NAPI poll() function may be passed a budget value of zero, i.e. during
netpoll, which isn't NAPI context.
Therefore, napi_consume_skb() must be given budget value instead of
!force to truly discern netpoll-like scenarios.
https://lore.kernel.org/netdev/20220707141056.2644-1-liew.s.piaw@gmail.com/t/#m470f5c20225e76fb08c44d6cfa2f1b739ffaaea4
Signed-off-by: Sieng-Piaw Liew <liew.s.piaw@gmail.com>
[improve code format]
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Use existing rx processed count to track against budget, thereby making
budget decrement operation redundant.
rx_desc_count can be calculated outside the rx loop, making the loop a
bit smaller.
Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
napi_build_skb() reuses NAPI skbuff_head cache in order to save some
cycles on freeing/allocating skbuff_heads on every new rx or completed
tx.
Use napi_consume_skb() to feed the cache with skbuff_heads of completed
tx so it's never empty.
Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
Check if we're in NAPI context when refilling rx. Normally we're almost
always running in NAPI context. Dispatch to napi_alloc_frag() directly
instead of relying on netdev_alloc_frag() which does the same but with
the overhead of local_bh_disable/enable.
Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
[improve code format]
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
We can increase the efficiency of rx path by using buffers to receive
packets then build SKBs around them just before passing into the network
stack. In contrast, preallocating SKBs too early reduces CPU cache
efficiency.
Performance is slightly increased but the changes allow more
potential optimizations.
Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
[improve code format]
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
bmips is using Broadcom B53 DSA driver which means the ethernet driver
must compensate for 6 bytes tags from the internal switch.
This means using NET_IP_ALIGN actually misaligns the skb, lowering
performance significantly. Therefore napi_alloc_skb() which uses
NET_IP_ALIGN is changed to netdev_alloc_skb().
Performance in iperf3 is increased from ~47Mbps to 52Mbps.
Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
Endianness depends on CPU architecture. CONFIG_CPU_(BIG/LITTLE)_ENDIAN should
be enabled on target or subtarget based on SoC architecture.
Fixes warning:
$ make kernel_oldconfig CONFIG_TARGET=subtarget
...
.config:1008:warning: override: CPU_LITTLE_ENDIAN changes choice state
....
Summary:
- ARC - only the CONFIG_CPU_BIG_ENDIAN symbol is defined for this architeture.
If it is disabled then the processor operates in LITTLE_ENDIAN mode (default),
- ARM32 - CONFIG_CPU_LITTLE_ENDIAN symbol available since kernel 5.19. This
option should be enabled after OpenWRT moves to kernel 6.x. After refreshing
the kernel, the symbol disappears,
- ARM64 - enabled CONFIG_CPU_LITTLE_ENDIAN,
- MIPS - enabled relevant symbols,
- POWERPC - enabled CONFIG_CPU_BIG_ENDIAN,
- UML - Symbols are not defined for this architecture,
- X86 - always little endian. Symbols are not defined for this architecture.
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
All targets expect the malta target already activate the CONFIG_GPIOLIB
option. Move it to generic kernel configuration and also activate it for
malta.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
This is now built-in, enable so it won't propagate on target configs.
Link: https://lkml.org/lkml/2022/1/3/168
Fixes: 79e7a2552e ("kernel: bump 5.15 to 5.15.44")
Fixes: 0ca9367069 ("kernel: bump 5.10 to 5.10.119")
Signed-off-by: Tomasz Maciej Nowak <tmn505@gmail.com>
(Link to Kernel's commit taht made it built-in,
CRYPTO_LIB_BLAKE2S[_ARM|_X86] as it's selectable, 5.10 backport)
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
ucidef_set_bridge_device is needed for DGND3700v2 network config since VLAN 1
must be used for the switch to be correctly configured.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
A devent amount of patches have been upstreamed, so maintaining linux 5.10 on
this target makes no sense.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Remove redundant target-level entries, noting that these settings will be
configured from "Kernel build options" of Kconfig.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
[remove from new configs introduced after patch submission]
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Each of
- CRYPTO_AEAD2
- CRYPTO_AEAD
- CRYPTO_GF128MUL
- CRYPTO_GHASH
- CRYPTO_HASH2
- CRYPTO_HASH
- CRYPTO_MANAGER2
- CRYPTO_MANAGER
- CRYPTO_NULL2
either directly required for mac80211 crypto support, or directly
selected by such options. Support for the mac80211 crypto was enabled in
the generic config since c7182123b9 ("kernel: make cryptoapi support
needed by mac80211 built-in"). So move the above options from the target
configs to the generic config to make it clear why do we need them.
CC: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Both CLANG_VERSION and LLD_VERISON are autogenerated runtime
configuration options, so add them to the kernel configuration filter
and remove from generic and per-target configs to keep configs clean.
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Deactivate all the symbols of the B53 DSA driver in the generic kernel
configuration. Multiple targets are now using this drivers and they
only need some of the options.
This fixes the bcm4908 build which didn't deactivate all of the options.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
CONFIG_RCU_{NEED_SEGCBLIST,STALL_COMMON} are set basically everywhere. Move them
to the generic kconfigs. And resort the generic kconfigs while at it.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
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>
MAC address retrieval was switched to more generic upstream (5.13) NVMEM
based solution in commit 32adbfc789 ("bmips: convert mtd-mac-address
to nvmem implementation"), but NVMEM subsystem wasn't enabled in the
kernel, so fix it now.
References: https://github.com/openwrt/openwrt/pull/4041#issuecomment-883322801
Fixes: 32adbfc789 ("bmips: convert mtd-mac-address to nvmem implementation")
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Define nvmem-cells and convert mtd-mac-address to nvmem implementation.
The conversion is done with an automated script.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Some targets select HZ=100, others HZ=250. There's no reason to select a higher
timer frequency (and 100 Hz are available in every architecture), so change all
targets to 100 Hz.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
For the targets which enable ubifs, these symbols are already part of the
generic kconfigs. Drop them from the target kconfigs.
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
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>
After backporting upstream ehci overcurrent patches we need to use spurious-oc
instead of ignore-oc.
Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
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>
Straightforward refresh of patches using update_kernel.
Run tested: x86_64 (apu2)
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
