savedscp is a method of storing the DSCP of an ip packet into conntrack
mark. In combination with a suitable tc filter action (conndscp but may
end up being integrated into connmark) DSCP values are able to be stored
on egress and restored on ingress across links that otherwise alter or
bleach DSCP.
This is useful for qdiscs such as CAKE which are able to shape according
to policies based on DSCP.
Ingress classification is traditionally a challenging task since
iptables rules haven't yet run and tc filter/eBPF programs are pre-NAT
lookups, hence are unable to see internal IPv4 addresses as used on the
typical home masquerading gateway.
The ingress problem is solved by the tc filter, but the tc people didn't
like the idea of tc setting conntrack mark values, though they are ok
with reading conntrack values and hence restoring DSCP from conntrack
marks.
x_tables CONNMARK with the new savedscp action solves the problem of
storing the DSCP to the conntrack mark.
It accepts 2 parameters. The mark is a 32bit value with usually one 1
bit set. This bit is set when savedscp saves the DSCP to the mark.
This is useful to implement a 'one shot'
iptables based classification where the 'complicated' iptables rules are
only run once to classify the connection on initial (egress) packet and
subsequent packets are all marked/restored with the same DSCP. A mark
of zero disables the setting of a status bit/s.
The mask is a 32bit value of at least 6 contiguous bits and represents
the area where the DSCP will be stored.
e.g.
iptables -A QOS_MARK_eth0 -t mangle -j CONNMARK --savedscp-mark 0xfc000000/0x01000000
Would store the DSCP in the top 6 bits of the 32bit mark field, and use
the LSB of the top byte as the 'DSCP has been stored' marker.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Commit fcb41decf6 ("config: enable some useful features on
!SMALL_FLASH devices") enabled netns, which in turn lead to the crash in
the flow offload target.
When the flow offloading framework intends to delete a flow from the
hardware table, it is necessary to retrieve the namespace from
nf_flowtable->ft_net. However, no one ever wrote the namespace into
nf_flowtable->ft_net in advance. So the framework will mistakenly use a
NULL namespace to execute dev_get_by_index_rcu(net, ifindex), leading to
the kernel panic.
Ref: FS#2321
Fixes: fcb41decf6 ("config: enable some useful features on !SMALL_FLASH devices")
Tested-by: Simon Tretter <simon@mediaarchitectu.re>
Signed-off-by: HsiuWen Yen <y.hsiuwen@gmail.com>
[merged patch into offload patch, fix for 4.19, SOB fix, commit subj/msg touches]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Memory is allocated with devm_kzalloc() on every page program
and leaks until device is closed (which never happens).
Convert to kzalloc() and handle error paths manually.
Signed-off-by: Mantas Pucka <mantas@8devices.com>
This patch removes 202-reduce_module_size.patch which is causing missing
debug symbols in kernel modules, leading to unusable
kernel-debug.tar.bz2 on all platforms, making debugging of release
kernel crashes difficult.
Cc: Felix Fietkau <nbd@nbd.name>
Acked-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
This patch uses nfct_help() to detect whether an established connection
needs conntrack helper instead of using test_bit(IPS_HELPER_BIT,
&ct->status).
The reason for this modification is that IPS_HELPER_BIT is only set when
the conntrack helper is attached by explicit CT target.
However, in the case that a device enables conntrack helper via the other
ways (e.g., command "echo 1 > /proc/sys/net/netfilter/nf_conntrack_helper")
, the status of IPS_HELPER_BIT will not present any change. That means the
IPS_HELPER_BIT might lose the checking ability in the context.
Signed-off-by: HsiuWen Yen <y.hsiuwen@gmail.com>
This reduces the needed modifications to the mainline Linux kernel and
also makes the regmap package work with an out of tree kernel which
does not have these modifications.
The regmap-core is only added when it is really build as a module.
The regmap-core is normally bool so it cannot be built as a module in an
unmodified kernel. When it is selected by on other kernel module it will
always be selected as build in and it also does not show up in
$(LINUX_DIR)/modules.builtin as it is not supposed to be a kernel module.
When it is not in $(LINUX_DIR)/modules.builtin the build system expects
it to be built as a .ko file.
Just check if the module is really there and only add it in that case.
This splits the regmap package into multiple packages, one for each bus type.
This way only the bus maps which are really needed have to be added.
This also splits the I2C, SPI and MMIO regmap into separate packages to not
require all these subsystems to build them, on an unmodified upstream kernel
this also causes problems in some situations.
Signed-off-by: Hauke Mehrtens <hauke.mehrtens@intel.com>
Refresh all patches
Remove upstream patch:
backport-4.14/424-v4.20-net-dsa-fix-88e6060-roaming.patch
Minor tweak to generic/hack-4.14/902-debloat_proc.patch to cleanly apply
after upstream changes.
Tested-on: ath79
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
The b53 driver was added as a dsa driver into the mainline Linux kernel,
but we still use the swconfig based driver. The header file b53.h is
used by both drivers, but the swconfig one needs an extra member, add
this one in a patch to not overwrite the version shipped with the
mainline kernel.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Refreshed all patches.
Altered patches:
- 666-Add-support-for-MAP-E-FMRs-mesh-mode.patch
New symbol for arm targets:
- HARDEN_BRANCH_PREDICTOR
Compile-tested on: ar71xx, cns3xxx, imx6
Runtime-tested on: ar71xx, cns3xxx, imx6
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Commit b7265c59ab ("kernel: backport a series of netfilter cleanup
patches to 4.14") added patch 302-netfilter-nf_tables_inet-don-t-use-
multihook-infrast.patch. That patch switches the netfilter core in the
kernel to use the new native NFPROTO_INET support. Unfortunately, the
new native NFPROTO_INET support does not exist in 4.14 and was not
backported along with this patchset. As such, nftables inet tables never
see any traffic.
As an example the following nft counter rule should increment for every
packet coming into the box, but never will:
nft add table inet foo
nft add chain inet foo bar { type filter hook input priority 0\; }
nft add rule inet foo bar counter
This commit pulls in the required backport patches to add the new
native NFPROTO_INET support, and thus restore nftables inet table
functionality.
Tested on Turris Omnia (mvebu)
Fixes: b7265c59ab ("kernel: backport a series of netfilter cleanup ...")
Signed-off-by: Brett Mastbergen <bmastbergen@untangle.com>
By default the RPS delegation will happen by masking the last few bits of
skb->hash. This patch adds an inermediate hash bucket that maps the masked
hash to a RPS core. This makes RPS results much more deterministic on SMP
systems.
Signed-off-by: John Crispin <john@phrozen.org>
Jonas Gorski commented on the previous patch:
|This is actually the wrong fix and papers over an issue in one of our
|local patches.
|
|We intentionally allow regmap to be built as a module, see
|
|/target/linux/generic/hack-4.14/259-regmap_dynamic.patch
|[...]
|[The regulator code] optionally supports regmap thanks to the stubs
|provided if regmap is disabled - which breaks if you compile regmap
|as a module.
In order to mitigate this issue, this patch reverts the previous patch
and replaces the existing IS_ENABLED(CONFIG_REGMAP) with
IS_REACHABLE(CONFIG_REGMAP). This solves this particular issue as the
regulator code will now automatically fallback to the regmap stubs in
case the kmod-regmap module is enabled, but nothing else sets
CONFIG_REGMAP=y.
Note: There's still a potential issue that this patch doesn't solve:
If someone ever wants to make a OpenWrt kernel package for a
regulator module that requires the REGMAP feature for a target that
doesn't set CONFIG_REGMAP=y but has CONFIG_REGULATOR=y, the resulting
kmod-regulator-xyz package will not work on the target.
Luckily, there aren't any in-tree OpenWrt kernel module packages for
regulators at the moment. On the bright side: regmap is a critical
part nowadays and all new and upcoming architectures require it by
default. This will likely only ever be a problem for legacy targets
and devices that cannot afford to enable REGMAP.
Cc: Jonas Gorski <jonas.gorski@gmail.com>
Cc: John Crispin <john@phrozen.org>
Fixes: d00913d121 ("kernel: modules: fix kmod-regmap")
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Newer kernels have a patch that implements compatible functionality
directly. Adjust the attribute of our own patch in preparation for
dropping it later
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Rereshed all patches
Reworked patches to match upstream:
335-v4.16-netfilter-nf_tables-add-single-table-list-for-all-fa.patch
Compile-tested on: cns3xxx, imx6, x86_64
Runtime-tested on: cns3xxx, imx6, x86_64
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
The sender domain has a DMARC Reject/Quarantine policy which disallows
sending mailing list messages using the original "From" header.
To mitigate this problem, the original message has been wrapped
automatically by the mailing list software.
Refresh patches.
Remove patch that can be reverse applied:
mvebu/patches-4.14/530-ATA-ahci_mvebu-enable-stop_engine-override.patch
mvebu/patches-4.14/531-ATA-ahci_mvebu-pmp-stop-errata-226.patch
Update patch that no longer applied:
ipq806x/patches-4.14/0035-clk-mux-Split-out-register-accessors-for-reuse.patch
Compiled-tested-for: lantiq, ramips
Run-tested-on: lantiq BT hh5a, ramips MIR3g
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Tested-by: Michael Yartys <michael.yartys@protonmail.com>
Tested-by: Rosen Penev <rosenp@gmail.com>
