Enabling KERNEL_TRANSPARENT_HUGEPAGE exposes 2 missing symbols:
* CONFIG_READ_ONLY_THP_FOR_FS
* TRANSPARENT_HUGEPAGE_ALWAYS
* TRANSPARENT_HUGEPAGE_MADVISE
The first one was added in 5.4, and is marked experimental there so just
disable it in the generic config.
For the latter two, we should not force the user to use either of them,
so add them as build-configurable kernel options.
Fixes: d1a8217d87 ("kernel: clean-up build-configurable kernel config symbols")
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
This symbol is exposed on ARM64 with EFI enabled in the kernel config.
Currently this happens only on ipq807x, but as there might be new ARM64
targets with EFI in the future it is better to add the symbol to the
generic config.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Jo-Philipp Wich <jo@mein.io>
Kernel v5.1 included an eBPF JIT for MIPS32 kernels, but problems were
discovered [1] and the changes later reverted in kernel v5.5 with commits:
* f8fffebdea75 ("MIPS: BPF: Disable MIPS32 eBPF JIT")
* 36366e367ee9 ("MIPS: BPF: Restore MIPS32 cBPF JIT")
Only the first of these was backported to LTS kernel 5.4, leaving cBPF
programs without a JIT and introducing a performance regression for any
such users e.g. libpcap, tcpdump, etc.
Restore cBPF performance by backporting the second commit above:
* 070-v5.5-MIPS-BPF-Restore-MIPS32-cBPF-JIT.patch
[1] https://lore.kernel.org/bpf/20191205182318.2761605-1-paulburton@kernel.org/
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Don't explicitely disable options in target/linux/generic/config-* if
they are already controlled in config/Config-kernel.in.
Add a bunch of new symbols and prepare defaults for using only unified
hierarchy (ie. cgroup2). Update symbol dependencies while at it
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
AR8327 datasheet[1] calls the register at address 0x0010
"Power-on Strapping Register". As it has nothing to do with "strip",
let's rename it to "POWER_ON_STRAP" to make it easier to grasp.
[1] https://lafibre.info/images/doc/201106_spec_AR8327.pdf
Signed-off-by: Sungbo Eo <mans0n@gorani.run>
update_kernel.sh refreshed all patches, no human interaction was needed
Build system: x86_64
Run-tested: Netgear R7800 (ipq806x)
Signed-off-by: John Audia <graysky@archlinux.us>
This patch adds cpu frequency setting for AR724x and AR7161.
It is unknown if all the AR71xx devices support the same value range,
so support is only enabled for AR7161 at this stage.
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
The current code would return an error on ath79 when the SoC isn't known
to the driver. Return the raw hex value instead, as happens with non
supported targets.
Also return the correctly incremented value from
sc_tag_cpufreq_ath79_arraysize()
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
RouterBOOT cpu frequency settings are letter-indexed. Follow the
same logic for the sysfs interface.
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
The routine would only accurately print out the first word.
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
Fixes: 5ecf7d96 ("generic: routerboot sysfs platform driver")
Fixes:
- CVE-2020-10757
The "mtd: rawnand: Pass a nand_chip object to nand_release()" commit was
backported which needed some adaptations to other code.
Run tested: ath79
Build tested: ath79
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Fixes:
- CVE-2020-10757
The "mtd: rawnand: Pass a nand_chip object to nand_release()" commit was
backported which needed some adaptations to other code.
Build tested: ramips
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This resolves a hazard between a mtc0 and a mfc0 instruction after
activating the DSP support. Without this fix the CPU could use the old
value again and the DSP support would not be active.
Fixes: FS#2928, FS#2899
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Tested-by: Stefan Lippers-Hollmann <s.l-h@gmx.de> [ath79/tl-wdr3600; ath79/tl-wdr4300]
Tested-by: David Bauer <mail@david-bauer.net> [ocedo_koala/ocedo_raccoon]
From upstream:
b8392808eb3f sch_cake: add RFC 8622 LE PHB support to CAKE diffserv handling
3f608f0c4136 sch_cake: fix a few style nits
8c95eca0bb8c sch_cake: don't call diffserv parsing code when it is not needed
9208d2863ac6 sch_cake: don't try to reallocate or unshare skb unconditionally
From netdev not yet accepted:
sch_cake: fix IP protocol handling in the presence of VLAN tags
The VLAN tag handling is actually wider than just cake so upstream are
working out how to fix it generically. We fix it here just for cake.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
This was backported to 4.19 and I clearly expected it to land in 5.4 but
it didn't (5.5) so backport it to 5.4 for consistency.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
With this change the LED subsystem is abused in the kernel to switch the
simswap. This change will be reverted, so we could use again the gpio
subsystem.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Backport upstream patches that make drivers/spi/spi-rb4xx.c device tree
aware, plus a null pointer fix.
Signed-off-by: Christopher Hill <ch6574@gmail.com>
USB support built in kernel isn't required for booting.
This patch move USB support to modules.
CONFIG_USB_STORAGE is removed because it can be found in kmod-usb-storage
and all usb support for this targed is moved to kmods.
Tested-by: Cezary Jackiewicz <cezary@eko.one.pl> [nsa310]
Tested-by: Alberto Bursi <bobafetthotmail@gmail.com> [nsa325, ea4500]
Reviewed-by: Sungbo Eo <mans0n@gorani.run>
Signed-off-by: Pawel Dembicki <paweldembicki@gmail.com>
[keep CONFIG_SCSI]
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
All devices are using nand images. Built-in SATA modules are not needed
anymore.
Signed-off-by: Pawel Dembicki <paweldembicki@gmail.com>
[only move CONFIG_SATA_MV]
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Tested-by: Pawel Dembicki <paweldembicki@gmail.com> [pogo v4]
The interface mode number of RGMII_33V is 7 on RTL8367, but it's 9 on
RTL8367B.
the external interface modes for RTL8367 are follows:
- 0, Disabled
- 1, RGMII
- 2, MII_MAC
- 3, MII_PHY
- 4, TMII_MAC
- 5, TMII_PHY
- 6, GMII
- 7, RGMII_33V
the external interface modes for RTL8367B are follows:
- 0, Disabled
- 1, RGMII
- 2, MII_MAC
- 3, MII_PHY
- 4, TMII_MAC
- 5, TMII_PHY
- 6, GMII
- 7, RMII_MAC
- 8, RMII_PHY
- 9, RGMII_33V
But the driver in U-Boot of RT-N56U GPL tar blocks using RGMII_33V (9)
mode and it seems to be unsupported on RTL8367B, so drop it from
switch-case in rtl8367b_extif_set_mode.
ref (RTL8367):
- TL-WR2453ND v1
ref (RTL8367B):
- ASUS RT-N56U
- TP-Link Archer C2 v1
Signed-off-by: INAGAKI Hiroshi <musashino.open@gmail.com>
While the other fq-based qdiscs take advantage of skb->hash and doesn't
recompute it if it is already set, sch_cake does not.
This was a deliberate choice because sch_cake hashes various parts of the
packet header to support its advanced flow isolation modes. However,
foregoing the use of skb->hash entirely loses a few important benefits:
- When skb->hash is set by hardware, a few CPU cycles can be saved by not
hashing again in software.
- Tunnel encapsulations will generally preserve the value of skb->hash from
before the encapsulation, which allows flow-based qdiscs to distinguish
between flows even though the outer packet header no longer has flow
information.
It turns out that we can preserve these desirable properties in many cases,
while still supporting the advanced flow isolation properties of sch_cake.
This patch does so by reusing the skb->hash value as the flow_hash part of
the hashing procedure in cake_hash() only in the following conditions:
- If the skb->hash is marked as covering the flow headers (skb->l4_hash is
set)
AND
- NAT header rewriting is either disabled, or did not change any values
used for hashing. The latter is important to match local-origin packets
such as those of a tunnel endpoint.
The immediate motivation for fixing this was the recent patch to WireGuard
to preserve the skb->hash on encapsulation. As such, this is also what I
tested against; with this patch, added latency under load for competing
flows drops from ~8 ms to sub-1ms on an RRUL test over a WireGuard tunnel
going through a virtual link shaped to 1Gbps using sch_cake. This matches
the results we saw with a similar setup using sch_fq_codel when testing the
WireGuard patch.
Fixes: 046f6fd5daef ("sched: Add Common Applications Kept Enhanced (cake) qdisc")
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
When a client moves from a DSA user port to a software port in a bridge,
it cannot reach any other clients that connected to the DSA user ports.
That is because SA learning on the CPU port is disabled, so the switch
ignores the client's frames from the CPU port and still thinks it is at
the user port.
Fix it by enabling SA learning on the CPU port.
To prevent the switch from learning from flooding frames from the CPU
port, set skb->offload_fwd_mark to 1 for unicast and broadcast frames,
and let the switch flood them instead of trapping to the CPU port.
Multicast frames still need to be trapped to the CPU port for snooping,
so set the SA_DIS bit of the MTK tag to 1 when transmitting those frames
to disable SA learning.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
Currently enabling VLAN filtering blocks all traffic in the bridge
immediately. That is because DSA ignores all VLAN setup when VLAN
filtering is disabled, and when it is enabled, there is no VLAN entry
in the VLAN table, causing all traffic to be blocked.
Add patches to allow VLAN setup even if VLAN filtering is disabled.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
Currently, setting a bridge's self PVID to other value and deleting
the default VID 1 renders untagged ports of that VLAN unable to talk to
the CPU port:
bridge vlan add dev br0 vid 2 pvid untagged self
bridge vlan del dev br0 vid 1 self
bridge vlan add dev sw0p0 vid 2 pvid untagged
bridge vlan del dev sw0p0 vid 1
# br0 cannot send untagged frames out of sw0p0 anymore
That is because the CPU port is set to security mode and its PVID is
still 1, and untagged frames are dropped due to VLAN member violation.
Set the CPU port to fallback mode so untagged frames can pass through.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
Remove dependencies on core kernel headers in host tools used to build perf,
which break on any non-linux system
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This is useful when booting OpenWrt from ramdisks in order to have both
images partitions defined.
Furthermore, instead of always using img2 for the inactive image, let's use
img1 or img2 accordingly.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Kernel config option LEDS_TRIGGER_IDE_DISK was renamed in kernel 4.8 to
CONFIG_LEDS_TRIGGER_DISK in upstream commit eb25cb9956cc ("leds: convert
IDE trigger to common disk trigger").
Removing it as it should be added only on targets which has usage for
this trigger.
Signed-off-by: Thomas Albers <thomas.gameiro@googlemail.com>
[commit description facelift]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
When firmware is flashed, cferam.000 extension is renamed to the next number.
When booting, CFE scans the NAND and picks the partition with the highest
cferam extension and ignores the other one.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
The only Sercomm WFI user has been migrated to a dedicated firmware parser.
Keep support for no cferam partition based on a boolean DT property.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Sercomm uses 2 bootflag partitions and boots the firmware with the highest
bootflag. Support splitting the firmware partition while keeping support for
unsplitted layout.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
This commit introduces support for R/W access to the CPU frequency
setting of routerboot on ath79 hardware.
On unsupported hardware, the sysfs attribute will expose the raw tag
value (read-only) to help with reverse engineering its meaning.
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
This routine will be shared between hard and soft config drivers.
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
This driver exposes the data encoded in the "soft_config" flash segment
of MikroTik RouterBOARDs devices. It presents the data in a sysfs folder
named "soft_config" through a set of human-and-machine-parseable
attributes. Changes can be discarded by writing 0 to the 'commit'
attribute, or they can be committed to flash storage by writing 1.
This driver does not reuse any of the existing code previously found in
the "rbcfg" utility and makes this utility obsolete by providing a clean
sysfs interface.
Like "rbcfg", this driver requires 4K_SECTORS support since the flash
partition in which these parameters are stored is typically 4KB in size.
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
This routine will be shared between hard and soft config drivers.
Also use scnprintf() instead of snprintf().
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
For the sake of strictly typed code, add a missing const qualifier.
Add a missing return value in error path.
Check the return value of mtd_read(), for good measure.
Also demote the error printks of failed sysfs file creation to warn
level since they are not fatal in the init() sequence.
Finally, add a note regarding PAGE_SIZE and clarify a comment.
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
The depends and select should apply to the sysfs driver, not the meta
config.
Tested-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
Between 4.19 and 5.4, the kernel moved the partition parsers into
the parsers subdirectory. This led to some necessary rebasing of
our local patches for parsers, which partially has been performed
without caring about where the code was inserted.
This commit tries to adjust our local patches so that parsers are
inserted at the "proper" positions with respect to alphabetic sorting
(if possible). Thus, the commit is cosmetic.
While this might look useless now, it will make life easier when
adding other parsers in the future or for rebasing on kernel changes.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
These trailing whitespaces were reported during kernel patch refresh.
While at it, harmonize a few indents as well.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
This patch has been backported to stable kernel 5.4 already.
Remove our local patch explicitly now, as by applying the patch
(or refreshing) the relevant code is actually added a second time.
Refresh remaining patches as well.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
Add i2c-pxa updates queued for v5.8, which add bus recovery to this
driver; this is needed for the uDPU platform.
Signed-off-by: Russell King <linux@armlinux.org.uk>
Backport the GPIO emulated open drain output fix from v5.5, which is
required for the i2c-pxa backport.
Signed-off-by: Russell King <linux@armlinux.org.uk>
This drops the obsolete version switches for non-supported kernels
from local drivers in generic target.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
31e99fe3da which introduced this code was unfortunately untested.
This commit fixes a number of issues and works around the fact that in
this particular scheme, the LZO payload may be padded at the end which
will trigger a harmless lzo decompression error.
This commit also disambiguates the debug printks.
Tested-by: Robert Marko <robimarko@gmail.com>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
Fixes: 31e99fe3da ("generic: platform/mikrotik: support LZOR encoding")
An e-mail response from MikroTik contained a minimal overview
on hardware option bits which mentioned a currently unknown bit.
While not being too detailed what it does, add it for documentation purposes.
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Fixes a build error seen when applying the kernel patches.
Applying patch generic/435-mtd-add-routerbootpart-parser-config.patch
patching file drivers/mtd/parsers/Kconfig
Hunk #1 FAILED at 160.
1 out of 1 hunk FAILED -- rejects in file drivers/mtd/parsers/Kconfig
patching file drivers/mtd/parsers/Makefile
Hunk #1 FAILED at 10.
1 out of 1 hunk FAILED -- rejects in file drivers/mtd/parsers/Makefile
This was missed as 5.4 is currently set as testing kernel
while the main kernel being used is 4.19
Fixes: 2976e423dc ("generic: routerboot partition build bits")
Signed-off-by: Koen Vandeputte <koen.vandeputte@ncentric.com>
Some newer MikroTik RouterBOARD devices use a new encoding scheme
for their WLAN calibration data. This patch provides support for
decoding this new scheme.
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
This driver exposes the data encoded in the "hard_config" flash segment
of MikroTik RouterBOARDs devices. It presents the data in a sysfs folder
named "hard_config". The WLAN calibration data is available on demand via
the 'wlan_data' sysfs file in that folder.
This driver permanently allocates a chunk of RAM as large as the
"hard_config" MTD partition (typically 4KB), although it is technically
possible to operate entirely from the MTD device without using a local
buffer (except when requesting WLAN calibration data), at the cost of a
performance penalty.
This driver does not reuse any of the existing code previously found in
routerboot.c.
This driver has been successfully tested on BE (ath79) and LE (ipq40xx
and ramips) hardware.
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Tested-by: Baptiste Jonglez <git@bitsofnetworks.org>
Tested-by: Tobias Schramm <t.schramm@manjaro.org>
Tested-by: Christopher Hill <ch6574@gmail.com>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
This driver provides an OF MTD parser to properly assign the RouterBoot
partitions on the flash. This parser builds from the "fixed-partitions"
one (see ofpart.c), but it can handle dynamic partitions as found on
routerboot devices.
The parent node must contain the following:
compatible = "mikrotik,routerboot-partitions";
#address-cells = <1>;
#size-cells = <1>;
Children routerbootpart DTS nodes are defined as follows:
For fixed partitions
node-name@unit-address {
reg = <prop-encoded-array>;
label = <string>;
read-only;
lock;
};
All properties but reg are optional.
For dynamic partitions:
node-name {
size = <prop-encoded-array>;
label = <string>;
read-only;
lock;
};
size property is mandatory unless the next partition is a fixed one or
a "well-known" one (matched from the strings defined below) in which case
it can be omitted or set to 0; other properties are optional.
By default dynamic partitions are appended after the preceding one, except
for "well-known" ones which are automatically located on flash.
Well-known partitions (matched via label or node-name):
- "hard_config"
- "soft_config"
- "dtb_config"
This parser requires the DTS to list partitions in ascending order as
expected on the MTD device.
This parser has been successfully tested on BE (ath79) and LE (ipq40xx
and ramips) hardware.
Tested-by: Baptiste Jonglez <git@bitsofnetworks.org>
Tested-by: Roger Pueyo Centelles <roger.pueyo@guifi.net>
Tested-by: Tobias Schramm <t.schramm@manjaro.org>
Tested-by: Christopher Hill <ch6574@gmail.com>
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
ar8229 and ar8236 don't allow unknown unicast/multicast frames and
broadcast frames to be flooded to cpu port. This isn't desired behavior
for swconfig as we treat it as a standalone switch.
Current code doesn't enable unicast frame flooding for ar8229 and uses
wrong setup for ar8236. This commit fixes both of them by enabling port
0 flooding for all unknown frames.
Fixes: FS#2848
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
Upstream commit dda9f4b9ca ("f2fs: fix to skip verifying block address
for non-regular inode").
On 4.14, attempting to perform operations on a non-regular inode
residing on an f2fs filesystem, such rm-ing a device node, would fail
and lead to a warning / call trace in dmesg. This fix was already
applied to other kernels upstream - including 4.19, from which the patch
was taken.
More info at https://bugzilla.kernel.org/show_bug.cgi?id=202495.
Signed-off-by: Matt Merhar <mattmerhar@protonmail.com>
This patch is needed for clause 37 1000Base-X encoding used in many SFP modules.
Signed-off-by: DENG Qingfang <dengqf6@mail2.sysu.edu.cn>
Signed-off-by: René van Dorst <opensource@vdorst.com>
Between 4.19 and 5.4, mtd parsers have been moved to "parsers"
subdirectory. Based on this, the selection of myloader.o in OpenWrt
was also moved to that subdirectory, while the Kconfig and our local
myloader.c file remained in /drivers/mtd.
This resulted in linking errors like the following (on ath25@5.4):
make[8]: *** No rule to make target 'drivers/mtd/parsers/myloader.o', ...
needed by 'drivers/mtd/parsers/built-in.a'. Stop.
make[7]: *** [scripts/Makefile.build:500: drivers/mtd/parsers] Error 2
make[6]: *** [scripts/Makefile.build:500: drivers/mtd] Error 2
Since myloader.c is not too big, this patch moves it to the kernel patches,
allowing to adjust the path for kernel 5.4 and keeping Makefiles and
file paths better in sync.
Other patches have been refreshed accordingly.
Signed-off-by: Adrian Schmutzler <freifunk@adrianschmutzler.de>
There are 2 different chips (w25q256fv and w25q256jv) that share
the same JEDEC ID. Only w25q256jv fully supports 4-byte opcodes.
Use SFDP header version to differentiate between them.
Fixes broken reboot on 8devices Habanero since f0f35fdac
Signed-off-by: Mantas Pucka <mantas@8devices.com>
From kernel 4.20 msm-gpio driver is broken and cause the
malfunction of the buttons on every ipq806x target.
Add a patch to fix this.
Tested-by: Hannu Nyman <hannu.nyman@iki.fi>
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Instead of resetting flash to 3B address on remove hook, this
implementation only enters 4B mode when needed, which prevents more
unexpected reboot stuck. This implementation makes it only break when
a kernel panic happens during flash operation on 16M+ areas.
Also silent broken-flash-reset warning. We are not dealing with vendors
and it's unpleasant for users to see that unnecessary and long WARN_ON
print.
Signed-off-by: Chuanhong Guo <gch981213@gmail.com>
The conditional check introduced by this patch may trigger a NULL pointer
dereference in case the result of dev_net() is NULL.
Since the purpose of this patch is neither sufficiently explained and since
this patch apparently has never been submitted upstream despite it being in
the pending-* patch directory, I propose to drop it without replacement.
If the performance implications of dropping this patch are found to be
significiant, it should be reintroduced with proper description and
benchmark results.
Ref: https://bugs.openwrt.org/index.php?do=details&task_id=2943
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
