This commit contains a series of fixes for DMA. The burst length
patch significantly improves Ethernet performance. Patches were
tested on the xRX200 and xRX330.
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
Update Telekom Speedport W921V firmware download URL.
Contained TAPI and VDSL firmware blobs are identical.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Enabling KERNEL_FTRACE exposes the HIST_TRIGGERS triggers symbol. Add a
build option for it to fix build failures with KERNEL_FTRACE enabled.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
19aae94 [build: avoid rebuilds of unset VARIANT packages] builds
packages defined without a VARIANT only once, using the first VARIANT
defined in the Makefile.
This caused problems with wpa-cli, as it is only built for variants that
include supplicant support, and the first VARIANT defined may not build
it.
The same happens to hostapd-utils, which is not built for
supplicant-only variants.
To circumvent this, set VARIANT=* for both packages so that they get
built for every defined variant. This should not cause spurious
rebuilds, since tey are not a dependency of any other package defined in
this Makefile.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
This allows a package to be rebuilt for all different VARIANTs.
To do so, set VARIANT=*.
The wpa-cli package is not getting built after changes made by
19aae94 [build: avoid rebuilds of unset VARIANT packages], because
wpa-cli is only built when compiling a variant that includes
supplicant support, and the first selected variant may not build it.
Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
Fix mac address increment patch. Permit to overflow to the next
byte and correctly calculate the incremented mac.
Reported-by: Chen Minqiang <ptpt52@gmail.com>
Fixes: d284e6ef0f ("treewide: convert mtd-mac-address-increment* to generic implementation")
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Backport of Ansuel Smith's "net: dsa: qca8k: make sure PAD0 MAC06
exchange is disabled", to ensure mac06 is disabled even if enabled by
the bootloader.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Signed-off-by: Matthew Hagan <mnhagan88@gmail.com>
The devicetree property mac-address is expected to be set by the
bootloader and has priority over the nvmem supplied one.
Drop the mac-address address property from the dtsi files, to let the
mac address from nvmem-cells get used.
Signed-off-by: Mathias Kresin <dev@kresin.me>
On lantiq a lot of stuff expects to be loaded to and executed at
0x80002000, including our own second stage bootloader.
For all build u-boots, the initial stack pointer is at 0x80008000. After
loading data to 0x80002000, every further stack operation corrupts the
loaded code.
Set the initial stack pointer to 0x80002000, to not overwrite code
loaded in memory. A stack of 0x2000 bytes has been proven as enough in
all done tests.
Signed-off-by: Mathias Kresin <dev@kresin.me>
On danube we only have 0x6800 bytes of usable SRAM. Everything behind
can't be written to and a SPL u-boot locks up during boot.
Since it's a hard to debug issue and took me more than two years to fix
it, I consider it worth to include fix albeit SPL u-boots are not build
in OpenWrt.
I faced the issue while trying to shrink the u-boot to 64K since some
boards only have an u-boot partition of that size from the days
ifx-uboot was used.
Signed-off-by: Mathias Kresin <dev@kresin.me>
Reviewed-by: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
With gcc10 the variables are placed more tightly to each other, which
uncovers a long existing bug in the lantiq DMA code. It can be observed
when using tftpboot with the filename parameter, which gets reset during
the tftpboot execution.
NetRxPackets[] points to cache line size aligned addresses. In
ltq_eth_rx_packet_align() the address NetRxPackets[] points to is
increased by LTQ_ETH_IP_ALIGN and the resulting not cache aligned
address is used further on. While doing so, the length/size is never
updated.
The "not cache aligned address" + len/size for a cache aligned address
is passed to invalidate_dcache_range(). Hence, invalidate_dcache_range()
invalidates the next 32 bit as well, which flashes the BootFile variable
as well.
variable BootFile is at address: 0x83ffe12c
NetRxPackets[] points to 0x83ffdb20 (len is 0x600)
data points to: 0x83ffdb22 (len is 0x600)
ltq_dma_dcache_inv: 0x83ffdb22 (for len 0x600)
invalidate_dcache_range: 0x83ffdb20 to 0x83ffe120 (size: 32)
invalidate_dcache_range: 0x83ffdb20 to 0x83ffdb40 (Bootfile: a.bin)
...
invalidate_dcache_range: 0x83ffe100 to 0x83ffe120 (Bootfile: a.bin)
invalidate_dcache_range: 0x83ffe120 to 0x83ffe140 (Bootfile: )
In ltq_dma_tx_map() and ltq_dma_rx_map() the start address passed to
ltq_dma_dcache_wb_inv() is incorrect. By considering the offset, the
start address passed to flush_dcache_range() is always aligned to 32, 64
or 128 bytes dependent on configured DMA burst size.
Fixes: FS#4113
Signed-off-by: Mathias Kresin <dev@kresin.me>
At least since gcc 7.3.0 (OpenWrt 18.06) lwr/lwl are used in the
assembly of LzmaProps_Decode. While the decission made by the compiler
looks perfect fine, it triggers some obscure hang on lantiq danube-s
v1.5 with MX29LV640EB NOR flash chips.
Only if the offset 1 is used, the hang can be observed. Using any other
offset works fine:
lwl s0,0(a1) - s0 == 0x6d000080
lwl s0,1(a1) - hangs
lwl s0,2(a1) - s0 == 0x0080xxxx
lwl s0,3(a1) - s0 == 0x80xxxxxx
It isn't clear whether it is a limitation of the flash chip, the EBU or
something else.
Force 8bit reads to prevent gcc optimizing the read with lwr/lwl
instructions.
Signed-off-by: Mathias Kresin <dev@kresin.me>
This patch fixes a blunder of mine. The include needed
for LED_COLOR_ID_BLUE property is missing.
This caused the builds to fail with:
|Error: arch/arm/boot/dts/qcom-ipq4019-r619ac.dtsi:91.13-14 syntax error
|FATAL ERROR: Unable to parse input tree
Fixes: 12d33d388c ("ipq40xx: add support for P&W R619AC (aka G-DOCK 2.0)")
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
The Zyxel NBG6617 already uses lzma to compress the kernel.
A local build with every module enabled (either as =Y or =M)
ended produced a 3058 KiB kernel (the kernel partition is 4MiB).
It booted just fine, let's reenable the device.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
P&W R619AC is a IPQ4019 Dual-Band AC1200 router.
It is made by P&W (p2w-tech.com) known as P&W R619AC
but marketed and sold more popularly as G-DOCK 2.0.
Specification:
* SOC: Qualcomm Atheros IPQ4019 (717 MHz)
* RAM: 512 MiB
* Flash: 16 MiB (NOR) + 128 MiB (NAND)
* Ethernet: 5 x 10/100/1000 (4 x LAN, 1 x WAN)
* Wireless:
- 2.4 GHz b/g/n Qualcomm Atheros IPQ4019
- 5 GHz a/n/ac Qualcomm Atheros IPQ4019
* USB: 1 x USB 3.0
* LED: 4 x LAN, 1 x WAN, 2 x WiFi, 1 x Power (All Blue LED)
* Input: 1 x reset
* 1 x MicroSD card slot
* Serial console: 115200bps, pinheader J2 on PCB
* Power: DC 12V 2A
* 1 x Unpopulated mPCIe Slot (see below how to connect it)
* 1 x Unpopulated Sim Card Slot
Installation:
1. Access to tty console via UART serial
2. Enter failsafe mode and mount rootfs
<https://openwrt.org/docs/guide-user/troubleshooting/failsafe_and_factory_reset>
3. Edit inittab to enable shell on tty console
`sed -i 's/#ttyM/ttyM/' /etc/inittab`
4. Reboot and upload `-nand-factory.bin` to the router (using wget)
5. Use `sysupgrade` command to install
Another installation method is to hijack the upgrade server domain
of stock firmware, because it's using insecure http.
This commit is based on @LGA1150(at GitHub)'s work
<a4932c8d5a>
With some changes:
1. Added `qpic_bam` node in dts. I don't know much about this,
but I observed other dtses have this node.
2. Removed `ldo` node under `sd_0_pinmux`, because `ldo` cause SD card not
working. This fix is from
<51143b4c75>
3. Removed the 32MB NOR variant.
4. Removed `cd-gpios` in `sdhci` node, because it's reported that it makes
wlan2g led light up.
5. Added ethphy led config in dts.
6. Changed nand partition label from `rootfs` to `ubi`.
About the 128MiB variant: The stock bootloader sets size of nand to 64MiB.
But most of this devices have 128MiB nand. If you want to use all 128MiB,
you need to modify the `MIBIB` data of bootloader. More details can be
found on github:
<https://github.com/openwrt/openwrt/pull/3691#issuecomment-818770060>
For instructions on how to flash the MIBIB partition from u-boot console:
<https://github.com/openwrt/openwrt/pull/3691#issuecomment-819138232>
About the Mini PCIe slot: (from "ygleg")
"The REFCLK signals on the Mini PCIe slot is not connected on
this board out of the box. If you want to use the Mini PCIe slot
on the board, you need to (preferably) solder two 0402 resistors:
R436 (REFCLK+) and R444 (REFCLK-)..."
This and much more information is provoided in the github comment:
<https://github.com/openwrt/openwrt/pull/3691#issuecomment-968054670>
Signed-off-by: Richard Yu <yurichard3839@gmail.com>
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
[Added comment about MIBIB+128 MiB variant. Added commit
message section about pcie slot. Renamed gpio-leds' subnodes
and added color, function+enum properties.]
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Since version 2.28, git has a config option init.defaultBranch to set the name
of the first branch created with git init. The env script expects this name to
be "master". This commit sets the initial branch name to "master"
instead of using the git configured one.
Signed-off-by: Arne Zachlod <arne@nerdkeller.org>
eb0a3ee fw4.uc: Do not quote port ranges
c5a8e3e tests: adapt test to new ICMP print logic
Also start using $(AUTORELEASE)
Signed-off-by: Paul Spooren <mail@aparcar.org>
Simply reading /proc/*/stat as a space-separated string will not work
as the process name may itself contain spaces. Hence we must match on
the '(' and ')' characters around the process name and can then handle
the remaining string as space-separated values.
This fixes shell error messages which have been popping up the console
due to spaces in process names being interpreted as field separators.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
The nl80211_set_wiphy() function was changed between kernel 5.11 and
5.12 to take the rdev->wiphy lock which should be freed at the end
again. The 500-mac80211_configure_antenna_gain.patch added some code
which just returned in some cases without unlocking. This resulted in a
deadlock with brcmfmac.
This patch fixes this by also jumping to the out label in case we want
to leave the function.
This fixes a hanging system when brcmfmac is in use. I do not know why
we do not see this with other driver.
The kernel returns very useful debug details when setting these OpenWrt
configuration options:
CONFIG_KERNEL_DETECT_HUNG_TASK=y
CONFIG_KERNEL_PROVE_LOCKING=y
Fixes: FS#4122
Fixes: b96c2569ac ("mac80211: Update to version 5.12.19-1")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
from https://git.kernel.org/pub/scm/network/iproute2/iproute2.git
changes since 5.14.0:
ad3a118f rdma: Fix SRQ resource tracking information json
7a235a10 man: devlink-port: fix pfnum for devlink port add
229eaba5 uapi: pickup fix for xfrm ABI breakage
a500c5ac lib/bpf: fix map-in-map creation without prepopulation
7c032cac man: devlink-port: remove extra .br
04ee8e6f man: devlink-port: fix style
14802d84 man: devlink-port: fix the devlink port add synopsis
897772a7 cmd: use spaces instead of tabs for usage indentation
e7a98a96 mptcp: unbreak JSON endpoint list
2f5825cb lib: bpf_legacy: fix bpffs mount when /sys/fs/bpf exists
d756c08a tc/f_flower: fix port range parsing
92e32f77 uapi: updates from 5.15-rc1
e7e0e2ce iptuntap: fix multi-queue flag display
deef844b man: ip-link: remove double of
a3272b93 configure: restore backward compatibility
ceba5930 tree-wide: fix some typos found by Lintian
7a705242 ip: remove leftovers from IPX and DECnet
8ab1834e uapi: update headers from 5.15 merge
6d0d35ba ip/bond: add lacp active support
926ad641 Update kernel headers
c730bd0b ip/tunnel: always print all known attributes
df8912ed ipioam6: use print_nl instead of print_null
7e7270bb tc/skbmod: Introduce SKBMOD_F_ECN option
86c596ed IOAM man8
2d83c710 New IOAM6 encap type for routes
f0b3808a Add, show, link, remove IOAM namespaces and schemas
acbdef93 Import ioam6 uapi headers
2d6fa30b Update kernel headers
508ad89c ipneigh: add support to print brief output of neigh cache in tabular format
* update patch 170-ip_tiny.patch to accomodate ioam.
Signed-off-by: Russell Senior <russell@personaltelco.net>
0750f2b4d329 README: dnsmasq integration is complete
8e48d0b0cbba bpf: add initial support for splitting map dscp value into ingress and egress
bfc2cafe2a8c map: add support for defining aliases
Signed-off-by: Felix Fietkau <nbd@nbd.name>
We can't use booleans, since we're not including stdbool.h. Use integers
instead.
Fixes: 0b79e7c01e ("dnsmasq: generate the dns object name dynamically")
Signed-off-by: Rui Salvaterra <rsalvaterra@gmail.com>
Similar to mt7623, also no longer use 'blockdev' and stop relying on
in-kernel partition parsers. Instead, strip off all metadata using
'fwtool' while writing the firmware image and scrape the number of
blocks written from 'dd', then use that block offset to stash the
configuration backup.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Re-reading the partition table doesn't work reliably, it fails if
anything on the device is still in use and it's not trivial to prevent
every possible case of a block device still being in use somehow.
Therefore, instead of relying on the in-kernel partition parser to know
where to write the configuration backup, use OpenWrt's format-agnostic
fwtool to strip off all metadata from the image and count its blocks
while writing. In that way we can know where to write the config backup
without needing the kernel to parse the MBR and FIT structures.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
dnsmasq may call hotplug.dhcp, hotplug.neigh and hotplug.tftp.
Only the first two callees were listed in the ACL, so add missing
hotplug.tftp.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
9d1431e jail: allow passing environment variable to procd jailed process
Fixes dnsmasq in ujail which needs USER_SCRIPT env variable to be
passed to jailed process.
Reported-by: Bastian Bittorf <bb@npl.de>
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
When __ieee80211_select_queue is called, skb->cb has not been cleared yet,
which means that info->control.flags can contain garbage.
In some cases this leads to IEEE80211_TX_CTRL_DONT_REORDER being set, causing
packets marked for other queues to randomly end up in BE instead.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
1fdf2a3 Fix kernel panic due to device deletion (#29)
e7b48d1 add the mutex lock for create/delete/config/insert nat46 devices to fix nat46 module crash issues. (#28)
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
A set of tags can be specified for --dhcp-host option to restrict the
assignment to the requests which match all the tags.
Example usage:
config vendorclass
option networkid 'udhcp'
option vendorclass 'udhcp'
config host
option mac '*:*:*:*:*:*'
list match_tag 'switch.10'
list match_tag 'udhcp'
option ip '192.168.25.10'
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Now that wildcard matching is supported, this makes it easier for packages
to supply their own qosify rules
Signed-off-by: Felix Fietkau <nbd@nbd.name>
737970946bc0 map: default to fnmatch matching for dns patterns. support regex via leading /
b56b112e62e2 ubus: fix crash caused by missing static keyword
3a420e272c18 qosify: support wildcards in classifier filenames
Signed-off-by: Felix Fietkau <nbd@nbd.name>
It isn't used anymore so there is no need to hack CPU port. Upstream
(DSA-based) b53 also supports all switch ports just fine.
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
