Robert Marko [Mon, 12 May 2025 18:14:52 +0000 (20:14 +0200)]
qualcommax: fix PCIe card wifi node address
Unlike other buses, for PCIe child nodes we should not use "wifi@1,0" as
the bus adress since for PCI devices the actual device adress on the bus is
actually set via the "reg" property and the node name is unused completely.
Though, it will cause DTC to throw warnings, so after investigation and
similar issue upstream[1] simply set the nodes to "wifi@0,0" instead.
Robert Marko [Tue, 13 May 2025 10:12:39 +0000 (12:12 +0200)]
qualcommax: ipq50xx: add missing bus-range to PCIe bridges
It seems that bus-range property was missed for the PCIe bridges on
IPQ50xx, thus causing DTC to warn on any bus adress used on nodes
under the PCIe bridges.
Nick Hainke [Sun, 11 May 2025 19:06:50 +0000 (21:06 +0200)]
libtracefs: update to 1.8.2
Changelog: 0379096 libtracefs: version 1.8.2 e2e0b4d sqlhist: Add bash completion for the sqlhist utility 78dd6d5 libtracefs: Add trace_sql.bash for tracefs_sql() bash completions 4f66bca libtracefs: Fix trace_sql() to handle buckets cast 6680dae libtracefs: Have tracefs_sql() cast handle stacktrace 4668195 libtracefs: Handle synthetic events with dynamic strings as fields 48a7752 libtracefs: Fix COMM for use as a field for tracefs_sqlhist() a8d57d2 libtracefs: Fix mistaken update to TRACEFS_STACKTRACE macro 6e913f7 libtracefs: utest: Do not fail if uprobes are not supported by kernel 5db5cf5 libtracefs: Rerun bison and flex via make sqlhist_remake 1514530 libtracefs: Add make sqlhist_remake to run bison and flex d49b6fa libtrace: Fix memory leak in tracefs_cpu 07cff64 libtracefs: Fix the read file failure code checking 457f02f trace-cmd: Unit test for tracefs_instance_file_append() API a2bfb49 libtracefs: utest: Rename private functions to fix static building 65f629e libtracefs utest: Do not test more events than what the ring buffer can hold 05a5dc4 libtracefs utest: Fix min percent test 0a46992 libtracefs utest: Add test to check handling multiple dynamic events bd47435 libtracefs: Destroy synthetic and eprobes before other events 78d8d2e libtracefs: Have tracefs_dynevent_get_all() find kprobes and uprobes properly ef70c00 libtracefs .gitignore: Ignore utest/trace-utest in git ff20336 libtracefs utest: Do not test kprobe interface if there's no kprobes 93d9049 libtracefs: Have tracefs_instance_tracers() return instance tracers 4cbebed libtracefs: Documentation: Add missing documentation to meson.build
Benjamin Berg [Wed, 30 Apr 2025 19:15:18 +0000 (21:15 +0200)]
mac80211: add patch to suppress PREP when mesh forwarding is disabled
This fixes a bug where mac80211 would respond to a PREQ frame when a
neighbor table entry exists locally even though it will not forward the
frame and the reported path does not actually work.
We are lucky to have a working realtek environment. But some things where mixed
heavily. To say it clear a bus is a bus and an ethernet is an ethernet. With
the new naming conventions and defines this becomes even more obvious.
Decouple it by moving the bus specific parts out of the ethernet device. To
make the code more readable rename bus_priv variables to priv and sort variable
definitions in inverse tree order (length descending) where it makes sense.
At least since 2022 there is a major bug in the MDIO driver that
produces out-of-bound reads and erratic behaviour during initialization.
- mdiobus_scan_bus_c22() scans the bus for 64 devices (PHY_MAX_ADDR)
- private bus structure only supports 57 entry arrays (MAX_PORTS)
All the bus/reader writer functions accept calls with addr>=57 and will
silently read beyond their limits. This can lead to ghost SERDES like
https://github.com/openwrt/openwrt/issues/18665#issuecomment-2846053813
Add proper boundary checks and end the functions with -ENODEV that is
the only accepted error code from the bus scan function.
realtek: move private bus structure closer to the bus
Relocate the bus structure definition into the MDIO source code area
of the ethernet driver. So if the real bus driver is forked from the
rest of the code only one area needs to be removed. Rename it to make
clear it belongs to the bus.
Although a dfine is used to set the maxiumum number of SMI
busses (=4) it is not used at all appropriate places in the code.
Replace hard coded constants with that define.
Inside the ethernet driver lives the mdio bus. It is not always clear
what belongs where. Prefix some leftovers from the kernel 6.6 refactor
to clearly state what belongs to the bus. Group all defines together
in one place. This commit has no functional changes.
qualcommbe: ipq95xx: fix ipq-uniphy crash on IPQ9554
IPQ9554 does not have uniphy1. The "gcc_uniphy1_sys_clk" cannot be
enabled. This causes the ipq-uniphy driver to crash on .probe().
Add a patch to resolve the crash.
Felix Fietkau [Sun, 11 May 2025 13:25:54 +0000 (15:25 +0200)]
ucode: update to Git HEAD (2025-05-11)
567207a9bf16 fs: implement fileno() on directory handles eb1d93235509 fs: support passing directory handle or fd in chdir() 38a2254337f1 build: detect whether toolchain employs default source fortification 6eddfc9dff17 resolv: fix fd leak in send_queries 3d36856b2dc5 uci: fix memory leak on cursor() error aafde95f1ecf uci: add cursor() flags argument d8cebc5a6bfd Revert "WIP: lib: support map() over objects" 2599cf80736b zlib: incorporate latest PR changes 830f316a7e49 socker: let sock.peercred() clear error on success 4cbac141406e types: rename u64_to_constant flag to ext_flag d802fe5da5cd types: add support for resources with embedded data/values 71b4fdc6f60b types: add support for setting resource persistent flag 141f799eba08 uloop: use container_of instead of direct pointer casts 1396f8f2988d uloop: use uc_fn_thisval 9a121fc7440c uloop: drop object_registry 11b804d97086 socket: do not clear last_error in socket.error() d5b3a9dc1091 socket: add strerror() method
Fixes: https://github.com/jow-/ucode/issues/285 Signed-off-by: Felix Fietkau <nbd@nbd.name>
This commit corrects led phy names. This has been discussed
at link [1] and corrected in the mediatek target in the
commit a51fd20e0d65ef836ee6c1e13c39a529855af4a0. Now it's time
to do the same thing in the lantiq target.
David Bauer [Sun, 4 May 2025 02:45:03 +0000 (04:45 +0200)]
ramips: support Genexis EX400 touch controller
Add the necessary package dependencies as well as device-tree properties
to support the touch-inputs as well as missing LEDs on the Genexis Pulse
EX400 range extender.
David Bauer [Sun, 4 May 2025 02:47:32 +0000 (04:47 +0200)]
ramips: add Semtech SX9512 touch input driver
This driver is required for the touch-inputs as well as some LEDs on the
Genexis Pulse EX400.
The driver was also sent upstream. As the EX400 is currently the only
consumer, the driver is added target-specific for ramips.
Once the driver has been accepted upstream and is provided in a kernel
release used by OpenWrt, the package should be moved to the global input
drivers.
cmake: use RelWithDebInfo, not Debug, with CONFIG_DEBUG
Given the description of the configuration option at its definition in
target/sdk/files/Config.in ("Compile packages with debugging info" and "Adds -g3
to the CFLAGS", the more appropriate CMAKE build type is `RelWithDebInfo` rather
than `Debug`.
INAGAKI Hiroshi [Sat, 10 May 2025 07:33:04 +0000 (16:33 +0900)]
qualcommax: ipq50xx: fix NAND ECC for ELECOM WRC-X3000GS2
Specify 4-bit ECC strength instead of 8-bit, to resolve NAND I/O errors
on ELECOM WRC-X3000GS2.
ELECOM WRC-X3000GS2 has a Macronix MX35UF1G24AD SPI-NAND chip as a flash
storage. That chip is registered as a chip that has 128 bytes OOB and
requires 8-bit ECC, and described as the same in the Macronix's datasheet.
But on WRC-X3000GS2, 8-bit ECC breaks NAND I/O on mtdblock devices with
the following errors and 4-bit ECC needs to be used to resolve this issue.
current (ECC strength=8):
```
[ 1.401566] 0x000000480000-0x000000500000 : "0:appsblenv"
[ 1.407128] 1 fixed-partitions partitions found on MTD device 0:appsblenv
[ 1.410663] Creating 1 MTD partitions on "0:appsblenv":
[ 1.417359] 0x000000000000-0x000000040000 : "env-data"
[ 1.479865] u-boot-env-layout: probe of 79b0000.qpic-nand:flash@0:partitions:partition-0-appsblenv:partition@0:nvmem-layout failed with error -74
root@OpenWrt:~# strings /dev/mtdblock10
[ 55.693754] mtdblock: MTD device '0:appsblenv' is NAND, please consider using UBI block devices instead.
[ 55.694660] I/O error, dev mtdblock10, sector 0 op 0x0:(READ) flags 0x80700 phys_seg 4 prio class 2
[ 55.703147] I/O error, dev mtdblock10, sector 8 op 0x0:(READ) flags 0x80700 phys_seg 3 prio class 2
[ 55.711884] I/O error, dev mtdblock10, sector 16 op 0x0:(READ) flags 0x80700 phys_seg 2 prio class 2
[ 55.720778] I/O error, dev mtdblock10, sector 24 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 2
[ 55.730310] I/O error, dev mtdblock10, sector 0 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 2
[ 55.738607] Buffer I/O error on dev mtdblock10, logical block 0, async page read
```
after fixing (ECC strength=4):
```
[ 1.402301] 0x000000480000-0x000000500000 : "0:appsblenv"
[ 1.407909] 1 fixed-partitions partitions found on MTD device 0:appsblenv
[ 1.411370] Creating 1 MTD partitions on "0:appsblenv":
[ 1.418075] 0x000000000000-0x000000040000 : "env-data"
(no error)
root@OpenWrt:~# strings /dev/mtdblock10
[ 39.131008] mtdblock: MTD device '0:appsblenv' is NAND, please consider using UBI block devices instead.
bootcmd=bootipq
baudrate=115200
wifi2=7C:F8:54:01:BD:F3
CRASH_COUNT=0
...
```
Fixes: 01b72ce61e ("qualcommax: ipq50xx: remove ECC user config from board files") Signed-off-by: INAGAKI Hiroshi <musashino.open@gmail.com> Link: https://github.com/openwrt/openwrt/pull/18770 Signed-off-by: Robert Marko <robimarko@gmail.com>
kmod-drm-rp1-dsi, kmod-drm-rp1-dpi and kmod-drm-rp1-vec should depend on
kmod-drm-dma-helper and kmod-drm-vram-helper in order to fix dependency errors
when all kernel modules are selected.
Shiji Yang [Fri, 9 May 2025 14:12:11 +0000 (22:12 +0800)]
realtek rtl931x: mark subtarget as source-only
There are no supported devices on this sub-target. It can be
considered that it is still under development. Therefore,
there is no need to make the buildbot build it every day.
Shiji Yang [Thu, 8 May 2025 11:23:14 +0000 (19:23 +0800)]
ramips: rt288x: mark subtarget as source-only
Due to RAM/ROM size limitations, this sub-target has been unable
to generate usable images for a long time. Mark them as source-only
to save some resource for the buildbot.
Shiji Yang [Thu, 8 May 2025 11:22:48 +0000 (19:22 +0800)]
lantig: xway_legacy: mark subtarget as source-only
Due to RAM/ROM size limitations, this sub-target has been unable
to generate usable images for a long time. Mark them as source-only
to save some resource for the buildbot.
Shiji Yang [Thu, 8 May 2025 11:22:16 +0000 (19:22 +0800)]
lantig: ase: mark subtarget as source-only
Due to RAM/ROM size limitations, this sub-target has been unable
to generate usable images for a long time. Mark them as source-only
to save some resource for the buildbot.
Eric Fahlgren [Thu, 8 May 2025 22:28:52 +0000 (15:28 -0700)]
imagebuilder: exclude metadata for profiles that have no kernel
Device profiles that specify 'DEFAULT := n' are being included
in the imagebuilder metadata, specifically in .profiles.mk, even
though there is no kernel built for the device. This results in
'make info' showing the device as valid, but then 'make image
PROFILE=xxx' failing with 'No rule to make target xxx-kernel.bin ...'
We exclude these profiles from the imagebuilder, avoiding these
errors.
Fixes: https://github.com/openwrt/openwrt/issues/18410 Signed-off-by: Eric Fahlgren <ericfahlgren@gmail.com> Link: https://github.com/openwrt/openwrt/pull/18748 Signed-off-by: Robert Marko <robimarko@gmail.com>
The RTL8214FC currently uses generic PHY functions. That makes it look like a copper
device. Switching to/from fibre works fortunately but the autonegotiation handling
still works on MII_LPA (PHY register 5) as if a copper link is used. Fix that by
- advertising a superset of TP/FIBRE features
- using clause 37 functions when on fibre
Additionally enhance the code of the driver to assist further development.
- log the speed of the inserted module to detect wrongly inserted 10gbase-r modules
- order phy driver functions alphabetically (keep match/name on top)
- remove genphy_loopback as the kernel uses it if not provided
Remark! The driver internally uses PORT_MII for the TP port. Align with that and
report MII to ethtool instead of TP. Other drivers do the same and it can be
changed in the future if needed.
Daniel Golle [Fri, 9 May 2025 02:46:29 +0000 (03:46 +0100)]
kernel: modules: fs: add missing backslash
Commit 7db0748b7b ("kernel: modules: split package fs-fscache") modified
the KCONFIG section of KernelPackage/fs-netfs but missed to add a
backslash to escape a linebreak. Add it now.
mediatek: filogic: increase spi flash memory speed on ZyXEL EX5601
Reduces boot time by 3s on a freshly installed image. This device
has a large flash and the gain can be higher with more packages
installed. According to the datasheet, this is the maximum frequency
supported by the Micron and Macronix chips that are installed in
these devices. Tested on three units over a two month period.
Before:
$ dd if=/dev/mtd5 of=/dev/null bs=10M count=1 status=progress
1+0 records in
1+0 records out 10485760 bytes (10 MB, 10 MiB) copied, 1.51901 s, 6.9 MB/s
After:
$ dd if=/dev/mtd5 of=/dev/null bs=10M count=1 status=progress
1+0 records in
1+0 records out 10485760 bytes (10 MB, 10 MiB) copied, 0.899474 s, 11.7 MB/s
Shiji Yang [Thu, 1 May 2025 03:35:36 +0000 (11:35 +0800)]
kernel: modules: split package fs-fscache
The module kmod-fs-fscache was merged into kmod-fs-netfs since
6.12 LTS kernel[1]. Therefore, we have to split fs-cachefiles
into a separate package. This patch also removed some outdated
config symbols that have already been removed in 5.15 kernel:
Lech Perczak [Sun, 12 Nov 2023 22:55:39 +0000 (23:55 +0100)]
ath79: support Mikrotik RouterBOARD 2011-UiAS-2HnD
Mikrotik RouterBOARD 2011 series are AR9344-based wired routers, with
optional 2,4GHz 802.11n 2x2:2 Wi-Fi radio, with desktop or rack-mount
capability, featuring 5 gigabit Ethernet ports and 5 fast Ethernet
ports. Different variants of the board are supported using a single
image.
Bare-bones RouterBOARD 2011LS is supported using the same image.
Specification:
- Power: 10-28V input via 5,5/2,1 DC barrel jack, or terminal block on
some boards
- CPU: AR9344 at 600 to 750MHz (configurable)
- RAM: 128MB (64MB in early versions) DDR2
- Flash: 128MB (64MB in early versions) NAND flash
- Gigabit Ethernet: 5 ports via AR8327 switch (eth1 to eth5)
- Fast Ethernet: 5 ports via AR9344 built-in switch (eth6 to eth10)
- SFP: single port through AR8327 switch (SGMII at port 6)
- PoE in: passive input up to 28V at eth1
- PoE out: passive output up to 28V at eth10
- USB: optional micro-AB host port or type-A host port
- Console: optional Cisco-style RS232 console at the back of the device,
on budget devices available as 3.3V UART via testpoints
- LCD: 2" touchscreen-equipped LCD (unsupported)
Installation:
- perform TFTP boot of initramfs image as for netinstall procedure
typical for Mikrotik devices
- when initramfs image boots, use sysupgrade to perform final
installation.
If upgrading from ar71xx, clean installation should be performed, as the
devices now uses yafut to handle kernel image.
TODO:
- 64MB NAND variant using K9F1208U0C isn't yet supported, because it
needs disabling subpage write support for that chip. A proper solution
would be to fix that in kernel, but I don't have this variant to test
with currently.
- The same is true for GD9FU1G8F3A NAND, which isn't ONFI compliant and
needs definition in kernel for proper geometry.
- SFP port probes over I²C and gigabit link is possible, but currently
the drivers lack support of handling link status information to
userspace, including swconfig. Link will appear as always up.
Currently, link status logging can detect something, but status both
for ports 0 and 6 is logged simultaneously,
but "swconfig dev switch0 show" will always show link up on port 6.
- Not yet fully known connection to display and touch panel, but some
documentation is available (seemingly connected via SPI, according to
Mikrotik's open source code at [1])
qualcommax: ipq50xx: remove ECC user config from board files
With the spi-qpic-snand driver now supporting reading ECC requirements
from the NAND chip itself, there's no need to set those in the DTS
anymore avoiding issues with devices of the same revision using
different NAND types with varying ECC requirements.
spi: spi-qpic-snand: update driver with latest patches from upstream
Updating the spi-qpic-snand driver with the latest patches sent
upstream and remove custom patches.
These patches add error handling fixes, use of right read location in
read ops, removal of unused variable, ability to read user config and
nand chip requirements, and support for 8bits ECC strength.
Tested on: Gl.iNet GL-B3000 & Linksys MR5500, MX2000, MX5500, SPNMX56
Henry Tung [Sun, 4 May 2025 18:29:54 +0000 (11:29 -0700)]
mac80211: ath11k: fix broadcast failures during GTK rekeying
Revert key_cipher change to NONE when clearing key.
Setting to NONE triggers asserts in the firmware and causes broadcast
drops with ath11k firmware 2.9.0.1, so until a future blob
fixes this, revert this as a workaround.
Imported from
https://git.codelinaro.org/clo/qsdk/oss/system/feeds/wlan-open/-/blob/win.wlan_host_opensource.3.0.r24/patches/ath11k/350-ath11k-Revert-clear-the-keys-properly-when-DISABLE_K.patch
Linus Lüssing [Wed, 7 May 2025 00:50:11 +0000 (02:50 +0200)]
realtek: rtl838x: fix broadcast flooding with many multicast entries
When many multicast entries are installed broadcast flooding might
potentially stop working for several ports. This is because the layer
2 broadcast flood port mask index has the wrong offset. It should be
9 bits, matching the 2^9 = 512 indexes on rtl838x, not 12.
The wrong offset leads to L2_BC_FLD_PMSK being set to 504, not 511
((511 << 12) >> 9) & 511 = 504). So, as by default an unset PMSK
is set to all ports, the issue would only become noticeable once
many multicast entries are installed, causing the 504th entry to be set
to something other than all ports.
Fixing this by setting the offset to 9 bits, to correctly point to our
511th reserved entry for all ports.
Tested-on: ZyXEL GS1900-24HP v1
Fixes: 28e972b2ea2f ("realtek: Configure initial L2 learning setup") Signed-off-by: Linus Lüssing <linus.luessing@c0d3.blue> Link: https://github.com/openwrt/openwrt/pull/18733 Signed-off-by: Robert Marko <robimarko@gmail.com>
This additional board.d script creates the "/tmp/sysinfo/oem_name" file
and populates it with the proper OEM "supported_devices" metadata entry
to be used by the new fwtool.sh oem detection feature.
Daniel Golle [Wed, 7 May 2025 02:58:17 +0000 (03:58 +0100)]
generic: 6.12: update block NVMEM driver
Update block NVMEM driver based on backported OF partition support.
Note the different MMC card DT binding compared to the previous
downstream solution: Instead of having a 'partitions' subnode inside
a 'block' node, the 'partitions' node now resides directly under the
node representing the card.
In order to make references to the 'boot0' or 'boot1' hw partitions
you will have to define 'partitions-boot0' or 'partitions-boot1', and
move the NVMEM layout into a partition (you may, of course, use
'fixed-partitions' for that, and cover the whole device, if needed).
See also https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/mmc/mmc-card.yaml?h=v6.13
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
ramips: update drivers to be compatible with kernel 6.12
Update drivers to be compatible with kernel 6.12:
- change deprecate function strlcpy to strscpy
- fix kernel warning "no previous prototype", by static declaration
- add missing of.h headers
Shiji Yang [Thu, 1 May 2025 02:18:52 +0000 (10:18 +0800)]
generic: fix MIPS -mno-long-calls patchset for 6.12 kernel
It seems that we need to override all execmem alloc/free callbacks,
not just for module. This is the default patch behavior in the 6.6
kernel. Fix the warning when the module is loaded:
Fixes: a9c0f28951fd ("generic: 6.12: move MIPS reloc patch from pending to hack and rework")
Ref: https://lore.kernel.org/all/20240505160628.2323363-1-rppt@kernel.org/ Signed-off-by: Shiji Yang <yangshiji66@outlook.com> Tested-by: Tony Ambardar <itugrok@yahoo.com> Link: https://github.com/openwrt/openwrt/pull/18721 Signed-off-by: Robert Marko <robimarko@gmail.com>
Some DTS files have a qsgmii link mode for the CPU port. This does
not harm but it is wrong. The CPU port of the realtek switch is always
directly connected to the switch by some unknown wiring and should
therefore be described as internal. Align the wrongly defined DTS
files to the standard.
This patch was added on linux v6.13 in order to avoid simple is_visible
functions on hwmon drivers.
See https://github.com/torvalds/linux/commit/79bc0af904db647979c735563299c9b0d820e432
Upstream e2015942e90a couldn't be backported to 6.6 because the following
symbols were missing:
- disable_work
- disable_work_sync
- enable_work
See https://github.com/torvalds/linux/commit/e2015942e90a021151a5751776f35830ba063be7
Reorganize patch numbers now that < 6.12 patches are no longer needed.
The following patches still differ from upstream:
- e340bff27e63
phy_set_eee_broken symbol is missing in 6.12
https://github.com/torvalds/linux/commit/e340bff27e63
The following patches can't be backported to 6.12 due to missing symbols:
- 5e7a74b6a357
phy_disable_eee_mode symbol is missing in 6.12
https://github.com/torvalds/linux/commit/5e7a74b6a357
projects / thirdparty / openwrt.git / log
