<feed xmlns='http://www.w3.org/2005/Atom'>
<title>staging/stintel, branch master</title>
<subtitle>Staging tree of Stijn Tintel</subtitle>
<id>https://git.openwrt.org/openwrt/staging/stintel/atom?h=master</id>
<link rel='self' href='https://git.openwrt.org/openwrt/staging/stintel/atom?h=master'/>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/'/>
<updated>2026-08-18T08:32:12Z</updated>
<entry>
<title>kernel: fix FM25G01B/FM25G02B Quad I/O read dummy</title>
<updated>2026-08-18T08:32:12Z</updated>
<author>
<name>Aleksandr Mineev</name>
</author>
<published>2026-07-17T08:59:53Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=6b98618527a12c52c04d1574202cdb4743728b7b'/>
<id>urn:sha1:6b98618527a12c52c04d1574202cdb4743728b7b</id>
<content type='text'>
The FM25G01B/FM25G02B datasheets specify a single dummy byte for the
0xEB Quad I/O read-from-cache, but the generic read_cache_variants set
issues two. The extra dummy byte shifts the data phase and returns
corrupted data with no ECC error, breaking boot on NC-1812/KN-1812.
Use a dedicated read-from-cache variant set with ndummy=1 for 0xEB.

Both the FM25G{01,02}B support (d5a5c9eb2ee9) and this fix
(8211f2d74b35) were merged into the mtd tree (nand/next). Backport both
patches to backport-6.18 using the actual upstream commits, so they
apply to the OpenWrt kernel tree and upstream nand/next as is.

FM25G01B datasheet: https://www.fmsh.com/nvm/FM25G01B_ds_eng.pdf
FM25G02B datasheet: https://www.fmsh.com/nvm/FM25G02B_ds_eng.pdf

Signed-off-by: Aleksandr Mineev &lt;sanderrrs@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24007
Signed-off-by: Jonas Jelonek &lt;jelonek.jonas@gmail.com&gt;
</content>
</entry>
<entry>
<title>mediatek: filogic: add support for GL.iNet BE10000</title>
<updated>2026-08-18T00:01:37Z</updated>
<author>
<name>GLiNet Tech</name>
</author>
<published>2026-08-15T06:35:38Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=00f7f33d2f458484416e87a066ee2e171c1578aa'/>
<id>urn:sha1:00f7f33d2f458484416e87a066ee2e171c1578aa</id>
<content type='text'>
This patch adds support for GL.iNet BE10000(Slate 7 Pro)
The GL.iNet BE10000 is a portable device with Wi-Fi 7, featuring two
2.5G Ethernet ports, one of which is a LAN port
and the other a WAN port.

Hardware Specification:
 - SoC: MediaTek MT7987A (Quad-core ARM Cortex-A53 2.0 GHz)
 - RAM: 1024 MiB DDR4
 - Flash: 512 MiB SPI-NAND
 - WiFi: MediaTek MT7995AV (WiFi 7)
   - 2.4GHz: b/g/n/ax/be (2x2 MIMO)
   - 5GHz: a/n/ac/ax/be (3x3 MIMO)
   - 6GHz: ax/be (3x3 MIMO)
 - Ethernet:
   - 1x 10/100/1000/2500 Mbps WAN (Realtek RTL8221 PHY)
   - 1x 10/100/1000/2500 Mbps LAN (SoC internal PHY)
 - Buttons: Reset,Switch
 - UART: 115200 8n1 (VCC, RX, TX, GND)

Flash Layout:
 - 0x0~0x100000 : bl2
 - 0x100000~0x180000 : u-boot-env
 - 0x180000~0x580000 : Factory
 - 0x580000~0x880000 : fip
 - 0x880000~0x8c0000 : log
 - 0x8c0000~0x900000 : CFG
 - 0x900000~0x1e000000 : ubi

MAC Addresses:
 - Base MAC located at Factory partition offset 0x4000
 - gmac0 (WAN) : Base - 0 (Label MAC)
 - gmac1 (LAN) : Base - 1

Installation:
 1. You can directly upgrade using the
    openwrt-mediatek-filogic-glinet_gl-be10000-squashfs-sysupgrade.bin
    image on the official GL firmware upgrade page.
 2. Access the device and use the sysupgrade command to upgrade.
    Note: When upgrading from official firmware, use the -n parameter
    to preserve no configuration.

Signed-off-by: GLiNet Tech &lt;glrouter@gl-inet.com&gt;
Link: https://github.com/openwrt/openwrt/pull/22485
Signed-off-by: Hauke Mehrtens &lt;hauke@hauke-m.de&gt;
</content>
</entry>
<entry>
<title>wifi-scripts: fix default cipher selection for 11ad in iface.uc</title>
<updated>2026-08-17T23:30:39Z</updated>
<author>
<name>Lech Perczak</name>
</author>
<published>2026-08-10T22:53:43Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=00045786e600de8f73c5b521fb97ef1d8e4045e0'/>
<id>urn:sha1:00045786e600de8f73c5b521fb97ef1d8e4045e0</id>
<content type='text'>
hw_mode property is only present in device configuration section,
not the interface section, so testing for config.hw_mode in
iface.parse_encryption is wrong. Test for (dev_config.band == 60g)
instead, this will cause GCMP to be properly selected instead of
CCMP unless set explicitly, and while at that will cover both hostapd
and wpa_supplicant cases.

Signed-off-by: Lech Perczak &lt;lech.perczak@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24586
Signed-off-by: Hauke Mehrtens &lt;hauke@hauke-m.de&gt;
</content>
</entry>
<entry>
<title>wifi-scripts: set hw_mode = 'ad' for 60GHz radios in hostapd config</title>
<updated>2026-08-17T23:30:39Z</updated>
<author>
<name>Lech Perczak</name>
</author>
<published>2026-08-09T23:13:04Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=aac62cf2e5e0f7ced53364cf6b1804f49dbc508b'/>
<id>urn:sha1:aac62cf2e5e0f7ced53364cf6b1804f49dbc508b</id>
<content type='text'>
ucode version of hostapd handler, when setting hw_mode property based on
band, incorrectly matches hw_mode = "a" for 5, 6, and 60GHz, while 60GHz
radios need this set explicitly as "ad".
This causes broken frequency display in LuCI, and potentially other issues,
so add an explicit check for 60GHz band.

Signed-off-by: Lech Perczak &lt;lech.perczak@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24586
Signed-off-by: Hauke Mehrtens &lt;hauke@hauke-m.de&gt;
</content>
</entry>
<entry>
<title>wifi-scripts: add 60G band to mac80211.uc to support 802.11ad</title>
<updated>2026-08-17T23:30:38Z</updated>
<author>
<name>Lech Perczak</name>
</author>
<published>2026-08-06T14:44:22Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=63d47187136e1c6e2d4cae5210baa75cd15edb5e'/>
<id>urn:sha1:63d47187136e1c6e2d4cae5210baa75cd15edb5e</id>
<content type='text'>
/etc/config/wireless is missing 60G band for routers supporting it.
This support was enabled in OpenWrt v23 but somehow disappeared in v24.

Run-tested-on: TP-Link Talon AD7200 v2

[Originally authored by Konstantin Glukhov, replaced SoB, because of
lack of real email, reworded commit message]

Signed-off-by: Lech Perczak &lt;lech.perczak@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24586
Signed-off-by: Hauke Mehrtens &lt;hauke@hauke-m.de&gt;
</content>
</entry>
<entry>
<title>realtek: add rtl960x Ethernet (GMAC) driver</title>
<updated>2026-08-17T19:30:40Z</updated>
<author>
<name>Taiga Ogawa</name>
</author>
<published>2026-08-03T14:05:34Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=1a77cc5e12812e16c341cee44bd4b6d8c819aa06'/>
<id>urn:sha1:1a77cc5e12812e16c341cee44bd4b6d8c819aa06</id>
<content type='text'>
Add a driver for the RTL960X on-chip Gigabit MAC: a single combined-NAPI
RX/TX datapath over the descriptor rings. When paired with the RTL960X DSA
switch driver, it converts the rtl_otto TX trailer to descriptor port
steering and passes the RX descriptor source port through
METADATA_HW_PORT_MUX. The DSA core derives offload_fwd_mark from bridge
membership; this metadata path cannot preserve the per-frame trap reason.

Configure TX completion coalescing with the 28-packet trigger,
TX_PKT_TMR=7, and REG_INI_TMR_SEL=3. On the tested hardware this gives a
roughly 469-472 us low-rate tail and fixed-window-side behavior; the exact
arming event remains unresolved. Retain the 10 ms software timer only as a
failure backstop.

Align the DMA burst size, early TX, TX FIFO threshold, RX FIFO size, and
RX interrupt trigger with the vendor driver. Keep the shorter RX packet
timer because the vendor field value reproducibly added about 0.9 ms of
ingress latency. The resulting IO_CMD value is 0xd71f31f0.

Program the RX filter through ndo_set_rx_mode, and use separate vendor
flow-control assert and deassert thresholds. The RX path bounds accepted
lengths to RX_BUF_SIZE - RX_SHIFT because the hardware writes each frame
2 bytes into the buffer.

The GMAC register block mixes 8/16/32-bit registers (IMR is 16-bit and
shares a 32-bit word with the write-1-to-clear ISR), so use three
width-specific regmap-MMIO instances over its private resource. Keep the
default value endianness to preserve the target's non-swapping MMIO
semantics.

Assisted-by: Codex:gpt-5.6-sol
Signed-off-by: Taiga Ogawa &lt;zectaiga@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24737
Signed-off-by: Markus Stockhausen &lt;markus.stockhausen@gmx.de&gt;
</content>
</entry>
<entry>
<title>realtek: move rtl838x Ethernet driver under realtek/</title>
<updated>2026-08-17T19:30:40Z</updated>
<author>
<name>Taiga Ogawa</name>
</author>
<published>2026-08-14T18:21:16Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=e5f9c21db508fae9bb89650de3d754e1cbfeca34'/>
<id>urn:sha1:e5f9c21db508fae9bb89650de3d754e1cbfeca34</id>
<content type='text'>
Move the existing rtl838x Ethernet driver under
drivers/net/ethernet/realtek/ without changing its source.

Moving NET_RTL838X under if NET_VENDOR_REALTEK would otherwise hide it
on MACH_REALTEK_RTL targets, which do not depend on PCI or parport.
Extending NET_VENDOR_REALTEK for MACH_REALTEK_RTL is the only
non-mechanical change and preserves the pre-move selection.

Assisted-by: Codex:gpt-5.6-sol
Signed-off-by: Taiga Ogawa &lt;zectaiga@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24737
Signed-off-by: Markus Stockhausen &lt;markus.stockhausen@gmx.de&gt;
</content>
</entry>
<entry>
<title>realtek: fix SMP IRQ 7 storm by restoring CEVT_R4K</title>
<updated>2026-08-17T19:09:19Z</updated>
<author>
<name>Luiz Angelo Daros de Luca</name>
</author>
<published>2026-08-11T17:30:50Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=b5d3a813191b5429c20e789f2785906943c327a7'/>
<id>urn:sha1:b5d3a813191b5429c20e789f2785906943c327a7</id>
<content type='text'>
The transition to generic machine initialization dropped the legacy
Realtek setup code. Without it, disabling CEVT_R4K results in an
unhandled IRQ 7 (MIPS CP0 timer) storm on secondary VPEs, leading
to an "irq 7: nobody cared" kernel panic.

The performance impact is negligible. Once the Realtek timer
(timer@3100) is loaded, it seamlessly takes over the tick duties due
to its higher rating. The MIPS R4K timer goes completely dormant and
only triggers harmlessly when the 32-bit counter overflows (a couple
of times per minute), avoiding the panic without core modifications.

Fixes: 7cc31af7bdd4 ("realtek: convert to generic machine initialization")
Assisted-by: Gemini:gemini-3.1-pro
Signed-off-by: Luiz Angelo Daros de Luca &lt;luizluca@gmail.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24702
Signed-off-by: Markus Stockhausen &lt;markus.stockhausen@gmx.de&gt;
</content>
</entry>
<entry>
<title>tools: zip: fix command injection in -T</title>
<updated>2026-08-17T08:36:17Z</updated>
<author>
<name>Peter Polonec</name>
</author>
<published>2026-08-14T16:25:19Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=56c66549576307f4c88cee6d1096471c6d6953ae'/>
<id>urn:sha1:56c66549576307f4c88cee6d1096471c6d6953ae</id>
<content type='text'>
[backport from Debian zip 3.0-16]
https://sources.debian.org/data/main/z/zip/3.0-16/debian/patches/fix-command-injection.patch

Info-ZIP 3.0 builds a shell command with the archive name when -T tests
an archive. A crafted archive name can inject additional shell commands.

Backport Debian's fix. Quote the archive name before it is added to the
test command.

Link: https://sintonen.fi/advisories/infozip-test-option-command-injection.txt
Signed-off-by: Peter Polonec &lt;polonec.devel@pm.me&gt;
Link: https://github.com/openwrt/openwrt/pull/24730
Signed-off-by: Robert Marko &lt;robimarko@gmail.com&gt;
</content>
</entry>
<entry>
<title>ipq40xx: enable display feature on generic subtarget</title>
<updated>2026-08-16T10:33:20Z</updated>
<author>
<name>Matt Eaton</name>
</author>
<published>2026-08-16T09:30:20Z</published>
<link rel='alternate' type='text/html' href='https://git.openwrt.org/openwrt/staging/stintel/commit/?id=20d94d5a2b27a25ea92b1d59f36f685b25e6a98b'/>
<id>urn:sha1:20d94d5a2b27a25ea92b1d59f36f685b25e6a98b</id>
<content type='text'>
The Ubiquiti UniFi Travel Router lists kmod-drm-panel-mipi-dbi and
kmod-backlight-pwm in DEVICE_PACKAGES. Both resolve through packages
gated on @DISPLAY_SUPPORT, which is only selected when the target
advertises the 'display' feature. Without it those packages are never
built, and the image step fails with:

  ERROR: unable to select packages:
    kmod-backlight-pwm (no such package)
    kmod-drm-panel-mipi-dbi (no such package)

This has broken every ipq40xx/generic snapshot build since r35795.

Signed-off-by: Matt Eaton &lt;git@divinehawk.com&gt;
Link: https://github.com/openwrt/openwrt/pull/24757
Signed-off-by: Robert Marko &lt;robimarko@gmail.com&gt;
</content>
</entry>
</feed>
