scripts: feeds: Don’t hardcode IS_TTY

When building in environments that set IS_TTY, the feeds script does not
honor it and passes a hardcoded value to scan.mk, causing unwanted
control characters to appear in stdout.

This commit addresses the issue by checking IS_TTY and MAKE_TERMOUT
variables and uses their values if defined.

Closes #8039

Signed-off-by: Ernestas Kulik <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20743
Signed-off-by: Christian Marangi <[email protected]>
(cherry picked from commit fd7e43d429a370665795fb14c859c046c761e4e9)

tools/ccache: same behaviour for local and CI builds

If the environment variable `CI` is set, ccache will enable the CMake
option CCACHE_DEV_MODE by default. This leads to differing behaviour
between local and CI builds which takes quite some time to debug. 🤯

Achieve consistent behaviour between local builds and CI builds by
setting CCACHE_DEV_MODE. Set it to OFF, because CCACHE_DEV_MODE amongst
other settings like linker choice, enables -Werror, which will lead to
potential compilation failures when the host compiler is updated. Using
-Werror for host utils is not desirable, because the compiler version
used is not controlled by the OpenWrt build system and host utils should
compile successfully on an as wide range of host OSes as possible.

Reported-by: Roman Azarenko <[email protected]>
Signed-off-by: Andreas Gnau <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20290
(cherry picked from commit 6f32c657db55a2e34f659d81d56aa385aaa7d69d)
Link: https://github.com/openwrt/openwrt/pull/20335
Signed-off-by: Christian Marangi <[email protected]>

kernel: r8127: update to v11.015.00

Changelog: https://github.com/openwrt/rtl8127/compare/11.014.00...11.015.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 1cffb44b3c7fb15a94638858a2625d4376755d1a)

kernel: r8126: update to v10.016.00

Changelog: https://github.com/openwrt/rtl8126/compare/10.015.00...10.016.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 27a3e5c7b7608009dd3166bf0983eb40ffdec275)

kernel: r8125: update to v9.016.01

Changelog: https://github.com/openwrt/rtl8125/compare/9.016.00...9.016.01

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 2dfad04a822b00d1cb62fc50d0ff7501dd7f5eaa)

kernel: r8168: add missing flags

PKG_MAKE_FLAGS is required when compiling r8168-rss

Signed-off-by: Andy Chiang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20001
Signed-off-by: Robert Marko <[email protected]>
Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 5719d6506467285db424814433e7b6d8779ee9d2)

OpenWrt v23.05.6: revert to branch defaults

Signed-off-by: Hauke Mehrtens <[email protected]>

OpenWrt v23.05.6: adjust config defaults

Signed-off-by: Hauke Mehrtens <[email protected]>

ath79: add extended AR9344 reset sequence

According to datasheet, on AR9344 the switch and switch analog need to
be reset first before initiating a full reset.

Resetting these systems fixes spurious reset hangs on Atheros AR9344
SoCs.

Link: https://github.com/freifunk-gluon/gluon/issues/2904
Signed-off-by: David Bauer <[email protected]>
(cherry picked from commit 144af32b47be6181f01815762159dff863f4616b)

account for
torvalds/linux/commit/41ccb3a0eb57741eebd72e51e011ef098bb1493d
torvalds/linux/commit/5eddff6add4feebac625f256d2fe292935351ce3
[[email protected]: adjust patch so it applies without commits above]
Signed-off-by: Felix Baumann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19759
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Add additional net scheduler fix

This fixes a problem introduced in the net scheduler fixes recently
backported.

Fixes: 242bca458fe4 ("kernel: backport fixes missing in stable")
Link: https://github.com/openwrt/openwrt/pull/19747
Signed-off-by: Hauke Mehrtens <[email protected]>

hostapd: fix logging of configuration content

This commit is an adoption of the fix for #14049 to
apply it for owrt-23.05.

Signed-off-by: Christian Korber <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17718
Link: https://github.com/openwrt/openwrt/pull/19658
Signed-off-by: Robert Marko <[email protected]>

kernel: backport fixes missing in stable

The a stable commit depend son these additional backports.

Without these additional backports this commit causes kernel warnings:
commit e269f29e9395527bc00c213c6b15da04ebb35070
Author: Lion Ackermann <[email protected]>
Date:   Mon Jun 30 15:27:30 2025 +0200
    net/sched: Always pass notifications when child class becomes empty
    [ Upstream commit 103406b38c600fec1fe375a77b27d87e314aea09 ]

Link: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.189&id=e269f29e9395527bc00c213c6b15da04ebb35070
Link: https://github.com/openwrt/openwrt/pull/19695
Signed-off-by: Hauke Mehrtens <[email protected]>

toolchain: glibc: Update glibc 2.37 to recent HEAD

23021eda75 malloc: Use __get_nprocs on arena_get2 (BZ 30945)
9a1bdd7df7 S390: Do not clobber r7 in clone [BZ #31402]
2fadb0d9b4 support: use 64-bit time_t (bug 30111)
08ac41cf08 make ‘struct pthread’ a complete type
9da8174362 nptl: Unconditionally use a 32-byte rseq area
a11c8d521d Include sys/rseq.h in tst-rseq-disable.c
829b64598b linux: Use rseq area unconditionally in sched_getcpu (bug 31479)
6f28bfa4a0 LoongArch: Correct {__ieee754, _}_scalb -> {__ieee754, _}_scalbf
314b731dd8 AArch64: Improve SVE memcpy and memmove
d0da4fb615 Add HWCAP2_MOPS from Linux 6.5 to AArch64 bits/hwcap.h
4c3985b4c8 AArch64: Add support for MOPS memcpy/memmove/memset
e7f8117b60 AArch64: Cleanup ifuncs
455b1b2b7f AArch64: Cleanup emag memset
cb48c2603a AArch64: Add memset_zva64
3dcb8fde81 AArch64: Remove Falkor memcpy
882b7dc2ab aarch64: correct CFI in rawmemchr (bug 31113)
bbc290aad4 aarch64: fix check for SVE support in assembler
fb8f66c22d AArch64: Check kernel version for SVE ifuncs
6eab323a3c powerpc: Fix ld.so address determination for PCREL mode (bug 31640)
89ce64b269 iconv: ISO-2022-CN-EXT: fix out-of-bound writes when writing escape sequence (CVE-2024-2961)
ae014a15b4 nptl: Fix tst-cancel30 on kernels without ppoll_time64 support
f75c298e74 CVE-2024-33599: nscd: Stack-based buffer overflow in netgroup cache (bug 31677)
5eea50c440 CVE-2024-33600: nscd: Do not send missing not-found response in addgetnetgrentX (bug 31678)
a8070b3104 CVE-2024-33600: nscd: Avoid null pointer crashes after notfound response (bug 31678)
6e106dc214 CVE-2024-33601, CVE-2024-33602: nscd: netgroup: Use two buffers in addgetnetgrentX (bug 31680)
1aa819a693 elf: Also compile dl-misc.os with $(rtld-early-cflags)
bcd2bf36b4 sparc: Remove 64 bit check on sparc32 wordsize (BZ 27574)
2a41cb9c71 login: Check default sizes of structs utmp, utmpx, lastlog
36ef0d2f2f login: structs utmp, utmpx, lastlog _TIME_BITS independence (bug 30701)
90304900c3 nscd: Use time_t for return type of addgetnetgrentX
f4de99ea83 hurd: fix build of tst-system.c
3956fd48d3 Force DT_RPATH for --enable-hardcoded-path-in-tests
340ca2d514 s390x: Fix segfault in wcsncmp [BZ #31934]
182f4a8c75 Add AT_RSEQ_* from Linux 6.3 to elf.h
a858172cf9 nptl: fix potential merge of __rseq_* relro symbols
643709c6ee elf: Make dl-rseq-symbols Linux only
7f769dd008 Linux: Make __rseq_size useful for feature detection (bug 31965)
40a1e1798b resolv: Allow short error responses to match any query (bug 31890)
3da1b0064a resolv: Do not wait for non-existing second DNS response after error (bug 30081)
708f507e30 resolv: Track single-request fallback via _res._flags (bug 31476)
88b3554ed9 tests: replace system by xsystem
40fb943e12 linux: Update the mremap C implementation [BZ #31968]
aaed1cc9fc mremap: Update manual entry
590e58b652 Add mremap tests
aca9c7d5de resolv: Fix tst-resolv-short-response for older GCC (bug 32042)
6e642a47fa Fix name space violation in fortify wrappers (bug 32052)
f82e0922de x86: Fix bug in strchrnul-evex512 [BZ #32078]
6eb9420551 support: Add FAIL test failure helper
5784740194 stdio-common: Add test for vfscanf with matches longer than INT_MAX [BZ #27650]
bb6cfec3dd Make tst-ungetc use libsupport
810fa2488c ungetc: Fix uninitialized read when putting into unused streams [BZ #27821]
5259c8f052 ungetc: Fix backup buffer leak on program exit [BZ #27821]
5b853745fd posix: Use <support/check.h> facilities in tst-truncate and tst-truncate64
3d0acc5b20 nptl: Use <support/check.h> facilities in tst-setuid3
0141e7c794 libio: Attempt wide backup free only for non-legacy code
a4d9258090 elf: Change ldconfig auxcache magic number (bug 32231)
4b2b81deb4 nptl: initialize rseq area prior to registration
b2ab6ce0f5 nptl: initialize cpu_id_start prior to rseq registration
71fca07505 x86: Avoid integer truncation with large cache sizes (bug 32470)
d5ff2fa5af x86_64: Sort fpu/multiarch/Makefile
3607863050 x86_64: Add log2 with FMA
c0cae389fb x86_64: Add expm1 with FMA
f2947239b6 x86_64: Add log1p with FMA
ec7c01efee x86: Check the lower byte of EAX of CPUID leaf 2 [BZ #30643]
0c6f7cd550 elf: Fix slow tls access after dlopen [BZ #19924]
f0c2fcce5f x86: Only align destination to 1x VEC_SIZE in memset 4x loop
ec40bb7938 sysdeps/x86/Makefile: Split and sort tests
b0eef2f27f x86_64: Fix missing wcsncat function definition without multiarch (x86-64-v4)
7e84df6892 x86: Improve large memset perf with non-temporal stores [RHEL-29312]
7d1c950582 x86/string: Fixup alignment of main loop in str{n}cmp-evex [BZ #32212]
19d8959b4d elf: Avoid some free (NULL) calls in _dl_update_slotinfo
abe2bb2b47 elf: Support recursive use of dynamic TLS in interposed malloc
9083997843 misc: Add support for Linux uio.h RWF_NOAPPEND flag
a3d7865b09 Fix underallocation of abort_msg_s struct (CVE-2025-0395)
24d69e33fc stdlib: Test using setenv with updated environ [BZ #32588]
5288c29247 assert: Reformat Makefile.
b989519fe1 assert: Add test for CVE-2025-0395
7feb4c414e AArch64: Improve generic strlen
7d16ea8b42 AArch64: Optimize memset
06eafb057a AArch64: Remove zva_128 from memset
93e4b0796e math: Improve layout of expf data
c747695762 AArch64: Add SVE memset
b4a783aa30 AArch64: Use prefer_sve_ifuncs for SVE memset
032545ebd3 math: Improve layout of exp/exp10 data
7403ede2d7 elf: Ignore LD_LIBRARY_PATH and debug env var for setuid for static
5b1ea2a223 support: Use const char * argument in support_capture_subprogram_self_sgid
1e29775735 support: Add support_record_failure_barrier
e8e6c54353 elf: Test case for bug 32976 (CVE-2025-4802)
0f4e5f764b x86_64: Fix typo in ifunc-impl-list.c.
fba1c47cb5 elf: Fix subprocess status handling for tst-dlopen-sgid (bug 32987)
9643281f9c support: Don't fail on fchown when spawning sgid processes
245dffa9ed support: Pick group in support_capture_subprogram_self_sgid if UID == 0
ea179f9a37 Fix error reporting (false negatives) in SGID tests
58b768addb posix: Fix double-free after allocation failure in regcomp (bug 33185)

Link: https://github.com/openwrt/openwrt/pull/19552
Signed-off-by: Hauke Mehrtens <[email protected]>

wireless-regdb: Update to version 2025.07.10

6953f19 wireless-regdb: Update regulatory info for Indonesia (ID) for 2025
2e8214e wireless-regdb: Permit 320 MHz bandwidth in 6 GHz band for GB
a94f685 wireless-regdb: Update regulatory info for Egypt (EG) for 2024
7628ce2 wireless-regdb: Update regulatory rules for Brazil (BR) on 6GHz
4411b39 wireless-regdb: Update regulatory info for Vietnam (VN) for 2025
490f136 wireless-regdb: Update regulatory info for Estonia (EE) for 2024
c56c663 wireless-regdb: update regulatory rules for Paraguay (PY) on 6 GHz for 2025
5a8ced5 wireless-regdb: Update regulatory info for CEPT countries for 6GHz listed by WiFi Alliance
5fd8ee3 wireless-regdb: update regulatory rules for Bosnia and Herzegovina (BA) for 6 GHz
e05260a wireless-regdb: update regulatory database based on preceding changes

Link: https://github.com/openwrt/openwrt/pull/19474
(cherry picked from commit 7fe06b7f56d95703a0cfb4c0e54dc8a6731dc70c)
Link: https://github.com/openwrt/openwrt/pull/19539
Signed-off-by: Hauke Mehrtens <[email protected]>

build: fix CMake generator for non-Ninja builds

OpenWRT by default uses the Ninja generator, but some packages disable
Ninja and use the default Unix Makefiles generator. This generator can
be overridden in the user environment with `CMAKE_GENERATOR`. This patch
explicitly sets the correct generator when `PKG_USE_NINJA:=0`.

In particular, the `mt76` package uses the Makefiles generator.

Signed-off-by: Ivan Romanov <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16263
(cherry picked from commit 4646aa169986036772b9f75393c08508d20ddf8b)
Link: https://github.com/openwrt/openwrt/pull/19525
Signed-off-by: Hauke Mehrtens <[email protected]>

libxml2: update to 2.14.5

Release Notes:
    https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.7
    https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.8
    https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.14.3
    https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.14.4
    https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.14.5

Fixes:
CVE-2025-32415 CVE-2025-32414 CVE-2025-6170 CVE-2025-49794 CVE-2025-49795 CVE-2025-49796

Build system: x86/64
Build-tested: x86/64-glibc
Run-tested: x86/64-glibc

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19383
(cherry picked from commit c08c2d6eb39dd5e040d6a50fdb46471b268fad1a)
Signed-off-by: Andreas Gnau <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19485
Signed-off-by: Hauke Mehrtens <[email protected]>

prereq-build: limit python distutils check to <v3.12

v3.12 won't have it anymore.

Signed-off-by: Andre Heider <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16699
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit 8191c8980f04087eaf2921dc327fff219999bd55)
Link: https://github.com/openwrt/openwrt/pull/19513
Signed-off-by: Hauke Mehrtens <[email protected]>

prereq-build: add Python 3.12 support

This reverts commit 09b7e1955c92e0b05197174b7bd4aef2c4073a35.

Signed-off-by: Paul Spooren <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16699
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit bf059f7108e75fdec618a272ab8c52d4780aafed)
Signed-off-by: Andreas Gnau <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19513
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.189

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.189

Manually rebased patches:
  bcm27xx/patches-5.15/950-0600-xhci-quirks-add-link-TRB-quirk-for-VL805.patch[1]
  bcm27xx/patches-5.15/950-0606-usb-xhci-add-VLI_TRB_CACHE_BUG-quirk.patch[1]
  bcm27xx/patches-5.15/950-0717-usb-xhci-add-a-quirk-for-Superspeed-bulk-OUT-transfe.patch[1]
  bcm53xx/patches-5.15/180-usb-xhci-add-support-for-performing-fake-doorbell.patch[1]

All other patches are automatically refreshed.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.189&id=e262ff8d634cf6e0028fb7b3643cdd8b758513be

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19466
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.188

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.188

There are no patches that need to be refreshed.

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19466
Signed-off-by: Hauke Mehrtens <[email protected]>

mac80211: Update to version 6.1.145-1

This updates mac80211 to version 6.1.145-1. This code is based on Linux
6.1.145 and contains all fixes included in the upstream wireless
subsystem from that kernel version. This includes many bugfixes and also
some security fixes.

The removed patches were applied upstream.

Link: https://github.com/openwrt/openwrt/pull/16948
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.187

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.187

New Kconfig symbol:
  x86: enable MITIGATION_TSA[1]

All patches are automatically refreshed.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.187&id=f2b75f1368af22bb290f128e29bc64b619a9d54d

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19250
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.186

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.186

Manually rebased patch:
  realtek/patches-5.15/800-net-mdio-support-hardware-assisted-indirect-access.patch[1-2]

All other patches are automatically refreshed.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.186&id=694456462ed63a06adbb0b7f2396a2eb5cc153c0
[2] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.186&id=014ad9210373d2104f6ef10e6bb999a7a0a4c50e

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19250
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: r8125: update to v9.016.00

Changelog: https://github.com/openwrt/rtl8125/compare/9.015.00...9.016.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 60606fa980d3fb8cd9b4bba1da9a321f9e72bebb)

kernel: r8168: add RSS variant

Instead of enabling RSS support, let's introduce a variant and let users
choose between both variants since it can cause network issues.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit cb3fc1aef9c01f23d7f740bf0c5f4b432d6d0e79)

kernel: r8127: rss: enable ENABLE_MULTIPLE_TX_QUEUE

We need both ENABLE_RSS_SUPPORT and ENABLE_MULTIPLE_TX_QUEUE in order to
support multiple RX and TX queues.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 7615de6ef0830d3917de66aaf4d5df6d3ff45c61)

kernel: r8126: rss: enable ENABLE_MULTIPLE_TX_QUEUE

We need both ENABLE_RSS_SUPPORT and ENABLE_MULTIPLE_TX_QUEUE in order to
support multiple RX and TX queues.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit a3e51a395652ec6b8b827de396b56ebc576d81e8)

kernel: r8125: rss: enable ENABLE_MULTIPLE_TX_QUEUE

We need both ENABLE_RSS_SUPPORT and ENABLE_MULTIPLE_TX_QUEUE in order to
support multiple RX and TX queues.

Signed-off-by: Pavel Kubelun <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18509
[bump release, improve commit description and package changes]
Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit d127963a4676909d2f58eb54859925a25cdaa063)

kernel: r8125: disable ASPM

Disable ASPM support for this NIC, fixing strange behavior problems, such as
increased latency, strange uneven throughput, etc.
With this option disabled the NIC achieves stable performance.
Upsteam r8169 driver disables ASPM by default for this NIC.

Signed-off-by: Pavel Kubelun <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18509
[bump release, fix commit description, split 2nd tx queue part]
Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit f99b39fd0caf4ba5f1be06841b772557243c4546)

kernel: r8127: load module at boot time

This allows the network interface naming to be stable, free from any
possible interaction from external USB network devices that might
claim usb* interface names.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit b5680bd11339b2ed489e1deeaaa2a3432691490a)

kernel: r8126: load module at boot time

This allows the network interface naming to be stable, free from any
possible interaction from external USB network devices that might
claim usb* interface names.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 9bb151d0c635241ceaa29525668cd225a10494d0)

kernel: r8125: load module at boot time

This allows the network interface naming to be stable, free from any
possible interaction from external USB network devices that might
claim usb* interface names.

Signed-off-by: Pavel Kubelun <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18509
Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit ce9f539672d2d55cea97e47a0e33459179237781)

kernel: r8168: load module at boot time

This allows the network interface naming to be stable, free from any
possible interaction from external USB network devices that might
claim usb* interface names.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 617961403c27733de5e9c11b0b2e6ea44d2dcec4)

kernel: r8101: load module at boot time

This allows the network interface naming to be stable, free from any
possible interaction from external USB network devices that might
claim usb* interface names.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 7aeb837bf46acdd68b5ed373741be897bbb2067d)

package: add kmod-r8127 ethernet driver

r8127 is an out of tree driver provided by Realtek for RTL8127 devices.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 851ea69d8ebf4bdc80b29562123b96cd94387e9e)

openssl: update to 3.0.16

Changes between 3.0.15 and 3.0.16 [11 Feb 2025]

CVE-2024-13176[1] - Fixed timing side-channel in ECDSA signature
computation.

There is a timing signal of around 300 nanoseconds when the top word of
the inverted ECDSA nonce value is zero. This can happen with significant
probability only for some of the supported elliptic curves. In
particular the NIST P-521 curve is affected. To be able to measure this
leak, the attacker process must either be located in the same physical
computer or must have a very fast network connection with low latency.

CVE-2024-9143[2] - Fixed possible OOB memory access with invalid
low-level GF(2^m) elliptic curve parameters.

Use of the low-level GF(2^m) elliptic curve APIs with untrusted explicit
values for the field polynomial can lead to out-of-bounds memory reads
or writes. Applications working with "exotic" explicit binary (GF(2^m))
curve parameters, that make it possible to represent invalid field
polynomials with a zero constant term, via the above or similar APIs,
may terminate abruptly as a result of reading or writing outside of
array bounds. Remote code execution cannot easily be ruled out.

1. https://www.openssl.org/news/vulnerabilities.html#CVE-2024-13176
2. https://www.openssl.org/news/vulnerabilities.html#CVE-2024-9143

Build system: x86/64
Build-tested: bcm27xx/bcm2712
Run-tested: bcm27xx/bcm2712

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17947
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit b4e6fd7b76440076eeff3a0789d40acbb5363ecf)
(cherry picked from commit 3abbc154546042af5efe6d299e3e40133dfb3d7f)
Link: https://github.com/openwrt/openwrt/pull/19073
Signed-off-by: Hauke Mehrtens <[email protected]>

wolfssl: Update to version 5.7.6

This fixes multiple bugs and also minor security problems.

Changelog:
https://github.com/wolfSSL/wolfssl/releases/tag/v5.7.4-stable
https://github.com/wolfSSL/wolfssl/releases/tag/v5.7.6-stable

The package size increases:
525814 bin/packages/mips_24kc/base/libwolfssl5.7.2.e624513f-5.7.2-r1.apk
549408 bin/packages/mips_24kc/base/libwolfssl5.7.6.e624513f-5.7.6-r1.apk

Link: https://github.com/openwrt/openwrt/pull/17742
(cherry picked from commit 0a7e92c2442bc09eec3875aae21996d9a5430806)
(cherry picked from commit 6f7bbd03954e8aa79484b7c5b727dd814bfee78d)
Link: https://github.com/openwrt/openwrt/pull/19073
Signed-off-by: Hauke Mehrtens <[email protected]>

grub2: set -std=gnu17

Arch Linux updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks grub compilation.

This is fixed in more recent grub2 versions.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23

Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

jsonfilter: update to Git HEAD (2025-04-18)

8a86fb78235b build: set GNU17 for lemon compilation

Link: https://github.com/openwrt/openwrt/pull/18525
(cherry picked from commit 891c8244f2b346b00190d25bd6970cf755a4aaa6)
Link: https://github.com/openwrt/openwrt/pull/18581
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit e0069e7587bbcd47821691586bffe65ec7b49534)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

toolchain: gdb: set -std=gnu17

Arch Linux updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks gdb compilation.

This is fixed in more recent gdb versions.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23

Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

toolchain: binutils: fix compilation with GCC15

GCC15 has switched the C language default from GNU17 to GNU23[1] and this
causes builds to fail with:
In file included from mips-opc.c:29:
mips-opc.c: In function 'decode_mips_operand':
mips-formats.h:86:7: error: expected identifier or '(' before 'static_assert'
   86 |       static_assert[(1 << (SIZE)) == ARRAY_SIZE (MAP)]; \
      |       ^~~~~~~~~~~~~
mips-opc.c:214:15: note: in expansion of macro 'MAPPED_REG'
  214 |     case 'z': MAPPED_REG (0, 0, GP, reg_0_map);
      |               ^~~~~~~~~~

So, backport upstream fix for this[2] to fix compilation with GCC15.
Patch for 2.40 was manually refreshed as part of the S390 code does not
exist in 2.40 as it was added after it.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23
[2] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=8ebe62f3f0d27806b1bf69f301f5e188b4acd2b4

Fixes: #18678
Link: https://github.com/openwrt/openwrt/pull/18681
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit d3216173abfc47917807423fc7578406ea29db5b)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: gmp: fix compilation with GCC15

Fedora 42 updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks GMP compilation by failing to find
a working compiler test.

Its been fixed upstream [2][3], so backport the fix to fix GCC15 compilation.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23
[2] https://gmplib.org/repo/gmp/rev/8e7bb4ae7a18
[3] https://gmplib.org/repo/gmp/rev/d66d66d82dbb

Link: https://github.com/openwrt/openwrt/pull/18506
(cherry picked from commit 31800db91d43042813b7249a09fd61c356b39767)
Link: https://github.com/openwrt/openwrt/pull/18581
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit d9fccc8e1a246bfa10a62ffaf60328d7f65178ed)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools/b43-tools: update to latest version, fix C23 build

Update b43-tools to the latest upstream version to fix a
compilation error with C23:

  util.h:25:15: error: 'bool' cannot be defined via 'typedef'
     25 | typedef _Bool bool;
        |               ^~~~

Changelog:
c6fc53f replace custom bool typedef with <stdbool.h>
dadf30c fix format warning in compilation
2fe10ea b43-fwdump: Fix forwarding of arguments to disassembler

Link: https://github.com/openwrt/openwrt/pull/18708
Signed-off-by: Nick Hainke <[email protected]>
(cherry picked from commit f8da45e25d9efcabbad65f1300de1bca7718b134)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: elfutils: fix compilation with GCC15

Backport fix from upstream to fix compilation with host GCC 15.
https://sourceware.org/git/?p=elfutils.git;a=commitdiff_plain;h=7508696d107ca01b65ce8273c881462a8658f90f

Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: elfutils: set -std=gnu17

Fedora 42 updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks m4 compilation.

This looks like a gnulib issue, so until that is updated/fixed
lets simply set C language version back to GNU17.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23

Link: https://github.com/openwrt/openwrt/pull/18506
(cherry picked from commit f165dd2f78f7aaafa85f59f5c35264dd03e3a6ce)
Link: https://github.com/openwrt/openwrt/pull/18581
Signed-off-by: Robert Marko <[email protected]>
[Rework patch to make it apply]
(cherry picked from commit 1e51fd037d780725ca31cc92b39078468a383295)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: e2fsprogs: fix compilation with GCC15

Backport fix from upstream to fix compilation with host GCC 15.
https://github.com/tytso/e2fsprogs/commit/49fd04d77b3244c6c6990be41142168eef373aef

Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: m4: set -std=gnu17

Fedora 42 updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks m4 compilation.

Its been reported upstream [2], so until its fixed lets simply set C
language version back to GNU17.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23
[2] https://savannah.gnu.org/support/?111150

Link: https://github.com/openwrt/openwrt/pull/18505
(cherry picked from commit 8c141e53d720b14a1d8595e0f756f1619df6e763)
Link: https://github.com/openwrt/openwrt/pull/18581
Signed-off-by: Robert Marko <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

tools: cpio: fix compilation with GCC15

Fedora 42 updated to GCC15 which now defaults to GNU23 as the default
instead of GNU17[1], and this breaks cpio compilation.

Its been reported upstream [2], so import the patch attached to the bug
report that fixes compilation with GCC15.

[1] https://gcc.gnu.org/gcc-15/porting_to.html#c23
[2] https://savannah.gnu.org/bugs/?66297

Link: https://github.com/openwrt/openwrt/pull/18506
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit c5041094ba06b3cdfb9458f1cb5acb00a9cbacd3)
Link: https://github.com/openwrt/openwrt/pull/19065
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.185

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.185

All patches are automatically refreshed.

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19028
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 5.15 to 5.15.184

Changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.184

New Kconfig symbols:
- enable CONFIG_MITIGATION_ITS[1]

All patches are automatically refreshed.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.15.184&id=858073be8899e16cc469be9c1837d6c3a64887fa

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19028
Signed-off-by: Hauke Mehrtens <[email protected]>

image: only filter out images when ImageBuilder is used

Currently, we are filtering out images if DEFAULT:=n or BROKEN:=y are set,
so if you are building from scratch and want to build custom images that
are stripped down to fit, you must edit the image recipe or its just
filtered out.

So, to allow this behaviour when building from scratch as we can assume
that person doing that knows what they are attempting to do lets just limit
the filtering to ImageBuilder.

Fixes: f060615a78e5 ("image: respect DEFAULT and BROKEN when Default profile is selected")
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit d0d1f190683ad5923182140c40afb479907a1c2b)

image: respect DEFAULT and BROKEN when Default profile is selected

Currently, when you select the Default profile it does not honor DEFAULT:=n
nor BROKEN:=y in device profiles but rather just tries to build all of them.

This may work when building directly, but when using Image Builder it will
always fail since no kernel or anything else is present for devices that
have DEFAULT:=n or BROKEN:=Y set since those are skipped during build.

So, lets look for DEFAULT being set to "n" or BROKEN being set to "y" and
then remove clear _PROFILE_SET so they dont end up being marked for
installation.

Fixes: #18410
Link: https://github.com/openwrt/openwrt/pull/18814
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit f060615a78e5c5e86829b2e40c3f77e5cf7033bf)

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 <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18748
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit b88f3ebb3341281fa7e61dc4f039062e72c86b9e)

kernel: Update to version 5.15.183

All patches automatically rebased.

changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.183

Signed-off-by: Antony Kolitsos <[email protected]>

kernel: Update to version 5.15.182

Manually adapted the following patches:
   generic/backport-5.15/828-v6.4-0002-of-Update-of_device_get_modalias.patch
   generic/backport-5.15/828-v6.4-0004-of-Move-of_modalias-to-module.c.patch

changelog: https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.182

Signed-off-by: Antony Kolitsos <[email protected]>

tools: add cmake dependency to bzip2

It's using cmake.mk so it needs CMake to build.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18880
(cherry picked from commit 9deb8fad2fbe6fcdc7fbe5d2eb6d8fd78992f485)
Signed-off-by: Felix Baumann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18885
Signed-off-by: Robert Marko <[email protected]>

ca-certificates: Update to 20241223

Update ca-certificates to version 20241223

  * Update Mozilla certificate authority bundle to version 2.70.
    The following certificate authorities were added (+):
    + Telekom Security TLS ECC Root 2020
    + Telekom Security TLS RSA Root 2023
    + FIRMAPROFESIONAL CA ROOT-A WEB
    + TWCA CYBER Root CA
    + SecureSign Root CA12
    + SecureSign Root CA14
    + SecureSign Root CA15
    The following certificate authorities were removed (-):
    - Security Communication Root CA (closes: #1063093)

Signed-off-by: Hannu Nyman <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18468
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit c79572210fa47d015b3bdc03f38c25ef5262a318)
Link: https://github.com/openwrt/openwrt/pull/18872
Signed-off-by: Hauke Mehrtens <[email protected]>

ca-certificates: update to version 20240203

Update Mozilla certificate authority bundle to version 2.64

Signed-off-by: Seo Suchan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18872
Signed-off-by: Hauke Mehrtens <[email protected]>
(cherry picked from commit 6b904fa95b50c1259954b2bff31fafe83290f182)

generic: drop extra-old-deprecated pending fix patch for sch codel

Remove patch 620-net_sched-codel-do-not-defer-queue-length-update from openwrt-23.05 as well.

Patch is actually an ancient patch that somehow manage to be ported for 7 solid years.

This comes from [1] where a fix patch was proposed. Nobody notice that the proposed patch was actually rejected upstream in favor of [2]. And the upstream fix patch is present in kernel from version 4.18.

This means that we were actually fixing for a non existant bug and maybe introducing regression down the line.

Drop the patch for good as we already have a fix for it in flace for a long time.

[1] https://bugzilla.kernel.org/show_bug.cgi?id=109581
[2] torvalds/linux@35b42da

Automatic refresh:
target/linux/generic/hack-5.15/660-fq_codel_defaults.patch

Tested on ramips/mt7621

Signed-off-by: Antony Kolitsos <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18791
Signed-off-by: Robert Marko <[email protected]>

wireless-regdb: Update to version 2025.02.20

b43aeb5 wireless-regdb: assert and correct maximum bandwidth within frequency difference
68588bf wireless-regdb: Update regulatory info for Syria (SY) for 2020
0dda57e wireless-regdb: Update regulatory info for Moldova (MD) on 6GHz for 2022
b19ab0b wireless-regdb: Update regulatory info for Azerbaijan (AZ) on 6GHz for 2024
f67f40d wireless-regdb: Update regulatory info for Oman (OM)
bd70876 wireless-regdb: Update regulatory rules for Armenia (AM) on 2.4 and 5 GHz
6c7cbcc wireless-regdb: Permit 320 MHz bandwidth in 6 GHz band in ETSI/CEPT
f9f6b30 wireless-regdb: Update regulatory rules for Austria (AT)
39b47ea wireless-regdb: Update regulatory info for Cayman Islands (KY) for 2024
3dd7ceb wireless-regdb: allow NO-INDOOR flag in db.txt
4d754a1 wireless-regdb: Update regulatory rules for Iran (IR) on both 2.4 and 5Ghz for 2021
8c8308a wireless-regdb: Update frequency range with NO-INDOOR for Oman (OM)
c2f11e2 wireless-regdb: update regulatory database based on preceding changes

Signed-off-by: Rudy Andram <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17957
Signed-off-by: Hauke Mehrtens <[email protected]>
(cherry picked from commit da2cc98458f46745de95e88b6066620fbd02b190)

mbedtls: update to 2.28.10

This release of Mbed TLS provides bug fixes and minor enhancements. This release includes fixes for security issues.

Mbed TLS 2.28.10 is the last release of the 2.28 LTS and won't receive bug fixes or security fixes anymore.
Users are advised to upgrade to a maintained version.

* CVE-2025-27810: Potential authentication bypass in TLS handshake [1]
* CVE-2025-27809: TLS clients should generally call mbedtls_ssl_set_hostname [2]

[1] https://mbed-tls.readthedocs.io/en/latest/security-advisories/mbedtls-security-advisory-2025-03-2/
[2] https://mbed-tls.readthedocs.io/en/latest/security-advisories/mbedtls-security-advisory-2025-03-1/

Full release announcement:
https://github.com/Mbed-TLS/mbedtls/releases/tag/mbedtls-2.28.10

Signed-off-by: Antony Kolitsos <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18793
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.181

Manually adapted the following patches:
   bcm27xx/patches-5.15/950-0612-media-i2c-ov7251-Make-the-enable-GPIO-optional.patch

Link: https://github.com/openwrt/openwrt/pull/18710
Signed-off-by: Hauke Mehrtens <[email protected]>

tools/ninja: fix build with python 3.13+

python 3.13 removed the pipes module. Replace pipes by shlex. shlex was introduced into python in 3.3

Signed-off-by: Arne Zachlod <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18582
Signed-off-by: Robert Marko <[email protected]>

kernel: add missing symbol (BLOCK_LEGACY_AUTOLOAD)

As of kernel v5.15.179 kernel configuration fails without a
reference to BLOCK_LEGACY_AUTOLOAD, apparently as a consequence
of commit 3e9899c12d5ab3a8ce376bf4365175ea3dbaffc0 (backport of
6c0f5898836c05c6d850a750ed7940ba29e4e6c5), so add this symbol as
unset which is the same state the symbol has in the 6.6 kernel
config for 24.10.

Fixes: dfe1b5e82a35 ("kernel: Update to version 5.15.179")
Signed-off-by: Andrew MacIntyre <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18552
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.180

No manual changes needed.

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.179

Removed the following patches because they were upstreamed:
   bcm27xx/patches-5.15/950-0020-drm-probe-helper-Create-a-HPD-IRQ-event-helper-for-a.patch [0]
   bcm47xx/patches-5.15/701-bgmac-reduce-max-frame-size-to-support-just-MTU-1500.patch [1]
   bcm53xx/patches-5.15/700-bgmac-reduce-max-frame-size-to-support-just-MTU-1500.patch [1]

Manually adapted the following patches:
   bcm27xx/patches-5.15/950-0151-usb-add-plumbing-for-updating-interrupt-endpoint-int.patch
   bcm27xx/patches-5.15/950-0600-xhci-quirks-add-link-TRB-quirk-for-VL805.patch

[0]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-5.15.y&id=5211d75e2cb75cb129602247e79058f26afb4104
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-5.15.y&id=3ce32a9fb1d210f1e8384ebbfc336b64d3c131bd

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.178

No manual changes needed.

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.177

No manual changes needed.

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.176

No manual changes needed.

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.175

No manual changes needed.

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: Update to version 5.15.174

Manually adapted the following patches:
   bcm27xx/patches-5.15/950-0156-drm-v3d-Skip-MMU-flush-if-the-device-is-currently-of.patch
   bcm27xx/patches-5.15/950-0173-drm-v3d-Delete-pm_runtime-support.patch
   generic/hack-5.15/780-usb-net-MeigLink_modem_support.patch

Link: https://github.com/openwrt/openwrt/pull/18012
Signed-off-by: Hauke Mehrtens <[email protected]>

ramips: fix image recipe for ASUS RT-N56U

The OpenWrt image metadata includes checksum validation. Therefore,
it must be generated at the end.

Fixes: https://github.com/openwrt/openwrt/issues/9045
Fixes: https://github.com/openwrt/openwrt/issues/13674
Link: https://patchwork.ozlabs.org/project/openwrt/patch/TYCP286MB08952FAACDFA234C5E052131BCD82@TYCP286MB0895.JPNP286.PROD.OUTLOOK.COM/
Signed-off-by: Shiji Yang <[email protected]>
Signed-off-by: Luke Morrison <[email protected]>
(cherry picked from commit 708b77f549e852d5156acde03cc76942da29494a)
Link: https://github.com/openwrt/openwrt/pull/17678
Signed-off-by: Hauke Mehrtens <[email protected]>

uboot-mediatek: fix image overlaps SPL issue

Backport u-boot linker symbol fixes from u-boot v2024.01 to fix the
image overlaps SPL issue.

Fixes: https://github.com/openwrt/openwrt/issues/17120
Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17264
Signed-off-by: Hauke Mehrtens <[email protected]>

libxml2: update to 2.13.6

Release Notes:
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.5
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.6

Fixes: CVE-2025-24928 CVE-2024-56171
Link: https://github.com/openwrt/openwrt/pull/18194
Signed-off-by: Nick Hainke <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18280
Signed-off-by: Robert Marko <[email protected]>

libxml2: update to 2.13.4

Release Notes:
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.0
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.1
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.2
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.3
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.13.4

Small size reduction:
415095 bin/packages/mips_24kc-old/base/libxml2_2.12.6-r1_mips_24kc.ipk
 87175 bin/packages/mips_24kc-old/base/libxml2-dev_2.12.6-r1_mips_24kc.ipk
 20190 bin/packages/mips_24kc-old/base/libxml2-utils_2.12.6-r1_mips_24kc.ipk
398070 bin/packages/mips_24kc-new/base/libxml2_2.13.4-r1_mips_24kc.ipk
 86760 bin/packages/mips_24kc-new/base/libxml2-dev_2.13.4-r1_mips_24kc.ipk
 19479 bin/packages/mips_24kc-new/base/libxml2-utils_2.13.4-r1_mips_24kc.ipk

Link: https://github.com/openwrt/openwrt/pull/16593
Signed-off-by: Hauke Mehrtens <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18280
Signed-off-by: Robert Marko <[email protected]>

libxml2: update to 2.12.9

Release Notes:
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.12.7
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.12.8
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.12.9

This fixes:
CVE-2024-34459: Fix buffer overread with xmllint --htmlout
CVE-2024-40896: Fix XXE protection in downstream code

Link: https://github.com/openwrt/openwrt/pull/16593
Signed-off-by: Hauke Mehrtens <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18280
Signed-off-by: Robert Marko <[email protected]>

libxml2: update to 2.12.6

Release Notes:
https://gitlab.gnome.org/GNOME/libxml2/-/releases/v2.12.6

Changelog (taken from the release notes):
- Regressions
  - parser: Fix detection of duplicate attributes in XML namespace
  - xmlreader: Fix xmlTextReaderConstEncoding
  - html: Fix htmlCreatePushParserCtxt with encoding
  - xmllint: Return error code if XPath returns empty nodeset

Compile-tested: x86_x64, Q35 VM, OpenWRT SNAPSHOT r26135-a8bfdf2ed4d9
Run-tested: x86_x64, Q35 VM, OpenWRT SNAPSHOT r26135-a8bfdf2ed4d9
Signed-off-by: Pascal Ernster <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18280
Signed-off-by: Robert Marko <[email protected]>

mac80211: fix compilation error for old stable kernel version

Fix compilation error for old stable version caused by
genlmsg_multicast_allns backport fix pushed middle version.

Version 5.15 version 0-169, 6.1 version 0-115, 6.6 version 0-58 have the
old genlmsg_multicast_allns version with flags variable.

Compiling backport project with these version result in a compilation
error. To handle this, introduce a backport function for the affected
kernel version.

Link: https://github.com/openwrt/openwrt/pull/18373
Signed-off-by: Christian Marangi <[email protected]>
(cherry picked from commit 40393881491aa5fd23e715d0bedf591e3b152ddd)

lldpd: add `custom-tlv` handling

Do not verify the format of TLV. Leave that to lldpd.

These lldpd config entries:

config custom-tlv
list ports 'eth0'
option tlv 'replace oui 33,44,55 subtype 254 oui-info 55,55,55,55,55'

config custom-tlv
option tlv 'oui 33,44,44 subtype 232'
list ports 'br-lan'
list ports 'eth0'

config custom-tlv # oui-info truncated
option tlv 'add oui 33,44,33 subtype 66 oui-info 5555555555'

config custom-tlv
option tlv 'add oui 33,44,31 subtype 44'

config custom-tlv # invalid oui
option tlv 'add oui 3322 subtype 79'

config custom-tlv # invalid oui
option tlv 'oui 3312 subtype 74'

Produce the following lldpd.conf content:

configure ports eth0 lldp custom-tlv replace oui 33,44,55 subtype 254 oui-info 55,55,55,55,55
configure ports br-lan,eth0 lldp custom-tlv oui 33,44,44 subtype 232
configure lldp custom-tlv add oui 33,44,33 subtype 66 oui-info 5555555555
configure lldp custom-tlv add oui 33,44,31 subtype 44
configure lldp custom-tlv add oui 3322 subtype 79
configure lldp custom-tlv oui 3312 subtype 74

And lldpd (v1.0.13 on v22) logs the following:

Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op replace oui 33:44:55 subtype fe
Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op add oui 33:44:44 subtype e8
Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op add oui 33:44:33 subtype 42
Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op add oui 33:44:33 subtype 42
Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op add oui 33:44:31 subtype 2c
Sat Mar 16 19:11:39 2024 daemon.info lldpd[10916]: custom TLV op add oui 33:44:31 subtype 2c
Sat Mar 16 19:11:39 2024 daemon.warn lldpcli[10915]: invalid OUI value '3322'
Sat Mar 16 19:11:39 2024 daemon.info lldpcli[10915]: an error occurred while executing last command
Sat Mar 16 19:11:39 2024 daemon.warn lldpcli[10915]: invalid OUI value '3312'
Sat Mar 16 19:11:39 2024 daemon.info lldpcli[10915]: an error occurred while executing last command
Sat Mar 16 19:11:39 2024 daemon.info lldpcli[10915]: lldpd should resume operations

( The last two TLV are invalid: their oui must be three hex bytes, comma
separated. Only the first hex byte of oui-info 5555555555 is used )

Depends on #14867 and its release version bump

Tested on: 22.03.6

Signed-off-by: Paul Donald <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/14872
(cherry picked from commit 8d1fe32c2c46a8de519f9926be2bc062cef66f24)
Link: https://github.com/openwrt/openwrt/pull/18344
Signed-off-by: Robert Marko <[email protected]>

lldpd: get_config_cid_ifaces() -> get_interface_csv()

where csv = comma separated value(s)

Make the function more generic. Can use it for not only 'config'.

Now it can be used to parse interfaces for additional lldpd settings,
e.g. custom-tlv.

Tested on: 22.03.6

Signed-off-by: Paul Donald <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/14872
(cherry picked from commit a015f59880c72a1bdd07de32491f85ce7f8d2a1c)
Link: https://github.com/openwrt/openwrt/pull/18344
Signed-off-by: Robert Marko <[email protected]>

generic: globally enable CONFIG_PCPU_DEV_REFCNT

Quoting the kconfig description for CONFIG_PCPU_DEV_REFCNT:

network device refcount are using per cpu variables if this option is
set. This can be forced to N to detect underflows (with a performance
drop).

This was introduced from kernel 5.13 and was wrongly set as disabled.
Some target actually enables it but this should be always enabled unless
refcount needs to be debugged (unlikely for production images)

Enable in generic and drop the entry in every other target.

Link: https://github.com/openwrt/openwrt/pull/18174
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit ea6f3e49d6509f3bddb43364cf7225b4935fee82)
[ adapt for kernel 5.15 ]
Signed-off-by: Christian Marangi <[email protected]>

scripts/feeds: shallow clone submodules

When a feed has submodules, all its submodules are fully cloned whereas
the feed itself is shallowed. Let's be consistent and perform shallow clones
as well for the submodules.

Signed-off-by: Cedric CHEDALEUX <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18003
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit 9ec32cfb2733856a2ab4caee07d9b3297568381d)
Link: https://github.com/openwrt/openwrt/pull/18168
Signed-off-by: Christian Marangi <[email protected]>

scripts/feeds: shallow clone for specific commit update

When a feed is referenced with a specific commit (i.e. <git_url>^<sha1>),
a full clone was performed and a branch was created from the sha1
and named with the sha1. Other git clones operations are shallowed.

As Git does not support clone at a specific commit, let's first perform
a shallow clone to latest commit, then fetch the relevant commit and
finally checkout it (no more 'pseudo' branch).

It saves bandwith and significantly speeds up the feed update process.

Signed-off-by: Cedric CHEDALEUX <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18003
Signed-off-by: Robert Marko <[email protected]>
(cherry picked from commit 32d0a57dc10811a24816e51f10e55963f40fe462)
Link: https://github.com/openwrt/openwrt/pull/18168
Signed-off-by: Christian Marangi <[email protected]>

kernel: r8126: update to v10.015.00

Changelog: https://github.com/openwrt/rtl8126/compare/10.014.01...10.015.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 3d3328bf5f019507896bfa9ffeb53777fc5bfeaa)

kernel: r8125: update to v9.015.00

Changelog: https://github.com/openwrt/rtl8125/compare/9.014.01...9.015.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit a8dea562aa6146f5955ef081ee3240f0b960edf9)

kernel: r8168: update to v8.055.00

Changelog: https://github.com/openwrt/rtl8168/compare/8.054.00...8.055.00

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 5dd763680e874b0082e4cfbbc729fa75ee056601)

mediatek: mt7622: fix SATA on BPi-R64

A commit which made their way into Linux stable down to 5.15 broke the
SATA support on the BPi-R64.

Fix this by importing a (still pending) patch re-adding the 'syscon'
compatible to the pciesys clock-controller which also contains phy-mode
bits referenced by the ahci_mtk driver expecting to access them using
syscon_regmap_lookup_by_phandle().

Signed-off-by: Daniel Golle <[email protected]>
(cherry picked from commit 69890e16b37d59b55ba64633522c727f957cb2fd)

bmips: shg2500: add missing led controller cells

Fixes the following DT warnings:
../dts/bcm63168-sercomm-shg2500.dts:76.4-14: Warning (reg_format): /ubus/spi@10001000/led-controller@1/led@1:reg: property has invalid length (4 bytes) (#address-cells == 2, #size-cells == 1)
../dts/bcm63168-sercomm-shg2500.dts:75.9-78.5: Warning (avoid_default_addr_size): /ubus/spi@10001000/led-controller@1/led@1: Relying on default #address-cells value
../dts/bcm63168-sercomm-shg2500.dts:75.9-78.5: Warning (avoid_default_addr_size): /ubus/spi@10001000/led-controller@1/led@1: Relying on default #size-cells value

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit fbe0bd5f6453a61fab871bee56883afc5c6308cf)

realtek: Fix old compatible for HPE 1920-8G PoE

Supported devices are listed in the metadata as the first part of the
DTS compatible. This normally follows the format "vendor,device".

When updating the device name of the 180W 1920-8G PoE an underscore was
used, instead of a comma, to join the vendor and device name. This will
lead to warnings for users wanting to sysupgrade a device with an older
compatible, as the device's info does not match the one the metadata.

Fixes: 987c96e88927 ("realtek: rename hpe,1920-8g-poe to match hardware")
Signed-off-by: Sander Vanheule <[email protected]>
(cherry picked from commit 6a7fa68569ac11bb2c2abb14026e8b84acd3a12f)

generic: backport pending gpio-regmap ops patch

This patch has been accepted for linux v6.14 so we can move it from pending
to backport.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit 2f2e21a52bdfdcf83868e3169768fda134aa883b)

generic: add gpio-regmap request/free ops patch

This patch is needed on bmips since it fixes issues with GPIOs not being
properly configured due to gpio_request_enable not being called on bcm63xx
devices. Therefore we can now drop the bcm63268 gpio function patch.

Backported from f5b1d340be9488af71e6522c3f10f1ebd1a35446 with the exception of
the realtek and bmips patches removal.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit f5b1d340be9488af71e6522c3f10f1ebd1a35446)

ath79: Add support for Sophos AP15C

The Sophos AP15C uses the same hardware as the AP15, but has a reset button.

Based on:
commit 6f1efb289837 ("ath79: add support for Sophos AP100/AP55 family")
author    Andrew Powers-Holmes <[email protected]>
          Fri, 3 Sep 2021 15:53:57 +0200 (23:53 +1000)
committer Hauke Mehrtens <[email protected]>
          Sat, 16 Apr 2022 16:59:29 +0200 (16:59 +0200)

Unique to AP15C:
 - Reset button
 - External RJ45 serial console port

Flashing instructions:

This firmware can be flashed either via a compatible Sophos SG or XG
firewall appliance, which does not require disassembling the device, or via
the U-Boot console available on the internal UART header.

To flash via XG appliance:
 - Register on Sophos' website for a no-cost Home Use XG firewall license
 - Download and install the XG software on a compatible PC or virtual
   machine, complete initial appliance setup, and enable SSH console access
 - Connect the target AP device to the XG appliance's LAN interface
 - Approve the AP from the XG Web UI and wait until it shows as Active
   (this can take 3-5 minutes)
 - Connect to the XG appliance over SSH and access the Advanced Console
   (Menu option 5, then menu option 3)
 - Run `sudo awetool` and select the menu option to connect to an AP via
   SSH. When prompted to enable SSH on the target AP, select Yes.
 - Wait 2-3 minutes, then select the AP from the awetool menu again. This
   will connect you to a root shell on the target AP.
 - Copy the firmware to /tmp/openwrt.bin on the target AP via SCP/TFTP/etc
 - Run `mtd -r write /tmp/openwrt.bin astaro_image`
 - When complete, the access point will reboot to OpenWRT.

To flash via U-Boot serial console:
 - Configure a TFTP server on your PC, and set IP address 192.168.99.8 with
   netmask 255.255.255.0
 - Copy the firmware .bin to the TFTP server and rename to 'uImage_AP15C'
 - Open the target AP's enclosure and locate the 4-pin 3.3V UART header [4]
 - Connect the AP ethernet to your PC's ethernet port
 - Connect a terminal to the UART at 115200 8/N/1 as usual
 - Power on the AP and press a key to cancel autoboot when prompted
 - Run the following commands at the U-Boot console:
    - `tftpboot`
    - `cp.b $fileaddr 0x9f070000 $filesize`
    - `boot`
 - The access point will boot to OpenWRT.

Signed-off-by: David Lutz <[email protected]>

bmips: dts: fix pinctrl error

The kernel logs the error "bcm6368_nand 10000200.nand: there is not valid
maps for state default" on boot and all nand pins show as UNCLAIMED in
sysfs pinmux-pins.

bcm6362.dtsi, bcm6368.dtsi and bcm63268.dtsi use the undocumented property
group which the driver doesn't understand. This has been documented upstream
in commit caf963efd4b0b9ff42ca12e52b8efe277264d35b.

Replacing group with pins allows the nand pins to be properly configured.

Signed-off-by: Kyle Hendry <[email protected]>
[add bcm636/bcm6368 and fix commit title]
Signed-off-by: Álvaro Fernández Rojas <[email protected]>
(cherry picked from commit d1e9c50d06a8cb618cb85ab489cbcccaec220636)

ipq40xx: make GL.iNet A1300 switch functional

Set the physical switch to KEY_RFKILL, since its previous value
(KEY_SETUP) is unsupported. This should also make the KEY_RESET button
functional, by allowing the gpio-button-hotplug kmod to load.

Signed-off-by: Chris Jones <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16564
Signed-off-by: Mieczyslaw Nalewaj <[email protected]>
(cherry picked from commit 83a04cd2b89f3bca2be3a4467b7ea710c7c6ba4f)
Link: https://github.com/openwrt/openwrt/pull/17228
Signed-off-by: Robert Marko <[email protected]>

build: Unconditionally append kmod feed for BuildBot

Recent changes to BuildBot config moved the kmods to a dedicated
directory and dropped them from the packages dir. This was needed as
both OPKG and APK gets confused if both entry are present.

To fix this, unconditionally append the kmod feed line if
CONFIG_BUILDBOT is enabled.

Fixes: #17146
Link: https://github.com/openwrt/openwrt/pull/17151
(cherry picked from commit 53ee2e8c0339c1d7ace47b8cf390147b86ea8d1d)
[ fix conflict error for missing APK support ]
Signed-off-by: Christian Marangi <[email protected]>

build: profiles.json: add kernel version information

Currently downstream tools like ASU lack information about kernel
version to find out the relevant kmod build folder on downloads server.

So lets fix it by providing a new `linux_kernel` JSON array which would
for the start provide Linux kernel version, revision and vermagic
information.

  "linux_kernel": {
     "release": "1",
     "vermagic": "b57450c07d3a786158c3601fc5cee57d",
     "version": "6.6.61"
   },

Fixes: openwrt/openwrt#17036
Fixes: efahl/owut#9
Co-developed-by: Petr Štetiar <[email protected]>
Signed-off-by: Eric Fahlgren <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17042
Signed-off-by: Petr Štetiar <[email protected]>
(cherry picked from commit c857145e034f623e31ab2028049a547ecd94ce85)

build: ensure silent Make behavior for json scripts

Run the invocation of Make with verbosity in order to
prevent the printing of Makefile level and subtarget status.
e.g.  make[3] -C target/linux val.DEFAULT_PACKAGES val.ARCH_PACKAGES

Remove piping of stderr, which is only useful when using
the "communicate" method over the "run" method,
and this script would not be written to handle a captured error anyway.

For error testing, stdout and stderr can be set to a file object
with the open() function like this:

out = open('json_out', 'w')
err = open('json_err', 'w')

...
...
stdout=out,
stderr=err,

Signed-off-by: Michael Pratt <[email protected]>
(cherry picked from commit fd3376c5eeccc1f1753483ed31ffff03808ce31d)

wifi-scripts: Fix parsing of Capabilities

Fixup capabilities parsing in iw output.

In addition to the normal capabilities iw now also outputs HE MAC, HE
PHY and EHT MAC and EHT PHY capabilities. Exclude them in the parsing.

The grep returns this with mac80211-hwsim:
```
root@OpenWrt:~# iw phy phy0 info | grep 'Capabilities:'
Capabilities: 0x107e
HE PHY Capabilities: (0x02bfce0000000000000000):
EHT PHY Capabilities: (0x7c0000feffff7f01):
HE PHY Capabilities: (0x02bfce0000000000000000):
EHT PHY Capabilities: (0x7c0000feffff7f01):
HE PHY Capabilities: (0x02bf000000000000000000):
Capabilities: 0x107e
HE PHY Capabilities: (0x1cbfce0000000000000000):
EHT PHY Capabilities: (0xfc1f3ffeffff7f37):
HE PHY Capabilities: (0x1cbfce0000000000000000):
EHT PHY Capabilities: (0xfc1f3ffeffff7f37):
HE PHY Capabilities: (0x1cbf000000000000000000):
HE PHY Capabilities: (0x1cbfce0000000000000000):
EHT PHY Capabilities: (0xfefffffeffffff7f):
HE PHY Capabilities: (0x1cbfce0000000000000000):
EHT PHY Capabilities: (0xfefffffeffffff7f):
HE PHY Capabilities: (0x1cbf000000000000000000):
Capabilities: 0x107e
```

With busybox 1.36.1 the ht_cap_mask variable will be set to
-72057598332895361. With busybox 1.37.0 it will be set to -1.
Both values are wrong, after this change it will be set to 4222
(0x107E).

Link: https://github.com/openwrt/openwrt/pull/17043
Signed-off-by: Hauke Mehrtens <[email protected]>
(cherry picked from commit adf958c9192d94fbeeb0ec8e16800b0071455909)

ath79: lzma-loader: fix syntax error

C compiler can't parse '#else if'.

Fixes: f84a9f7dc095 ("ath79: add support for Huawei AP6010DN")
Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16989
Signed-off-by: Hauke Mehrtens <[email protected]>
(cherry picked from commit 98f26346cbbb45e5c0417f57846ce1783e884ea2)

ath79: add support for Huawei AP6010DN

Huawei AP6010DN is a dual-band, dual-radio 802.11a/b/g/n 2x2 MIMO
enterprise access point with one Gigabit Ethernet port and PoE
support.

Hardware highlights:
- CPU: AR9344 SoC at 480MHz
- RAM: 128MB DDR2
- Flash: 32MB SPI-NOR
- Wi-Fi 2.4GHz: AR9344-internal radio
- Wi-Fi 5GHz: AR9580 PCIe WLAN SoC
- Ethernet: 10/100/1000 Mbps Ethernet through Atheros AR8035 PHY
- PoE: yes
- Standalone 12V/2A power input
- Serial console externally available through RJ45 port
- External watchdog: CAT706SVI (1.6s timeout)

Serial console:
  9600n8 (9600 baud, no stop bits, no parity, 8 data bits)

MAC addresses:
  Each device has 32 consecutive MAC addresses allocated by
  the vendor, which don't overlap between devices.
  This was confirmed with multiple devices with consecutive
  serial numbers.
  The MAC address range starts with the address on the label.
  To be able to distinguish between the interfaces,
  the following MAC address scheme is used:
    - eth0 = label MAC
    - radio0 (Wi-Fi 2.4GHz) = label MAC + 1
    - radio1 (Wi-Fi 5GHz) = label MAC + 2

Installation:
0. Connect some sort of RJ45-to-USB adapter to "Console" port of the AP

1. Power up the AP

2. At prompt "Press f or F  to stop Auto-Boot in 3 seconds",
   do what they say.
   Log in with default admin password "[email protected]".

3. Boot the OpenWrt initramfs from TFTP using the hidden script "run ramboot".
   Replace IP address as needed:

   > setenv serverip 192.168.1.10
   > setenv ipaddr 192.168.1.1
   > setenv rambootfile openwrt-ath79-generic-huawei_ap6010dn-initramfs-kernel.bin
   > saveenv
   > run ramboot

4. Optional but recommended as the factory firmware cannot be downloaded publicly:
   Back up contents of "firmware" partition using the web interface or ssh:

   $ ssh [email protected] cat /dev/mtd11 > huawei_ap6010dn_fw_backup.bin

5. Run sysupgrade using sysupgrade image. OpenWrt
   shall boot from flash afterwards.

Return to factory firmware (using firmware upgrade package downloaded from non-public Huawei website):
1. Start a TFTP server in the directory where
   the firmware upgrade package is located

2. Boot to u-boot as described above

3. Install firmware upgrade package and format the config partitions:

   > update system FatAP6X10XN_SOMEVERSION.bin
   > format_fs

Return to factory firmware (from previously created backup):
1. Copy over the firmware partition backup to /tmp,
   for example using scp

2. Use sysupgrade with force to restore the backup:
   sysupgrade -F huawei_ap6010dn_fw_backup.bin

3. Boot AP to U-Boot as described above

Quirks and known issues:
- The stock firmware has a semi dual boot concept where the primary
kernel uses a squashfs as root partition and the secondary kernel uses
an initramfs. This dual boot concept is circumvented on purpose to gain
more flash space and since the stock firmware's flash layout isn't
compatible with mtdsplit.
- The external watchdog's timeout of 1.6s is very hard to satisfy
during bootup. This is why the GPIO15 pin connected to the watchdog input
is configured directly in the LZMA loader to output the AHB_CLK/2 signal
which keeps the watchdog happy until the wdt-gpio kernel driver takes
over. Because it would also take too long to read the whole kernel image
from flash, the uImage header only includes the loader which then reads
the kernel image from flash after GPIO15 is configured.

Signed-off-by: Marco von Rosenberg <[email protected]>
(cherry picked from commit f84a9f7dc095967e41ced96e1ea5f9b2b9f40af6)
Link: https://github.com/openwrt/openwrt/pull/16978
Signed-off-by: Hauke Mehrtens <[email protected]>