upstream default for tickless
kernels, CONFIG_CPU_IDLE_GOV_MENU. Without this commit, my AMD
Ryzen 7 5800U can only achieve a minimum core frequency of 1,384
MHz which is over 3x higher than the processor's minimum
frequency of 400 MHz which is accessible with this modification.
In addition to the lower clock rate, I have seen a concomitant
reduction in both idle temps and at-the-wall power consumption.
Summary:
* Idle CPU freqs dropped from 1,384 MHz to 400 Mhz.
* Idle power consumption dropped from 7 W avg to 5 W.
* Idle temps have dropped from 50C on avg to 43C.
There are other well known reasons to switch to a tickless
timer including: reduced interrupt overhead, better use of CPU
resources, and reduced latency to name a few.
Build system: x86/64
Build-tested: x86/64/AMD Cezanne
Run-tested: x86/64/AMD Cezanne
Signed-off-by: John Audia <therealgraysky@proton.me>
Link: https://github.com/openwrt/openwrt/pull/16317
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
The ppp package can support all features since Linux 4.7.0 kernel.
Therefore, most kernel version checks can pass unconditionally on
OpenWrt v18.06 and later version. This patch can reduce the size
of ppp package by approximately 2.5 KB.
Signed-off-by: Shiji Yang <yangshiji66@qq.com>
Link: https://github.com/openwrt/openwrt/pull/16695
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
These permissions are not needed. Systemd also mounts these file systems
without these permissions on other Linux distributions.
Dropping these permissions should make the system more secure.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Link: https://github.com/openwrt/openwrt/pull/16960
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Create the folder /run and /run/lock using symlinks. Other Linux
distributions also have these folders and some applications might already
depend on them. Just create symlinks pointing to the older folder.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Link: https://github.com/openwrt/openwrt/pull/16961
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Do not add the file /run/apk/db.lock to the root file system. The /run
folder should be on a tmpfs.
At runtime we should make /run point to a tmpfs. At build time we should
just remove the folder.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Link: https://github.com/openwrt/openwrt/pull/16961
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
i915 driver requires to load correct firmware to work on latest x86
GPU, it is more reasonable to make it as a kernel module, so that
initramfs is not required, and it can also save some space from the
kernel image comparing being a built-in driver
Signed-off-by: Joe Zheng <joe.zheng@intel.com>
Link: https://github.com/openwrt/openwrt/pull/16276
Signed-off-by: Robert Marko <robimarko@gmail.com>
Change the package name from intel-igpu-firmware-* to i915-firmware-*,
the prefix "intel-igpu" is misleading, i915 firmware is not only for
iGPU but also for dGPU now.
Remove the redundant "intel" as i915 is already well known.
More accurate file classification to handle following files correctly:
adlp_dmc.bin
mtl_huc.bin
mtl_huc_gsc.bin
mtl_gsc_1.bin
The pattern in regex is "([[:alnum:]]+)_([[:alnum:]]+)(_[\w-.]+)?\.bin",
where $1 is the platform, $2 is the firmware type (dmc, guc, huc, etc.),
and the optional $3 which is revision or other suffix.
Glob first to narrow down the target file set, and then split with "_"
to extract the firmware type (remove the ".bin" in case there is no $3)
Add package "i915-firmware" as a meta package to install all the i915
firmwares, it is a balance between simplicity and optimization.
* Installing all the available firmwares as a whole, can support all the
platforms, not only the current one but also the future ones. The
price to pay is the increased size.
* If we want to minimize the storage, we can customize to install the
necessary ones only, even for the target platform only (e.g. ADL) and
skip the others. The price to pay is the time to tune.
What I am going to do is:
* Let drm-i915 driver depend on i915-firmware-dmc, which is small and
can cover most of the old platforms
* Let the user select i915-firmware to install all the i915 firmwares as
a whole to cover the latest or future platforms
Signed-off-by: Joe Zheng <joe.zheng@intel.com>
Link: https://github.com/openwrt/openwrt/pull/16276
Signed-off-by: Robert Marko <robimarko@gmail.com>
Add a new utility, `omnia-eeprom`, which can be used to print / set
EEPROM fields on Turris Omnia.
One example when this utility might be useful is if the board
experiences random crashes due to newer versions of the DDR training
algorithm in newer U-Boot. The user can change the DDR speed from 1600K
to 1333H to solve these issues, with
```
omnia-eeprom set ddr_speed 1333H
```
Signed-off-by: Marek Behún <kabel@kernel.org>
Link: https://github.com/openwrt/openwrt/pull/16264
Signed-off-by: Robert Marko <robimarko@gmail.com>
In preparation for APK version bump, use /dev/null instead of /dev/zero
for --repositories-file to mute an error in recent APK files.
New APK version use modern istream logic that are more sensible to the
kind of file passed and /dev/null is required to correctly handle an
empty repository file.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Currently its not possible to generate apk package.adb package index if
the package has longer description field, which leads to following
failure:
(2352/2353) Installing zoneinfo-all (2024b-r1)
(2353/2353) Installing zstd (1.5.6-r1)
ERROR: System state may be inconsistent: failed to write database: No buffer space available
1 error; 2704 MiB in 2353 packages
The code to read/write installeddb does not really handle long
description well. Until the database is converted to apkv3 format,
truncate the apkv3 descriptions to allow existing code to work.
APKv3 index and packages still contain the original long description
unmodified, so no package rebuild will be needed.
Fixing the issue by backporting the single upstream fix as its not
possible to to upstep apk to the latest Git HEAD due to some
regressions, see commit 692052cdc0e7 ("Revert "apk: update to Git
417a93ceae540444fdbd3f76d1dadf0e15621fdc (2024-11-13)"") for more
details.
Fixes: #16929
References: https://gitlab.alpinelinux.org/alpine/apk-tools/-/issues/11038
Upstream-Status: Backport [417a93ceae]
Link: https://github.com/openwrt/openwrt/pull/16951
Signed-off-by: Petr Štetiar <ynezz@true.cz>
7330fa5 initd: mount /sys and /proc with MS_RELATIME
Fixes mounting /proc in unpriviledged user namespace.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
8dfead68c202 wifi: mt76: mt7915: hold dev->mutex while interacting with the thermal state
d508a6eb935d wifi: mt76: mt7996: use mac80211 .sta_state op
57019e663f57 wifi: mt76: add code for emulating hardware scanning
dc4c2bdf7c56 wifi: mt76: add support for allocating a phy without hw
8cd0263f92e1 wifi: mt76: rename struct mt76_vif to mt76_vif_link
99df84d62883 wifi: mt76: add vif link specific data structure
dcc6f158d759 wifi: mt76: mt7996: split link specific data from struct mt7996_vif
d388deab9e73 wifi: mt76: initialize more wcid fields mt76_wcid_init
d026be405c54 wifi: mt76: add chanctx functions for multi-channel phy support
0b05795ca81c wifi: mt76: remove dev->wcid_phy_mask
0b526090de95 wifi: mt76: add multi-radio support to a few core hw ops
aeedee5c0a2c wifi: mt76: add multi-radio support to tx scheduling
fc0ff17b53ff wifi: mt76: add multi-radio support to scanning code
f19cbcf83400 wifi: mt76: add multi-radio remain_on_channel functions
42429ae0eaf6 wifi: mt76: mt7996: use emulated hardware scan support
f9d593d4a6b4 wifi: mt76: mt7996: pass wcid to mt7996_mcu_sta_hdr_trans_tlv
807090b28661 wifi: mt76: mt7996: prepare mt7996_mcu_add_dev/bss_info for MLO support
252baa7bf477 wifi: mt76: mt7996: prepare mt7996_mcu_add_beacon for MLO support
9ee990050305 wifi: mt76: mt7996: prepare mt7996_mcu_set_tx for MLO support
bf12cc404334 wifi: mt76: mt7996: prepare mt7996_mcu_set_timing for MLO support
60bf2bef95dc wifi: mt76: connac: prepare mt76_connac_mcu_sta_basic_tlv for MLO support
1289737e12a8 wifi: mt76: mt7996: prepare mt7996_mcu_update_bss_color for MLO support
7c00df0e7e57 wifi: mt76: connac: rework connac helpers
484e3f289a40 wifi: mt76: mt7996: move all debugfs files to the primary phy
d258f4e3e1ca wifi: mt76: mt7996: switch to single multi-radio wiphy
c246fa545119 wifi: mt76: mt7996: fix monitor mode
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Explicitly enable help so that the build will fail when an appropriate
lua binary isn't found instead of silently disabling the feature.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Most apk meson args are shared between host and target builds, so lets
have a common variable for them.
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Commit b2a84c86e3 ("apk: host: make the help available") enabled the
host build for lua (5.1) as a dependency, but apk defaults to lua 5.3.
This caused it to ignore our lua, and only build the help if the host
system had lua 5.3 installed.
Fix this by explicitly setting the lua version to ours.
Fixes: b2a84c86e3 ("apk: host: make the help available")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
This reverts commit c6a3a11c07 as it was
reported, that it causes issues:
ERROR: failed to read repositories: /dev/zero: No buffer space available
ERROR: unable to select packages:
ca-bundle-20240203-r1:
error: uninstallable
arch: all
satisfies: world[ca-bundle]
Reported-by: xiao bo <peterwillcn@gmail.com>
References: c6a3a11c07 (commitcomment-149072959)
Signed-off-by: Petr Štetiar <ynezz@true.cz>
This reverts commit e1043a746a, that
attempts to nest partitions that overlap but are not nested. This
causes the 'ubi' partition to be truncated, making rootfs inaccessible
and bricking the device.
Also, had this commit worked, it would have renumbered MTD partitions
in a way that would have broken documented scripts for installation and
update of main and recovery OSes, making backups, return to stock, etc,
and broken user configurations that put the 'extra' partition to use.
Signed-off-by: Rodrigo Balerdi <lanchon@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/16944
Signed-off-by: Robert Marko <robimarko@gmail.com>
The GatoNetworks GDSP is a re-branded version of the R5000 5G Industrial
router from Yinghua Technologies.
The re-branded device comes with OpenWrt preinstalled, and an OpenWrt-based
U-Boot bootloader version. While the flash layout has been kept compatible
with the OpenWrt version found on the stock device (see [5]), the image format
changed, making a bootloader upgrade necessary.
Specifications:
SoC: Mediatek MT7981BA
RAM: 256MB
Flash: SPI-NOR 32 MiB (Winbond W25Q256)
WLAN: MT7976CN DBDC AX Wi-Fi
Switch: MT7531AE (4x LAN Gigabit ports, 1x WAN Gigabit port)
5G: Quectel RM520N modem
Watchdog: an external WDT connected to GPIO 6 is present and always running;
the built-in Mediatek watchdog is also present and effective, but
not used at the moment.
This porting has been tested only with 1x 5G modems installed (the device
supports up to two).
Installation:
Installation is possible via sysupgrade both in the stock device and
re-branded version. However, in the former case, updating the bootloader is
required.
OpenWrt-based U-Boot Bootloader installation
--------------------------------------------
The firmware flashed in the re-branded device at manifacturing time will
flash an OpenWrt-based U-Boot bootloader with some extra recovery features
(see [1]) at first boot.
To update the bootloader, you need to install the mtd-rw module and
insmod it:
insmod mtd-rw i_want_a_brick=1
Then update relevant flash partitions:
mtd erase u-boot-env
mtd erase BL2
mtd erase FIP
mtd write openwrt-mediatek-filogic-gatonetworks_gdsp-preloader.bin BL2
mtd write openwrt-mediatek-filogic-gatonetworks_gdsp-bl31-uboot.fip FIP
And reboot, making sure all previous commands ran succesfully.
If something goes wrong, you can recover your device via the mtk_uartboot
tool.
In my testing, it was possible to start the process even without (un)-plugging
the device, may be handy for remote recovery.
Installation from stock device and firmware
-------------------------------------------
To install vanilla OpenWrt in the stock device (R5000 5G Industrial router
from Yinghua Technologies) running the stock vendor firmware, you will need
to update your bootloader as described in previous section. Remember to use
-F (force upgrade) and -n (not keeping settings).
U-Boot Recovery
---------------
This procedure has been tested only with the OpenWrt-based U-boot bootloader.
Assign your system static IP address 192.168.1.1 and start a TFTP server. The
device will look for an initramfs image named
openwrt-mediatek-filogic-gatonetworks_gdsp-initramfs-kernel.bin
(so you may use openwrt/bin/targets/mediatek/filogic as root dir for your
TFTP server).
Power on the device while keeping the reset button pressed, until you see
a TFTP request from 192.168.1.10. Your environment will be restored to it's
default state.
MAC addresses assignment
------------------------
MAC addresses are assigned slightly differently than in stock firmware. In
particular, the 5 GHz Wi-Fi uses 2.4 GHZ MAC + 1, rather than reusing it with
LA bit set as done in stock firmware. This MAC address is allocated to the
device, so it can be used.
The 2.4 GHz Wi-Fi MAC address is the label MAC. LAN MAC is used to set the
special U-Boot environment ethaddr variable.
device MAC address U-Boot env variable factory partition offset
2.4 GHz Wi-Fi :84 wifi_mac 0x4
5.8 GHz Wi-Fi :85 not present not present
WAN :86 wan_mac 0x24
LAN :87 lan_mac 0x2A
Notes
-----
[1]: the OpenWrt-based U-Boot bootloader you will find installed in the
re-branded device is configured to request for the initramfs image via
TFTP for $gdsp_tftp_tries times before trying normal boot from NOR flash.
Setting this U-Boot environment variable to 0x0 will disable the feature,
which is not implemented in this patch.
[2]: the exposed UART port is connected to ttyS1; the ttyS0 console port is
not exposed.
[3]: the provided bootloader environment has no provision for operating on
BL2 and the FIP partitions. This is an intentional choice to make it
(slightly) more difficult to brick the device.
[4]: it seems GPIO 6 is used both for the "SYS" LED and external WDT.
[5] BL2 expects to find FIP payload at a fixed offset, so some constraints
apply.
Signed-off-by: Enrico Mioso <mrkiko.rs@gmail.com>
