packages: assign PKG_CPE_ID for all missing packages
The PKG_CPE_ID links to NIST CPE version 2.2.
Assign PKG_CPE_ID to all remaining package which have a CPE ID.
Not every package has CPE id.
Related: https://github.com/openwrt/packages/issues/8534 Signed-off-by: Alexander Couzens <lynxis@fe80.eu> Signed-off-by: Arne Zachlod <arne@nerdkeller.org>
- CVE-2022-25640: A TLS v1.3 server who requires mutual authentication
can be bypassed. If a malicious client does not send the
certificate_verify message a client can connect without presenting a
certificate even if the server requires one.
- CVE-2022-25638: A TLS v1.3 client attempting to authenticate a TLS
v1.3 server can have its certificate heck bypassed. If the sig_algo in
the certificate_verify message is different than the certificate
message checking may be bypassed.
Petr Štetiar [Wed, 6 Apr 2022 17:58:56 +0000 (19:58 +0200)]
kernel: generic: add missing symbol for arm64 spectre mitigation
Upstream in commit 3e3904125fcc ("arm64: Mitigate spectre style branch
history side channels") introduced new config symbol
MITIGATE_SPECTRE_BRANCH_HISTORY which I missed in commit d39a6c67dcb4
("kernel: bump 4.14 to 4.14.275") and buildworkers for arm64 targets
started complaining:
Mitigate Spectre style attacks against branch history (MITIGATE_SPECTRE_BRANCH_HISTORY) [Y/n/?] (NEW) aborted!
Fixes: d39a6c67dcb4 ("kernel: bump 4.14 to 4.14.275") Signed-off-by: Petr Štetiar <ynezz@true.cz>
Petr Štetiar [Thu, 24 Mar 2022 05:52:37 +0000 (06:52 +0100)]
imagebuilder: fix broken image generation with external targets
When using external targets there is a symlink being created for the
target under target/linux which then becomes dangling under Image
Builder. Fix it by dereferencing the possible symlink.
Tested on IB with external target, ipq40xx and mvebu.
Petr Štetiar [Thu, 31 Mar 2022 07:44:25 +0000 (09:44 +0200)]
patchelf: backport fix for rpath endianness
This is backport of upstream fix introduced in commit e88d83c8b4e4
("patchelf: Check ELF endianness before writing new runpath") which
fixes broken rpath handling on big endian systems:
Petr Štetiar [Thu, 24 Mar 2022 05:45:04 +0000 (06:45 +0100)]
zlib: backport security fix for a reproducible crash in compressor
Tavis has just reported, that he was recently trying to track down a
reproducible crash in a compressor. Believe it or not, it really was a
bug in zlib-1.2.11 when compressing (not decompressing!) certain inputs.
Tavis has reported it upstream, but it turns out the issue has been
public since 2018, but the patch never made it into a release. As far as
he knows, nobody ever assigned it a CVE.
Runtime tested on ipq40xx/glinet-b1300 and mvebu/turris-omnia.
Petr Štetiar [Wed, 16 Mar 2022 18:23:08 +0000 (19:23 +0100)]
kernel: bump 4.14 to 4.14.272
Added new config symbol `HARDEN_BRANCH_HISTORY` in order to harden
Spectre style attacks against branch history and fixed rejects in
following patches:
Martin Schiller [Wed, 16 Mar 2022 14:04:56 +0000 (15:04 +0100)]
openssl: bump to 1.1.1n
This is a bugfix release. Changelog:
*) Fixed a bug in the BN_mod_sqrt() function that can cause it to loop
forever for non-prime moduli. (CVE-2022-0778)
*) Add ciphersuites based on DHE_PSK (RFC 4279) and ECDHE_PSK
(RFC 5489) to the list of ciphersuites providing Perfect Forward
Secrecy as required by SECLEVEL >= 3.
Rafał Miłecki [Tue, 1 Mar 2022 17:46:27 +0000 (18:46 +0100)]
base-files: call "sync" after initial setup
OpenWrt uses a lot of (b)ash scripts for initial setup. This isn't the
best solution as they almost never consider syncing files / data. Still
this is what we have and we need to try living with it.
Without proper syncing OpenWrt can easily get into an inconsistent state
on power cut. It's because:
1. Actual (flash) inode and data writes are not synchronized
2. Data writeback can take up to 30 seconds (dirty_expire_centisecs)
3. ubifs adds extra 5 seconds (dirty_writeback_centisecs) "delay"
Some possible cases (examples) for new files:
1. Power cut during 5 seconds after write() can result in all data loss
2. Power cut happening between 5 and 35 seconds after write() can result
in empty file (inode flushed after 5 seconds, data flush queued)
Above affects e.g. uci-defaults. After executing some migration script
it may get deleted (whited out) without generated data getting actually
written. Power cut will result in missing data and deleted file.
There are three ways of dealing with that:
1. Rewriting all user-space init to proper C with syncs
2. Trying bash hacks (like creating tmp files & moving them)
3. Adding sync and hoping for no power cut during critical section
This change introduces the last solution that is the simplest. It
reduces time during which things may go wrong from ~35 seconds to
probably less than a second. Of course it applies only to IO operations
performed before /etc/init.d/boot . It's probably the stage when the
most new files get created.
All later changes are usually done using smarter C apps (e.g. busybox or
uci) that creates tmp files and uses rename() that is expected to be
atomic.
Matthew Gyurgyik [Fri, 13 Nov 2020 19:21:29 +0000 (11:21 -0800)]
imagebuilder: fix partition signature
When building images with the imagebuilder, the partition signature
never changes. The signature is generated by hashing SOURCE_DATE_EPOCH
and LINUX_VERMAGIC which are undefined. Prepopulate these variables, as
done by the SDK.
Petr Štetiar [Tue, 22 Feb 2022 19:00:28 +0000 (20:00 +0100)]
wolfssl: fix API breakage of SSL_get_verify_result
Backport fix for API breakage of SSL_get_verify_result() introduced in
v5.1.1-stable. In v4.8.1-stable SSL_get_verify_result() used to return
X509_V_OK when used on LE powered sites or other sites utilizing
relaxed/alternative cert chain validation feature. After an update to
v5.1.1-stable that API calls started returning X509_V_ERR_INVALID_CA
error and thus rendered all such connection attempts imposible:
$ docker run -it openwrt/rootfs:x86_64-21.02.2 sh -c "wget https://letsencrypt.org"
Downloading 'https://letsencrypt.org'
Connecting to 18.159.128.50:443
Connection error: Invalid SSL certificate
Petr Štetiar [Mon, 21 Feb 2022 06:41:05 +0000 (07:41 +0100)]
ubus: backport fixes for UAF and other issues
Backporting following fixes:
a72457b61df0 libubus: increase stack depth for processing obj msgs ef038488edc3 libubus: process pending messages in data handler if stack depth is 0 2099bb3ad997 libubus: use list_empty/list_first_entry in ubus_process_pending_msg
where at least commit 2099bb3ad997 ("libubus: use
list_empty/list_first_entry in ubus_process_pending_msg") fixes UAF
issue I've introduced in commit c5f2053dfcfd ("workaround possibly false
positive uses of memory after it is freed") while fixing another false
positive UAF reported[1] by clang's static analyzer.
Those fixes are being used in master/21.02 for about 6 months, so should
be tested enough and considered for backporting. I've runtested those
fixes on mvebu/turris-omnia and ipq40xx/glinet-b1300 devices.
The diff of security related changes we would need to backport would be
so huge, that there would be a high probability of introducing new
vulnerabilities, so it was decided, that bumping to latest stable
release is the prefered way for fixing following security issues:
* OCSP request/response verification issue. (fixed in 4.8.0)
* Incorrectly skips OCSP verification in certain situations CVE-2021-38597 (fixed in 4.8.1)
* Issue with incorrectly validating a certificate (fixed in 5.0.0)
* Hang with DSA signature creation when a specific q value is used (fixed in 5.0.0)
* Client side session resumption issue (fixed in 5.1.0)
* Potential for DoS attack on a wolfSSL client CVE-2021-44718 (fixed in 5.1.0)
* Non-random IV values in certain situations CVE-2022-23408 (fixed in 5.1.1)
Cc: Hauke Mehrtens <hauke@hauke-m.de> Cc: Eneas U de Queiroz <cotequeiroz@gmail.com> Signed-off-by: Petr Štetiar <ynezz@true.cz> Acked-by: Hauke Mehrtens <hauke@hauke-m.de> Acked-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Petr Štetiar [Wed, 16 Feb 2022 16:50:59 +0000 (17:50 +0100)]
kernel: bump 4.14 to 4.14.267
All patches refreshed automagically without conflicts, but test builds
choked on new BPF_UNPRIV_DEFAULT_OFF kernel config symbol introduced in
upstream commit e69f08ba23a3 ("bpf: Add kconfig knob for disabling
unpriv bpf by default").
Run tested on ipq40xx/glinet-b1300 and mvebu/turris-omnia.
Hauke Mehrtens [Sat, 29 Jan 2022 10:56:27 +0000 (11:56 +0100)]
mbedtls: Update to version 2.16.12
This fixes the following security problems:
* Zeroize several intermediate variables used to calculate the expected
value when verifying a MAC or AEAD tag. This hardens the library in
case the value leaks through a memory disclosure vulnerability. For
example, a memory disclosure vulnerability could have allowed a
man-in-the-middle to inject fake ciphertext into a DTLS connection.
* Fix a double-free that happened after mbedtls_ssl_set_session() or
mbedtls_ssl_get_session() failed with MBEDTLS_ERR_SSL_ALLOC_FAILED
(out of memory). After that, calling mbedtls_ssl_session_free()
and mbedtls_ssl_free() would cause an internal session buffer to
be free()'d twice. CVE-2021-44732
The sizes of the ipk changed on MIPS 24Kc like this:
182454 libmbedtls12_2.16.11-2_mips_24kc.ipk
182742 libmbedtls12_2.16.12-1_mips_24kc.ipk
Hauke Mehrtens [Sat, 12 Feb 2022 22:13:47 +0000 (23:13 +0100)]
tcpdump: Fix CVE-2018-16301
This fixes the following security problem:
The command-line argument parser in tcpdump before 4.99.0 has a buffer
overflow in tcpdump.c:read_infile(). To trigger this vulnerability the
attacker needs to create a 4GB file on the local filesystem and to
specify the file name as the value of the -F command-line argument of
tcpdump.
Paul Spooren [Sun, 1 Nov 2020 21:53:39 +0000 (11:53 -1000)]
build: store SOURCE_DATE_EPOCH in JSON info files
The source date epoch is the only reproducible date close to the actual
build date. It can be used for tooling like the firmware wizard to show
the image age.
Signed-off-by: Paul Spooren <mail@aparcar.org>
(cherry picked from commit 165f0b00cdd2f763c1d478c2f58c535fc19b13bd)
[store source_date_epoch as integer] Signed-off-by: Paul Spooren <mail@aparcar.org>
*) Avoid loading of a dynamic engine twice.
*) Fixed building on Debian with kfreebsd kernels
*) Prioritise DANE TLSA issuer certs over peer certs
*) Fixed random API for MacOS prior to 10.12
Petr Štetiar [Thu, 23 Dec 2021 10:37:55 +0000 (11:37 +0100)]
kernel: bump 4.14 to 4.14.259
All patches refreshed automagically without conflicts, but upstream in
commit 48c2461f28fe ("ARM: 8800/1: use choice for kernel unwinders")
added new config options UNWINDER_ARM and UNWINDER_FRAME_POINTER so we
need to adjust default configs as well.
Run tested on ipq40xx/glinet-b1300 and mvebu/turris-omnia.
Signed-off-by: Petr Štetiar <ynezz@true.cz> Acked-by: Hauke Mehrtens <hauke@hauke-m.de>
Hauke Mehrtens [Sun, 12 Dec 2021 20:43:21 +0000 (21:43 +0100)]
mac80211: Update to version 4.19.221
The following patch was backported from upstream before and is not
needed any more:
package/kernel/mac80211/patches/ath/980-ath10k-fix-max-antenna-gain-unit.patch
Roman Yeryomin [Fri, 3 Sep 2021 14:31:11 +0000 (17:31 +0300)]
iproute2: m_xt.so depends on dynsyms.list
When doing parallel build on a fast machine with bottleneck in i/o,
m_xt.so may start linking faster than dynsyms.list gets populated,
resulting in error:
ld:dynsyms.list:0: syntax error in dynamic list
Fix this by adding dynsyms.list as make dependency to m_xt.so
Described also here:
https://bugs.openwrt.org/index.php?do=details&task_id=3353
Change from v1:
- add dynsysms.list dependancy only when shared libs are enabled
At least since gcc 7.3.0 (OpenWrt 18.06) lwr/lwl are used in the
assembly of LzmaProps_Decode. While the decission made by the compiler
looks perfect fine, it triggers some obscure hang on lantiq danube-s
v1.5 with MX29LV640EB NOR flash chips.
Only if the offset 1 is used, the hang can be observed. Using any other
offset works fine:
e983a25 Update regulatory rules for Ecuador (EC) a0bcb88 wireless-regdb: Update regulatory rules for Norway (NO) on 6 and 60 GHz cdf854d wireless-regdb: Update regulatory rules for Germany (DE) on 6GHz 86cba52 wireless-regdb: reduce bandwidth for 5730-5850 and 5850-5895 MHz in US 6fa2384 wireless-regdb: remove PTMP-ONLY from 5850-5895 MHz for US 9839e1e wireless-regdb: recent FCC report and order allows 5850-5895 immediately 42dfaf4 wireless-regdb: update 5725-5850 MHz rule for GB
1. These patches (ath, ath9k, mac80211) were included in kernel
versions since 4.14.245 and 4.19.205. They fix security vulnerability
CVE-2020-3702 [1] similar to KrØØk, which was found by ESET [2].
Thank you Josef Schlehofer for reporting this problem.
Petr Štetiar [Sun, 1 Nov 2020 16:31:40 +0000 (17:31 +0100)]
sdk: fix missing include directories
It's not possible to compile some applications which are using
`-Werror=missing-include-dirs` compiler flags with the SDK as some
target directories are missing in the SDK tarball:
cc1: error: staging_dir/target/usr/include: No such file or directory [-Werror=missing-include-dirs]
cc1: error: staging_dir/target/include: No such file or directory [-Werror=missing-include-dirs]
Fix this by adding the missing directories in the SDK.
Sven Eckelmann [Sat, 11 Sep 2021 19:03:34 +0000 (21:03 +0200)]
uboot-tegra: Fix build with GCC-10 as host compiler
The package uses the host compiler to build the dtc binary. With gcc-10,
the option -fno-common is now the default behavior. Thus multiple
definitions of the same variable are now forbidden and results in following
error during linking:
HOSTLD scripts/dtc/dtc
/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x10): multiple definition of `yylloc'; scripts/dtc/dtc-lexer.lex.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
The easiest workaround is to add the upstream commit 018921ee79d3 ("Remove
redundant YYLOC global declaration").
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Sven Eckelmann [Sat, 11 Sep 2021 19:03:34 +0000 (21:03 +0200)]
uboot-mvebu: Fix build with GCC-10 as host compiler
The package uses the host compiler to build the dtc binary. With gcc-10,
the option -fno-common is now the default behavior. Thus multiple
definitions of the same variable are now forbidden and results in following
error during linking:
HOSTLD scripts/dtc/dtc
/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x10): multiple definition of `yylloc'; scripts/dtc/dtc-lexer.lex.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
The easiest workaround is to add the upstream commit 018921ee79d3 ("Remove
redundant YYLOC global declaration").
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Sven Eckelmann [Sat, 11 Sep 2021 19:03:34 +0000 (21:03 +0200)]
uboot-kirkwood: Fix build with GCC-10 as host compiler
The package uses the host compiler to build the dtc binary. With gcc-10,
the option -fno-common is now the default behavior. Thus multiple
definitions of the same variable are now forbidden and results in following
error during linking:
HOSTLD scripts/dtc/dtc
/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x10): multiple definition of `yylloc'; scripts/dtc/dtc-lexer.lex.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
The easiest workaround is to add the upstream commit 018921ee79d3 ("Remove
redundant YYLOC global declaration").
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Sven Eckelmann [Sat, 11 Sep 2021 19:03:34 +0000 (21:03 +0200)]
uboot-sunxi: Fix build with GCC-10 as host compiler
The package uses the host compiler to build the dtc binary. With gcc-10,
the option -fno-common is now the default behavior. Thus multiple
definitions of the same variable are now forbidden and results in following
error during linking:
HOSTLD scripts/dtc/dtc
/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x10): multiple definition of `yylloc'; scripts/dtc/dtc-lexer.lex.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
The easiest workaround is to add the upstream commit 018921ee79d3 ("Remove
redundant YYLOC global declaration").
Signed-off-by: Sven Eckelmann <sven@narfation.org>
This sets the --cross-compile-prefix option when running Configure, so
that that it will not use the host gcc to figure out, among other
things, compiler defines. It avoids errors, if the host 'gcc' is
handled by clang:
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com> Tested-by: Rosen Penev <rosenp@gmail.com>
(cherry picked from commit 2f75348923e564f1b73fbc32f7cabc355cd6e2b9)
Petr Štetiar [Thu, 3 Jun 2021 07:49:58 +0000 (09:49 +0200)]
ubus: update to version 2021-06-03
This update cherry picks following changes:
* cmake: add a possibility to set library version
* ubusd: protect against too-short messages
* ubusd: add per-client tx queue limit
* ubusd: convert tx_queue to linked list
* lua: avoid truncation of large numeric values
Fixes: FS#1525 Signed-off-by: Petr Štetiar <ynezz@true.cz>
Petr Štetiar [Thu, 29 Jul 2021 19:49:18 +0000 (21:49 +0200)]
ubus: backport SOVERSION support
Add a support for setting of new `ABIVERSION` CMake define which allows
to control the SOVERSION used for the built shared library. This is
needed for downstream packaging to properly track breaking ABI changes
when updating to newer versions of the library.
Michael Yartys [Thu, 3 Jun 2021 21:50:12 +0000 (23:50 +0200)]
ath10k-ct: add security fixes
This rebases -ct changes on top of upstream stable kernel's latest code.
Including the wifi security fixes that recently went in.
Removed upstreamed 203-ath10k-Limit-available-channels-via-DT-ieee80211-fre.patch
and refreshed patches.
Signed-off-by: Michael Yartys <michael.yartys@protonmail.com> Signed-off-by: Petr Štetiar <ynezz@true.cz> [backport]
(backported from commit 2e10ed925e1e07c28570731a429efa5e7de3b826)
The system init script currently sets /tmp/localinfo when zoneinfo is
populated. However, zoneinfo has spaces in it whereas the actual files
have _ instead of spaces. This made the if condition never return true.
Example failure when removing the if condition:
/tmp/localtime -> /usr/share/zoneinfo/America/Los Angeles
This file does not exist. America/Los_Angeles does.
mac80211: distance config: allow "auto" as a value
The user can now enable the ACK timeout estimation algorithm (dynack)
for drivers that support it.
It is also expected that the distance config accepts the same values as:
$ iw phyX set distance XXX
Hauke Mehrtens [Sun, 16 May 2021 21:49:48 +0000 (23:49 +0200)]
tools/mklibs: Fix compile with GCC 11
GCC 11 defaults to C++17, but mklibs does not compile when using the
C++17 standard. This patch switches back to the gnu++98 version like
done in master commit 9437012b9ee4 ("tools/mklibs: update to 0.1.44 and
convert to Python 3")
This fixes the following compile error message:
elf.hpp:52:56: error: ISO C++17 does not allow dynamic exception specifications
52 | const section &get_section(unsigned int i) const throw (std::out_of_range) { return *sections.at(i); };
Hauke Mehrtens [Sun, 16 May 2021 13:31:49 +0000 (15:31 +0200)]
openwrt-keyring: Only copy sign key for 19.07 and 21.02
Instead of adding all public signature keys from the openwrt-keyring
repository only add the key which is used to sign the OpenWrt 19.07
feeds and the 21.02 feeds to allow checking the next release.
If one of the other keys would be compromised this would not affect
users of 19.07 release builds.
Bas Mevissen [Mon, 19 Apr 2021 23:08:19 +0000 (01:08 +0200)]
Extend checks on build prerequisites for building OpenWRT core
OpenWRT requires a number of Perl modules to be installed. It wasn't checking on all of them.
This patch adds checks for Perl FindBin, File::Copy, File::Compare and Thread::Queue modules.
Failing to install these, will have the build break at some point. By adding these to the
prereq-build.mk script, they are checked on forehand.
Tested on a Fedora 33 and 34 (beta) that was freshly installed. Fedora appears to
break up Perl modules into small packages that need to be installed for the build to succeed.
Rosen Penev [Mon, 30 Mar 2020 01:13:56 +0000 (18:13 -0700)]
prereq-build: test for perl's Data::Dumper
Required for installation of autoconf:
make[5]: Entering directory `/openwrt/build_dir/host/autoconf-2.69'
Making all in bin
make[6]: Entering directory `/openwrt/build_dir/host/autoconf-2.69/bin'
autom4te_perllibdir='..'/lib AUTOM4TE_CFG='../lib/autom4te.cfg'
../bin/autom4te -B '..'/lib -B '..'/lib --language M4sh --cache
'' --melt ./autoconf.as -o autoconf.in
Can't locate Data/Dumper.pm in @INC (@INC contains: ../lib
/usr/local/lib64/perl5 /usr/local/share/perl5 /usr/lib64/perl5/vendor_perl
/usr/share/perl5/vendor_perl /usr/lib64/perl5 /usr/share/perl5 .) at
../lib/Autom4te/C4che.pm line 33.
BEGIN failed--compilation aborted at ../lib/Autom4te/C4che.pm line 33.
Compilation failed in require at ../bin/autom4te line 40.
BEGIN failed--compilation aborted at ../bin/autom4te line 40.
make[6]: *** [autoconf.in] Error 2
Petr Štetiar [Fri, 9 Apr 2021 12:52:05 +0000 (14:52 +0200)]
tplink-safeloader: fix C7v5 factory flashing from vendor fw > v1.1.x
Currently it's not possible to flash factory images on devices shipped
with vendor firmware versions 1.1.0 Build 20201120 rel. 50406 (published
2020-12-22):
Hauke Mehrtens [Sun, 2 May 2021 15:35:16 +0000 (17:35 +0200)]
dropbear: Fix CVE-2020-36254
This backports a fix from dropbear 2020.81.
CVE-2020-36254 description:
scp.c in Dropbear before 2020.79 mishandles the filename of . or an empty filename, a related issue to CVE-2018-20685.
David Bauer [Sun, 2 May 2021 22:08:38 +0000 (00:08 +0200)]
ramips: backport unlocked mdiobus accessors
Commit 718e97c5c843 ("ramips: mt7530 swconfig: fix race condition in
register access") backports a fix which depends on unlocked MMD
accessors, however these were not yet included in Kernel 4.14 and they
were not backported yet.
Magnus Kroken [Wed, 21 Apr 2021 20:10:58 +0000 (22:10 +0200)]
openvpn: update to 2.4.11
Fixes two related security vulnerabilities (CVE-2020-15078) which under
very specific circumstances allow tricking a server using delayed
authentication (plugin or management) into returning a PUSH_REPLY before
the AUTH_FAILED message, which can possibly be used to gather
information about a VPN setup.
This release also includes other bug fixes and improvements.
Magnus Kroken [Wed, 21 Apr 2021 20:10:57 +0000 (22:10 +0200)]
openvpn: update to 2.4.9
This is primarily a maintenance release with bugfixes and improvements.
This release also fixes a security issue (CVE-2020-11810) which allows
disrupting service of a freshly connected client that has not yet
negotiated session keys. The vulnerability cannot be used to
inject or steal VPN traffic.
Release announcement:
https://openvpn.net/community-downloads/#heading-13812
Full list of changes:
https://community.openvpn.net/openvpn/wiki/ChangesInOpenvpn24#OpenVPN2.4.9
The mt7530_{r,w}32 operation over MDIO uses 3 mdiobus operations and
does not hold a lock, which causes a race condition when multiple
threads try to access a register, they may get unexpected results.
To avoid this, handle the MDIO lock manually, and use the unlocked
__mdiobus_{read,write} in the critical section.
This fixes the "Ghost VLAN" artifact[1] in MT7530/7621 when the VLAN
operation and the swconfig LED link status poll race between each other.
projects / thirdparty / openwrt.git / log
