Use the kernel's built-in formula for computing this value.
The value applied by OpenWRT's sysctl configuration file does not scale
with the available memory, under-using hardware capabilities.
Also, that formula also influences net.netfilter.nf_conntrack_buckets,
which should improve conntrack performance in average (fewer connections
per hashtable bucket).
Backport upstream commit for its effect on the number of connections per
hashtable bucket.
Apply a hack patch to set the RAM size divisor to a more reasonable value (2048,
down from 16384) for our use case, a typical router handling several thousands
of connections.
Signed-off-by: Vincent Pelletier <plr.vincent@gmail.com>
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
Backport patch
8b6836d82470 ("net: dsa: mv88e6xxx: keep the pvid at 0 when VLAN-unaware")
from 5.15.
Keeping the pvid at 0 when VLAN-unaware makes it possible to drop the
hack introduced in commit 920eaab1d8 ("kernel: DSA roaming fix for
Marvell mv88e6xxx"). Dropping the hack makes it possible to use VLAN
interfaces with VID 1 on DSA ports without problems with FDB.
Signed-off-by: Marek Behún <kabel@kernel.org>
In commit ab143647ef ("kernel: generic: improve FIT partition parser")
part_bits was bumped to 2 in order to allow up to 3 additional FIT
sub-images mapped into sub-partitions.
This change has to be reflected also in our local patch
420-mtd-set-rootfs-to-be-root-dev.patch
which still assumed part_bits==1 for mtdblock devices in case of
CONFIG_FIT_PARTITION=y.
Fixes: #9557
Fixes: ab143647ef ("kernel: generic: improve FIT partition parser")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
* only map filesystems configured in 'loadables'
* allow mapping more than one filesystem (e.g. customization/branding
or localization in addition to rootfs)
* small cleaning here and there
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Add patch found in Teltonika RUT9_R_00.07.01.4 GPL SDK download[1]
adding USB IDs of the MeigLink SLM750 to the relevant kernel drivers.
Newer versions of Teltonika's 2G/3G/4G RUT9XX WWAN router series come
with this kind of modem.
[1]: https://wiki.teltonika-networks.com/view/GPL
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
flowtable->net was initialized too late, and this could be triggered even
without hardware offload support on the device
Signed-off-by: Felix Fietkau <nbd@nbd.name>
The issue of EAP frames sent to group address (or the wrong address) has been
addressed in mac80211, so this hack is no longer needed
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This caches flows between MAC addresses on separate ports, including their VLAN
in order to bypass the normal bridge forwarding code.
In my test on MT7622, this reduces LAN->WLAN bridging CPU usage by 6-10%,
potentially even more on weaker platforms
Signed-off-by: Felix Fietkau <nbd@nbd.name>
The kernel configuration allows us to select a default qdisc. Let's do this for
5.10 (as 5.4 is on its way out) and get rid of the hacky patch we've been
carrying.
Acked-by: Jo-Philipp Wich <jo@mein.io>
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
This moves bootargs-append support patch from ipq40xx and ipq806x to
generic. This way we can append additional boot arguments from DTS instead
of only being able to overwrite them.
This is a preparation for kirkwood support of ipTIME NAS1.
Signed-off-by: Sungbo Eo <mans0n@gorani.run>
Preparation for supporting BMT on MT7621. Move source files to the files/
subdirectory in order to simplify maintenance
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Set policy bit to force read-only mode on uImage.FIT filesystem
sub-images mapped as block partitions by the FIT partition parser.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Add and enable a new kconfig knob to disable unprivileged eBPF by default.
Patches automatically rebased.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
* correctly set system side interface, the original patch was
errornous and there is a follow-up fix for it
* enable phy statistics for AQR112(+R/C) and ARQ412
(ethtool --phy-statistics ethX)
Tested, including phy-statistics, on
- IEI Puzzle M901 (AQR112, AQR112C, AQR112R)
- IEI Puzzle M902 (AQR113, AQR112R)
- Ubiquiti UniFi 6 LR (AQR112C)
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Backport qca8k new feature:
- Ageing configuration support
- Add 2 missing counter on qca8337
- Convert to regmap
- Standardize define and code with GENMASK AND BITFILED macro
- Add mdb add/del support
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Assuming cmdlinepart is only one level deep partition scheme and that
static partition are also defined in DTS, we can assign an of_node for
partition declared from bootargs.
The cmdlinepart parser is the first parser checked and if it does find
some partition declared in the bootargs, every other parser is ignored.
This means that the fixed-partition parser is ignored and an of_node for
the mtd is never assigned.
Fix this by searching a defined of_node using a similar fixed_partition
parsing logig, check if a partition is present with the same label, check
that it has the same offset and size and finally assign an of_node to the
mtd. The NVMEM can now find the of_node for the mtd and correctly works.
Fixes: abc17bf306 ("ath79: convert mtd-mac-address to nvmem implementation")
Tested-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Copy and refresh patch enabling AQR112 and AQR412 Ethernet PHY from
layerscape (5.4) target to generic (5.10) as AQR112 can also be found
on other targets as well.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
I had quite a hard time to understand what the change to net/core/dev.c
is supposed to do.
Simplify the change by returning NETDEV_TX_OK in case a eth_mangle_tx
callback was set but returned NULL instead of setting the return value
in the else branch.
Signed-off-by: Mathias Kresin <dev@kresin.me>
ar8216 switches have a hardware bug, which renders normal 802.1q support
unusable. Packet mangling is required to fix up the vlan for incoming
packets.
The patch was ommited at the time kernel 5.10 support was added but is
still required for ar8216 switches.
Signed-off-by: Mathias Kresin <dev@kresin.me>
Don't patch the kernel to expose the partition name in sysfs as it is
already exposed via 'uevent'.
All previous users have been converted to use 'uevent', so we can
safely drop the custom patch.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Deleted (upstreamed):
bcm27xx/patches-5.10/950-0145-xhci-add-quirk-for-host-controllers-that-don-t-updat.patch [1]
Manually rebased:
bcm27xx/patches-5.10/950-0355-xhci-quirks-add-link-TRB-quirk-for-VL805.patch
bcm53xx/patches-5.10/180-usb-xhci-add-support-for-performing-fake-doorbell.patch
Note: although automatically rebaseable, the last patch has been edited to avoid
conflicting bit definitions.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-5.10.y&id=b6f32897af190d4716412e156ee0abcc16e4f1e5
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
SATA_PMP option is no longer exposed when no SATA host driver is enabled
since upstream linux commit bd322af15ce9 ("ata: make SATA_PMP option
selectable only if any SATA host driver is enabled").
Commit 1bb3f593ee ("kirkwood: update config for kernel 5.10") manually
added CONFIG_SATA_PMP=y to config file, but the config will disappear for
every kernel_oldconfig refresh.
To prevent this, a small hack is added, which selects SATA_HOST
automatically when SATA_PMP is enabled. This patch can be dropped if
SATA_MV is ever re-added into kernel config file.
Tested-by: Pawel Dembicki <paweldembicki@gmail.com>
Signed-off-by: Sungbo Eo <mans0n@gorani.run>
[Move patch to generic/hack-5.10]
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
There is a longlasting warn about unused defined try_to_force_load.
Fix this by including this function only when actually used.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
The current patch produces the following error when CONFIG_DMABUF_HEAPS is
enabled:
drivers/built-in.a: member drivers/dma-buf/heaps in archive is not an object
Fixes: b10d604459 ("kernel: add linux 5.10 support")
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
When a target configuration has unser Kconfig symbols, the build will
fail when OpenWrt is compiled with V=s and stdin is connected to a tty.
In case OpenWrt is compiled without either of these preconditions, the
build will uscceed with the symbols in question being unset.
Modify the kernel configuration in a way it fails on unset symbols
regardless of the aformentioned preconditions.
Signed-off-by: David Bauer <mail@david-bauer.net>
kirkwood build broke due to missing include needed for ETH_ALEN.
Add patch (sent upstream as well) to address that.
Refresh patches for 5.4 and 5.10.
Fixes: 91a52f22a1 ("treewide: backport support for nvmem on non platform devices")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
In the current state, nvmem cells are only detected on platform device.
To quickly fix the problem, we register the affected problematic driver
with the of_platform but that is more an hack than a real solution.
Backport from net-next the required patch so that nvmem can work also
with non-platform devices and rework our current patch.
Drop the mediatek and dsa workaround and rework the ath10k patches.
Rework every driver that use the of_get_mac_address api.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
This reverts commit b309248730.
This commit could create a property without allocated memory, breaking
subsequent reads over a nodes property. Also, the mac-address-increment
was not applied when reading from nvmem.
Revert this commit for now, which breaks the label-mac-address logic.
Possibly, traversing the device-tree from the netdev side is easier
anyways.
Signed-off-by: David Bauer <mail@david-bauer.net>
Add the new CONFIG_BATTERY_RT5033 to the generic configuration, as reported by
Paul Blazejowski. Resort the kconfig while at it.
No deleted or manually refreshed patches.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
Latest binutils (2.37) exposed a long-standing bug. The kernel linking stage
would break at the SORTTAB step, due to the exception table having been
previously purged from vmlinux, as its section wasn't marked as unconditionally
kept. Fix thusly.
Additionally, the "#define ARM_MMU_DISCARD(x) KEEP(x)" change is bogus. It
would only apply to !CONFIG_MMU devices (which we don't support in OpenWrt), and
it would even break the build if referenced. Drop it.
While at it, rename the patch in order to make it obvious that it's
arm-specific.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
[Add same changes for kernel 5.4 too]
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Nvmem require the device node to be registered with the of_platform.
Register the device node so that nvmem can correctly find the dev and
correctly load the mac-addr stored in the nvmem cell declared in the dts.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Traversing the device-tree by referencing a network device to determine
a devices labe-mac does not work with the generic nvmem implementation,
as the userspace expects the MAC-address to be available as a
device-tree property.
The legacy mtd-mac-address implementation did create such a node. Do the
same when using the nvmem implementation to allow reading the MAC
address.
Fixes commit d284e6ef0f ("treewide: convert mtd-mac-address-increment*
to generic implementation")
Signed-off-by: David Bauer <mail@david-bauer.net>
Introduce new partition type 0x2e representing uImage.FIT and trigger
FIT partition parser on partitions having that type.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
ath79, lantiq, ipq40xx, ramips all use the OpenWrt-specific gpio-export
functionality. Consolidate the patch that adds it under hack-5.10 since
this logic is obviously not target-specific. For those who want to
disable it, unsetting CONFIG_GPIO_SYSFS symbol will disable this code.
Signed-off-by: Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
Check if gendisk is already up instead of trying to figure out parent
mtd device (which didn't work well).
Reported-by: DENG Qingfang <dqfext@gmail.com>
Reported-by: Koen Vandeputte <koen.vandeputte@citymesh.com>
Fixes: 2809d00007 ("kernel: support FIT partition parser on mtdblock devices")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Move 480-mtd-set-rootfs-to-be-root-dev.patch to hack-5.10 and fix assumption
about the block device index.
Fixes: 2809d00007 ("kernel: support FIT partition parser on mtdblock devices")
Acked-by: Daniel Golle <daniel@makrotopia.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Allow using the FIT block-device partition partition parser to work on
top of mtdblock devices.
This is more tricky than it sounds as it requires to reorganize the
procedure of registering mtdblock devices in order to avoid locking
troubles caused by the block partition parsers then trying to open
the mtdblock device for reading while locks are still being held.
Fix that by moving the adding of the disks after the entire device
has been probed when locks no longer need to be held.
Also address issue with mtdsuper surfaced when using sub-partitions
which prevented mounting JFFS2 using the /dev/mtdblock* device.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Use update_kernel to refresh all patches, required manual updates to:
610-netfilter_match_bypass_default_checks.patch
611-netfilter_match_bypass_default_table.patch
762-net-bridge-switchdev-Refactor-br_switchdev_fdb_notif.patch
764-net-bridge-switchdev-Send-FDB-notifications-for-host.patch
Run-tested: x86_64
Nothing screamed out but any funny business with linux bridging should
suspect this update first.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
The uImage.FIT partition parser used to squeeze in FIT partitions in
the range where partition editor tools (fdisk and such) expect the
regular partition. This is confusing people and tools when adding
additional partitions on top of the partition used for OpenWrt's
uImage.FIT.
Instead of squeezing in the additional partitions, rather start with
all uImage.FIT partitions at offset 64.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Straightforward refresh of patches using update_kernel.
Removed (reverse-applicable):
bmips/patches-5.10/010-v5.11-net-dsa-implement-a-central-TX-reallocation-procedur.patch
Run tested: x86_64 (apu2)
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Also add a new kconfig symbol (CONFIG_KCMP) to the generic config,
disabling the SYS_kcmp syscall (it was split from
CONFIG_CHECKPOINT_RESTORE, which is disabled by default, so the
previous behaviour is kept).
Removed (upstreamed) patches:
070-net-icmp-pass-zeroed-opts-from-icmp-v6-_ndo_send-bef.patch
081-wireguard-device-do-not-generate-ICMP-for-non-IP-pac.patch
082-wireguard-queueing-get-rid-of-per-peer-ring-buffers.patch
083-wireguard-kconfig-use-arm-chacha-even-with-no-neon.patch
830-v5.12-0002-usb-serial-option-update-interface-mapping-for-ZTE-P685M.patch
Manually rebased patches:
313-helios4-dts-status-led-alias.patch
104-powerpc-mpc85xx-change-P2020RDB-dts-file-for-OpenWRT.patch
Run tested:
ath79 (TL-WDR3600)
mvebu (Turris Omnia)
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
Introduce a magic GUID_PARTITION_LINUX_FIT_GUID to designate a GPT
partition to be interpreted by the FIT partition parser.
In that way, sub-partitions for (external-data) uImage.FIT stored
directly in a partition can be split, similar like we do for devices
with raw flash storage.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Allow for single (external-data) FIT image to hold kernel, dtb and
squashfs. In that way, the bootloader verifies the system integrity
including the rootfs, because what's the point of checking that the
hash of the kernel is correct if it won't boot in case of squashfs
being corrupted? Better allow bootloader to check everything needed
to make it at least up to failsafe mode. As a positive side effect
this change also makes the sysupgrade process on nand potentially
much easier as it is now.
In short: mkimage has a parameter '-E' which allows generating FIT
images with 'external' data rather than embedding the data into the
device-tree blob itself. In this way, the FIT structure itself remains
small and can be parsed easily (rather than having to page around
megabytes of image content). This patch makes use of that and adds
support for adding sub-images of type 'filesystem' which are used to
store the squashfs. Now U-Boot can verify the whole OS and the new
partition parsers added in the Linux kernel can detect the filesystem
sub-images, create partitions for them, and select the active rootfs
volume based on the configuration in FIT (passing configuration via
device tree could be implemented easily at a later stage).
This new FIT partition parser works for NOR flash (on top of mtdblock),
NAND flash (on top of ubiblock) as well as classic block devices
(ie. eMMC, SDcard, SATA, NVME, ...).
It could even be used to mount such FIT images via `losetup -P` on a
user PC if this patch gets included in Linux upstream one day ;)
Signed-off-by: John Crispin <john@phrozen.org>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Marvell mv88e6xxx switch series cannot perform MAC learning from
CPU-injected (FROM_CPU) DSA frames, which results in 2 issues.
- excessive flooding, due to the fact that DSA treats those addresses
as unknown
- the risk of stale routes, which can lead to temporary packet loss
Backport those patch series from netdev mailing list, which solve these
issues by adding and clearing static entries to the switch's FDB.
Add a hack patch to set default VID to 1 in port_fdb_{add,del}. Otherwise
the static entries will be added to the switch's private FDB if VLAN
filtering disabled, which will not work.
The switch may generate an "ATU violation" warning when a client moves
from the CPU port to a switch port because the static ATU entry added by
DSA core still points to the CPU port. DSA core will then clear the static
entry so it is not fatal. Disable the warning so it will not confuse users.
Link: https://lore.kernel.org/netdev/20210106095136.224739-1-olteanv@gmail.com/
Link: https://lore.kernel.org/netdev/20210116012515.3152-1-tobias@waldekranz.com/
Ref: https://gitlab.nic.cz/turris/turris-build/-/issues/165
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
This applies another patch from 5.4 to 5.10 as well:
de09355f74 ("kernel/hack-5.4: make UDP tunneling user-selectable")
UDP tunneling support isn't user-selectable, but it's required by WireGuard
which is, for the time being, an out-of-tree module. We currently work around
this issue by selecting an unrelated module which depends on UDP tunnelling
(VXLAN). This is inconvenient, as it implies this unrelated module needs to be
built-in when doing a monolithic build.
Fix this inconvenience by making UDP tunneling user-selectable in the kernel
configuration.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
