This patch initially introduced in 94b4da9b4a
breaks mvebu devices when using vlan filtering with kernel 5.15 or 6.1,
it was working ok in 5.10.
With this patch, frame that should exit untagged from the switch exit tagged.
Running 'tcpdump -Q out -evnnli eth1' (eth1 being the dsa interface)
- with the hack, frame is sent directly to the
destination port 4 with VID 2, so the switch leave the tag as instructed:
11:22:33:44:55:66 > 77:88:99:aa:bb:cc, Marvell EDSA ethertype 0xdada (Unknown), rsvd 0 0, mode From CPU, target dev 0, port 4, tagged, VID 2, FPri 0, ethertype ARP (0x0806), length 50: Ethernet (len 6), IPv4 (len 4), Request who-has 5.6.7.8 tell 1.2.3.4, length 28
- without the hack, frame is sent to the switch that
performs the forwarding decision and untagging:
11:22:33:44:55:66 > 77:88:99:aa:bb:cc, Marvell EDSA ethertype 0xdada (Unknown), rsvd 0 0, mode Forward, dev 1, port 0, tagged, VID 2, FPri 0, ethertype ARP (0x0806), length 50: Ethernet (len 6), IPv4 (len 4), Request who-has 5.6.7.8 tell 1.2.3.4, length 28
Removing this patch makes my Turris Omnia usable with vlan filtering,
ie wifi device can talk to wired device again.
Using kernel 5.15 some broadcast/multicast traffic is still leaked
(on a VLAN 2 access port I see tagged VLAN 3 frame),
using kernel 6.1 fixes that.
People needing the extra performance should try the bridger package.
Acked-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Etienne Champetier <champetier.etienne@gmail.com>
(cherry picked from commit 244328b19c)
Some devices rename cferam bootloader using specific patterns and don't follow
broadcom standards for renaming cferam files. This requires supporting
different cferam file names.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
(cherry picked from commit 8813edd8d9)
Upstream DSA driver is exporting symbols with the same name as our
downstream swconfig driver, so lets rename the downstream symbols to make
them unique and avoid the conflict on 6.1 kernel.
Without this change, building 6.1 with kmod-switch-bcm53xx would conflict
with the B53 DSA driver and CI would fail.
Signed-off-by: Robert Marko <robimarko@gmail.com>
(cherry picked from commit effccdd444)
Instead of reference vlan and do strange subtraction, use the handy
struct_group() to create a virtual struct of the same size of the
members. This permits to have a more secure memset and fix compilation
warning in 6.1 where additional checks are done.
Fix compilation warning:
| inlined from 'psb6970_reset_switch' at drivers/net/phy/psb6970.c:275:2:
| ./include/linux/fortify-string.h:314:25: error: call to '__write_overflow_field'
| declared with attribute warning: detected write beyond size of field
| (1st parameter); maybe use struct_group()? [-Werror=attribute-warning]
| 314 | __write_overflow_field(p_size_field, size);
| | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|cc1: all warnings being treated as errors
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
(cherry picked from commit d69becd307)
Now that it is possible to load several device tree overlays by
appending their config names to bootconf the uImage.FIT partition
parser need to discard everything after the first '#' character in
bootconf when looking up the config node to be used.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
(cherry picked from commit 07bca1adaa)
Recent libcap versions (>= 2.60) cause problems with BPF kselftests, so
backport an upstream patch that replaces libcap and drops the dependency.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit 04981c716a)
The compilation warning was triggered by wrongly set FRAME_WARN to 1024
even for 64bit. This was recently fix by correctly setting the
FRAME_WARN to 2048 for 64bit systems.
The compilation warning would still be triggered on 32bit system but the
actual code is never reached as ARCH_USE_GNU_PROPERTY is only set on
arm64 arch.
Drop the patch as kmalloc cause perf regression as suggested by upstream
maintainers.
Fixes: fa79baf4a6 ("generic: copy backport, hack, pending patch and config from 5.15 to 6.1")
Fixes: 5913ea1ba2 ("generic: 5.15: add pending patch fixing binfmt compilation warning")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
(cherry picked from commit 62338f4162)
Backport Russell King's series [1]
net: mvneta: reduce size of TSO header allocation
to pending-5.15 to fix random crashes on Turris Omnia.
This also backports two patches that are dependencies to this series:
net: mvneta: Delete unused variable
net: mvneta: fix potential double-frees in mvneta_txq_sw_deinit()
[1] https://lore.kernel.org/netdev/ZCsbJ4nG+So%2Fn9qY@shell.armlinux.org.uk/
Signed-off-by: Marek Behún <kabel@kernel.org>
Signed-off-by: Christian Lamparter <chunkeey@gmail.com> (squashed)
(cherry picked from commit 7b31c2e9ed)
This set the CONFIG_FRAME_WARN option depending on some target settings.
It will use the default from the upstream kernel and not the hard coded
value of 1024 now.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
(cherry picked from commit 16a20512d8)
The USXGMII driver in SDK was heavily refactored, some bugs have been
fixed and it has switched to use phylink_pcs. Follow up with changes
in SDK driver and sync our on-top-of-mainline driver with the SDK
driver.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
(cherry picked from commit ba58245e83)
CONFIG_PAHOLE_HAS_SPLIT_BTF should be runtime detected as it depends on
pahole being available on the host, so filter it out of configs.
Signed-off-by: Robert Marko <robimarko@gmail.com>
(cherry picked from commit 3591353f51)
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Backport commits adding support for the MT7988 built-in switch to the
mt7530 driver.
This change results in the Kconfig symbol NET_DSA_MT7530 to be extended
by NET_DSA_MT7530_MDIO (everything formally covered by NET_DSA_MT7530)
and NET_DSA_MT7530_MMIO (a new driver for the MMIO-connected built-in
switch of the MT7988 SoC).
Select NET_DSA_MT7530_MDIO for all targets previously selecting
NET_DSA_MT7530, with the exception of mediatek/filogic which also
selects NET_DSA_MT7530_MMIO.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
(cherry picked from commit 958fdf36e3)
In order to support Ethernet on the MT7988 SoC add support for NETSYS v3
as well as new paths and USXGMII SerDes to the mtk_eth_soc driver.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
(cherry picked from commit 6983a215d9)
This allows loading modules with large memory requirements, recently needed
while testing on armvirt/32. Past forum discussions [1] and bug reports [2]
also raised this and the ipq806x target already set it in response [3].
Given this increases kernel image size by only ~1KB, is generally useful on
multi-platform kernels, and enabled by default on upstream arm32 Linux, add
it to the generic config.
The setting has similar utility on arm64, is a requirement for KASLR, and
already enabled on most OpenWrt aarch64 targets, so pull this into the
top-level generic config.
[1]: https://forum.openwrt.org/t/vmap-allocation-for-size-442368-failed-use-vmalloc-size-to-increase-size/34545/7
[2]: https://github.com/openwrt/openwrt/issues/8282
[3]: f81e148eb6 ("ipq806x: update 4.19 kernel config").
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit c2d194a34e)
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
This removes unneeded kernel version switches from the targets after
kernel 5.10 has been dropped.
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Manually adjusted before running quilt due to new location in tree:
backport-5.15/780-v5.16-bus-mhi-pci_generic-Introduce-Sierra-EM919X-support.patch
backport-5.15/781-v6.1-bus-mhi-host-always-print-detected-modem-name.patch
pending-5.15/790-bus-mhi-core-add-SBL-state-callback.patch
All other patches automatically rebased.
Build system: x86_64
Build-tested: bcm2711/RPi4B, ramips/tplink_archer-a6-v3, filogic/xiaomi_redmi-router-ax6000-ubootmod
Run-tested: bcm2711/RPi4B, ramips/tplink_archer-a6-v3, filogic/xiaomi_redmi-router-ax6000-ubootmod
Signed-off-by: John Audia <therealgraysky@proton.me>
Import patch to prevent crashes in case WO firmware is missing on devices
with only a single frontend (eg. MT7981+MT7976).
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
It has just been queued for the v6.5. It allows Ethernet drivers to read
MACs from NVRAM using NVMEM interface.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
All targets are bumped to 5.15. Remove the old 5.10 patches, configs
and files using:
find target/linux -iname '*-5.10' -exec rm -r {} \;
Further, remove the 5.10 include.
Signed-off-by: Nick Hainke <vincent@systemli.org>
This patch does not have a valid patch headers and does not apply on
an external git tree with 'git am'. To fix this add the missing header.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
The OF LED configuration patch fails on MT7621, as the necessary headers
were only included for the mediatek subtarget with an additional patch.
Fixes: 242fe8634e ("generic: add hack for MT753x LED configuration")
Signed-off-by: David Bauer <mail@david-bauer.net>
Drop unused variable in psb6970_config_init. This variable should have
been dropped in 0806f8fc80 ("lantiq: add Linux 5.10 support as testing kernel").
This fix compilation warning:
drivers/net/phy/psb6970.c: In function 'psb6970_config_init':
drivers/net/phy/psb6970.c:310:28: warning: unused variable 'dev' [-Wunused-variable]
310 | struct net_device *dev = pdev->attached_dev;
|
cc1: all warnings being treated as errors
Fixes: 0806f8fc80 ("lantiq: add Linux 5.10 support as testing kernel")
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Backport patch present upstream fixing error for bcm63xx:
drivers/watchdog/bcm63xx_wdt.c: In function 'bcm63xx_wdt_ioctl':
drivers/watchdog/bcm63xx_wdt.c:208:17: error: this statement may fall through [-Werror=implicit-fallthrough=]
208 | bcm63xx_wdt_pet();
| ^~~~~~~~~~~~~~~~~
drivers/watchdog/bcm63xx_wdt.c:210:9: note: here
210 | case WDIOC_GETTIMEOUT:
| ^~~~
cc1: all warnings being treated as errors
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Add pending patch fixing binfmt compilation warning on some arch.
This fix compilation warning:
fs/binfmt_elf.c: In function 'parse_elf_properties.isra':
fs/binfmt_elf.c:821:1: error: the frame size of 1040 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
821 | }
| ^
cc1: all warnings being treated as errors
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Current gcc set a limit for each stack and complain if this limit is not
followed. JFSS2 in build_xattr function currently exceed the limit by
allocating a massive array of struct of 128 elements. Introduce a
pending patch that allocate this temp array dynamically fixing the
compilation warning:
fs/jffs2/xattr.c: In function 'jffs2_build_xattr_subsystem':
fs/jffs2/xattr.c:887:1: error: the frame size of 1088 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
887 | }
| ^
Affecting many target that require jffs2 support.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
From 5.15 and up linux kernel introduced CONFIG_WERROR to flag any
warning as error. To improve code quality, enable this by default to
catch any warning and fix it.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
This adds a small hack similar to the one used for ar8xxx switches to
read a reg:value map for configuring the LED configuration registers.
This allows OpenWrt to write device-specific LED action as well as blink
configurations. It is unlikely to be accepted upstream, as upstream
plans on integrating their own framework for handling these LEDs.
Signed-off-by: David Bauer <mail@david-bauer.net>
The MT753x switches provide a switch-internal MDIO bus for the embedded
PHYs.
Register a OF sub-node on the switch OF-node for this internal MDIO bus.
This allows to configure the embedded PHYs using device-tree.
Signed-off-by: David Bauer <mail@david-bauer.net>
Since CONFIG_DYNAMIC_DEBUG is already managed via the KERNEL_DYNAMIC_DEBUG
setting in Config-kernel.in (default N), remove or disable it in target
configs which unconditionally enable it, along with the related setting
CONFIG_DYNAMIC_DEBUG_CORE. This saves several KB in the kernels for
ipq40xx, ipq806x, filogic, mt7622, qoriq, and sunxi.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
This activates CONFIG_SLAB_FREELIST_RANDOM.
This option make the free list less predictable. This makes it harder to
exploit heap based security vulnerabilities.
This adds a little bit more code to the kernel and a small additional
compute overhead.
This option is activated in Debian by default.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Early versions (?) of the RTL8221B PHY cannot be identified in a regular
Clause-45 bus scan as the PHY doesn't report the implemented MMDs
correctly but returns 0 instead.
Implement custom identify function using the PKGID instead of iterating
over the implemented MMDs to work-around this problem.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
This activates the CONFIG_SCHED_STACK_END_CHECK option.
The kernel will check if the kernel stack overflowed in the schedule()
function. This just adds a very small computational overhead.
This option is activated in Debian by default.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This activates some extra checks in SLAB or SLUB to make it harder to
execute kernel heap exploits. This adds a minor performance
degradation which I haven't measured-.
Many mainstream Linux distributions also activate this option.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This activates the following kernel options by default:
* CONFIG_RANDOM_TRUST_CPU
* CONFIG_RANDOM_TRUST_BOOTLOADER
With these option Linux will also use data from the CPU RNG e.g. RDRAND
and the bootloader to initialize the Linux RNG if such sources are
available.
These random bits are used in addition to the other sources, no other
sources are getting deactivated. I read that the Chacha mixer isn't
vulnerable to injected entropy, so this should not be a problem even if
these sources might inject bad random data.
The Linux kernel suggests to activate both options, Debian also
activates them. This does not increase kernel code size.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
OpenWrt's developer guide prefers having actual patches so they an be
sent upstream more easily.
However, in the case of hack-5.15 patches which are not meant for
upstream, adding proper fields allows for `git am` to properly function.
This commit tries to rectify that, by digging in the history to find
where and how it was first added.
Signed-off-by: Olliver Schinagl <oliver@schinagl.nl>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
The current bcma SPROM extraction from upstream only supports SPROMs with
revisions from 8 to 11.
Let's align the downstream fallback driver with upstream.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
- Remove unneeded mac bytes from struct (it's already present in the SPROM).
- Convert devid_override to boolean.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
- Remove unneeded mac bytes from struct (it's already present in the SPROM).
- Convert devid_override to boolean.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
The previous attempt to replace an open coded paged read in the RealTek
Ethernet PHY driver was too naive and resulted in breaking the r8169
PCIe Ethernet driver which also makes use of the RealTek Ethernet PHY
driver.
Fix this by instead of using the (not yet populated) paged operations
rather use rtl821x_write_page and protect the whole paged read operation
using the MDIO bus mutex.
Fixes: 998b973157 ("kernel: net: phy: realtek: improve RealTek 2.5G PHY driver")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
This adds generic kernel support for Broadcom Fallback SPROMs so that it can be
used in any target, even non Broadcom ones.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
A patch was added in kernel 5.4 to support the fiber operation of
AR8033 with ramips devices. In kernel 5.18 similar enhancements
were added to the kernel. Those patches are required for other
fiber based devices but when added, build fails for ramips targets.
This commit removes the ramips patch and adds the kernel 5.18 ones.
Signed-off-by: Daniel Kestrel <kestrel1974@t-online.de>
[ split commit,refresh patch and improve commit message ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
A patch was added in kernel 5.4 to support the fiber operation of
AR8033 with ramips devices. In kernel 5.18 similar enhancements
were added to the kernel. Those patches are required for other
fiber based devices but when added, build fails for ramips targets.
This commit removes the ramips patch and adds the kernel 5.18 ones.
Signed-off-by: Daniel Kestrel <kestrel1974@t-online.de>
[ split commit, refresh patch and improve commit title ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Replace pending 730-net-phy-at803x-fix-feature-detection.patch with
upstream version and move it to backport.
Refresh other related patch while moving it.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
The patch needs to be refreshed to apply cleanly.
Fixes: 998b973157 ("kernel: net: phy: realtek: improve RealTek 2.5G PHY driver")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
* use interface mode switching only when operating in C45 mode
Linux prevents switching the interface mode when using C22 MDIO,
hence use rate-adapter mode in case the PHY controlled via C22.
* use phy_read_paged where appropriate
* use existing generic inline functions to handle 10GbE advertisements
instead of redundantly defining register macros in realtek.c which
are not actually vendor-specific.
* make sure 10GbE advertisement is valid, preventing false-positive
warning "Downshift occurred from negotiated speed 2.5Gbps to actual
speed 1Gbps, check cabling!" with some link-partners using 1G mode.
* Support Link Down Power Saving Mode (ALDPS)
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Instead of use mac-address-ascii in nvmem_get_mac_address
function, move it into of_get_mac_addr_nvmem function to
support more devices.
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
MEMREAD is a new ioctl for MTD character devices that was first included
in Linux 6.1. It allows userspace applications to use the Linux
kernel's OOB autoplacement mechanism while reading data from NAND
devices. The Yafut tool needs this ioctl to do its job.
Signed-off-by: Michał Kępień <openwrt@kempniu.pl>
Import commit "ubi: Fix failure attaching when vid_hdr offset equals to
(sub)page size" which did not yet make it to stable upstream Linux trees.
Fixes: #12232Fixes: #12339
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
They add NVMEM layouts support. It allows handling NVMEM content
independently of NVMEM device access.
Skip U-Boot env data patch for now as it break our downstream MAC hacks.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
MAC drivers don't use SGMII in-band autonegotiation unless told to do so
in device tree using 'managed = "in-band-status"'. When using MDIO to
access a PHY, in-band-status is unneeded as we have link-status via
MDIO. Switch off SGMII in-band autonegotiation using magic values.
Reported-by: Chen Minqiang <ptpt52@gmail.com>
Reported-by: Chukun Pan <amadeus@jmu.edu.cn>
Reported-by: Yevhen Kolomeiko <jarvis2709@gmail.com>
Tested-by: Yevhen Kolomeiko <jarvis2709@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Some vendor bootloaders do weird things with those PHYs which result in
link modes being reported wrongly. Start from a clean sheet by resetting
the PHY.
Reported-by: Yevhen Kolomeiko <jarvis2709@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Add dynamic interface mode update for the rtl8221 phy to match various
wire speeds. 10M/100M/1000M use SGMII, 2500M uses 2500Base-X.
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
The HG MXPD-483II 1310nm SFP module is meant to operate with 2500Base-X,
however, in their EEPROM they incorrectly specify:
Transceiver type : Ethernet: 1000BASE-LX
...
BR, Nominal : 2600MBd
Use sfp_quirk_2500basex for this module to allow 2500Base-X mode anyway.
https://forum.banana-pi.org/t/bpi-r3-sfp-module-compatibility/14573/60
X-Patchwork-Id: 13197378
X-Mailing-List: netdev@vger.kernel.org
X-Patchwork-Delegate: kuba@kernel.org
Reported-by: chowtom <chowtom@gmail.com>
Tested-by: chowtom <chowtom@gmail.com>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
The XMC XM25QH64C is a 8MB SPI NOR chip. The patch is verified on TL-WPA8631P v3.
Datasheet available at https://www.xmcwh.com/uploads/442/XM25QH64C.pdf
Signed-off-by: Joe Mullally <jwmullally@gmail.com>
Pahole version is being autodetected during runtime since kernel 5.15.96
via in-kernel scripts/pahole-version.sh so add CONFIG_PAHOLE_VERSION to
kernel filter in order to prevent it from being added to target configs.
Signed-off-by: Robert Marko <robimarko@gmail.com>
Use scratch buffer for DMA operetations. Passing a pointer to a stack
variable won't work and results in bogus bit flips being reported.
Patch was submitted upstream and is part of Linux 6.3.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
The MT7531 switch IC comes with SerDes ports with PCS identical to
what is also used in MediaTek's SoCs. Make use of the shared driver
to ease maintainance and reduce code duplication.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Using 2500Base-T SFP modules e.g. on the BananaPi R3 requires manually
disabling auto-negotiation, e.g. using ethtool. While a proper fix
using SFP quirks is being discussed upstream, bring a work-around to
restore user experience to what it was before the switch to the
dedicated SGMII PCS driver.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Backport patch allowing to set the MDIO bus clock frequency.
By default the MDIO bus clock runs on 2.5 MHz, allow increasing it
up to 25 MHz.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
