This reverts commit b7f9742da8.
There are several reports of regressions with this commit. Will be added
back once I've figured out and fixed the cause
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When hitting a timing window where ubus configuration calls are hitting hostapd
and wpa_supplicant simultaneously, they can deadlock waiting for each other.
Fix this by using a lock around the ubus calls.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Frequent crashes have been observed on MT7916 based platforms. While the
root of these crashes are currently unknown, they happen when decoding
rate information of connected STAs in AP mode. The rate-information is
associated with a band which is not available on the PHY.
Check for this condition in order to avoid crashing the whole system.
This patch should be removed once the roout cause has been found and
fixed.
Link: https://github.com/freifunk-gluon/gluon/issues/2980
Signed-off-by: David Bauer <mail@david-bauer.net>
Add modules.builtin.modinfo to the kernel package, to support presence
testing using module aliases and printing module details with 'modinfo'.
With related kmodloader changes this adds ~2 KB to compressed image sizes.
root@OpenWrt:/# modinfo unix
name: unix
filename: (builtin)
alias: net-pf-1
license: GPL
root@OpenWrt:/# modprobe net-pf-1 && echo SUCCESS || echo FAIL
SUCCESS
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Add modules.builtin to the kernel package for improved handling of loadable
and builtin modules. As with other distros, this allows 'modprobe <module>'
to consistently return success for both loaded/built-in modules, a useful
feature for presence-testing.
Given OpenWrt's few built-in modules, this change and related kmodloader
support add ~1 KB to the compressed image size.
Using sch_fq_codel (builtin) and sch_cake (loadable) for example:
root@OpenWrt:/# modprobe sch_fq_codel && echo SUCCESS || echo FAIL
SUCCESS
root@OpenWrt:/# modprobe sch_cake && echo SUCCESS || echo FAIL
SUCCESS
root@OpenWrt:/# rmmod sch_fq_codel
module is builtin
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
fast-xmit must only be enabled after the sta has been uploaded to the driver,
otherwise it could end up passing the not-yet-uploaded sta via drv_tx calls
to the driver, leading to potential crashes because of uninitialized drv_priv
data.
Add a missing sta->uploaded check and re-check fast xmit after inserting a sta.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
This fixes WARN_ONs when using AP_VLANs after station removal. The flush
call passed AP_VLAN vif to driver, but because these vifs are virtual and
not registered with drivers, we need to translate to the correct AP vif
first.
Fixes: openwrt#12420
Signed-off-by: Oldřich Jedlička <oldium.pro@gmail.com>
Detach of-mdio dependency from stmmac-core kmod to fix support for
x86_64 target. This target doesn't use OpenFirmware infrastructure and
stmmac-core for the dwmac-intel driver doesn't depends on it.
Add kmod-of-mdio to any other user of stmmac-core as it's not inherit
from stmmac-core anymore.
Fixes: #14209
Fixes: 4b4c940fbc ("x86: Add kmod-dwmac-intel")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Move pcs-xpcs kmod from armsr modules.mk to generic modules package.
Also add additional dependency to x86_64 as stmmac-core it's now used
by x86_64 target and depends on this package.
Fixes: 4b4c940fbc ("x86: Add kmod-dwmac-intel")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Might be a typo in drv->txtstamp function:
+ phy_rxtstamp(ndev->phydev, skb, 0);
to
+ phy_txtstamp(ndev->phydev, skb, 0);
Reviewed-by: Robert Marko <robimarko@gmail.com>
Signed-off-by: Kristian Skramstad <kristian+github@83.no>
Remove upstreamed patches:
* replace usage of the deprecated "pci-dma-compat.h" API (6c03b27)
* remove uaccess and get_fs calls from PCIe for Kenel >= 5.18 (1d0d08c)
Signed-off-by: Stefan Kalscheuer <stefan@stklcode.de>
Backport merged upstream patch that adds support for firmware loader
from NVMEM or attached filesystem for Aquantia PHYs.
Refresh all kernel patches affected by this change.
Also update the path for aquantia .ko that got moved to dedicated
directory upstream.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
ISISC is the QCA codename for their Atheros switch family including
AR237, QCA8337 etc.
Since we have qca8k support in OpenWrt, there is no need to have SSDK
support for these switches, and boards that also have external switches
can just use qca8k.
Disable QCA803x PHY support as well, since all of those are supportable
via at803x driver.
Signed-off-by: Robert Marko <robimarko@gmail.com>
PTP and swconfig support in SSDK require kernel modifications we dont need
nor we want to support for now, so move the PTP and swconfig disablement
into general build options as they are not ipq807x specific.
Signed-off-by: Robert Marko <robimarko@gmail.com>
Recent SSDK versions started also parsing the "SoC" variable to identify
the SoC along with the "CHIP_TYPE".
We are not passing "SoC" currently and this leads to components we dont
need like MHT (New 2.5G quad port switch) being compiled and then unused,
so lets just pass the "SoC" as well.
Signed-off-by: Robert Marko <robimarko@gmail.com>
The RTW88 PCI/USB driver uses the same firmware,
so add firmware dependencies.
Also CI report that:
Package kmod-rtw88-usb is missing dependencies for the following libraries:
usbcore.ko
This commit fixes it.
Fixes: 3538a19 ("mac80211: split rtw88 configuration for each supported chip")
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
Commit 947b44d ("ipq807x: fix wrong define for LAN and WAN ess mask")
started fixing wrong switch_lan_bmp that defined lan there weren't
actually present. This displayed a fragility in the malibu phy init code
in qca-ssdk.
Add patch to fix this. Also update each DTS with the new required
property if needed.
The new binding malibu_phy_start_addr is required with devices that
place the malibu first PHY referring port1 on a different PHY addres
than 0. The most common configuration is 0 but some device (for example
Qnap 301W) place the malibu PHY at an offset to address 16.
Refer to ipq8074-ess dtsi for extensive description on how to derive
this value.
Quoting the patch detailed description:
The usage of first_phy_addr is EXTREMELY FRAGILE and results
in dangerous results if the OEM (or anyone that by chance try to
implement things in a logical manner) deviates from the default values
from the "magical template".
To be in more details. With QSDK 12.4, some tweaks were done to improve
autoneg and now on every call of port status, the phydev is tried to
add. This resulted in the call and log spam of an error with ports that
are actually not present on the system with qsdk reporting phydev is
NULL. This itself is not an error and printing the error is correct.
What is actually an error from ages is setting generic bitmap reporting
presence of port that are actually not present. This is very common on
OEM where the switch_lan_bmp is always a variant of 0x1e (that on bitmap
results in PORT1 PORT2 PORT3 PORT4 present) or 0x3e (PORT1 PORT2 PORT3
PORT4 PORT5). Reality is that many device are used as AP with one LAN
port or one WAN port. (or even exotic configuration with PORT1 not
present and PORT2 PORT3 PORT4 present (Xiaomi 3600)
With this finding one can say... ok nice, then lets update the DT and
set the correct bitmap...
Again world is a bad place and reality is that this cause wonderful
regression in some case of by extreme luck the first ever connected
port working and the rest of the switch dead.
The problem has been bisected to all the device that doesn't have the
PORT1 declared in any of the bitmap.
With this prefaction in mind, on to the REAL problem.
malibu_phy_hw_init FOR SOME REASON, set a global variable first_phy_addr
to the first detected PHY addr that coincidentally is always PORT1.
PORT1 addr is 0x0. The entire code in malibu_phy use this variable to
derive the phy addrs in some function.
Declaring a bitmap where the PORT1 is missing (or worse PORT4 the only
one connected) result in first_phy_addr set to 1 or whatever phy addr is
detected first setting wrong value all over the init stage.
To fix this, introduce a new binding malibu_first_phy_addr to manually
declare the first phy that the malibu PHY driver should use and permit
to detach it from port bmp detection. The legacy detection is kept for
compatibility reason.
Fixes: #13945
Fixes: 947b44d9ae ("ipq807x: fix wrong define for LAN and WAN ess mask")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Tested-by: Robert Marko <robimarko@gmail.com> # Qnap 301W
Reviewed-by: Robert Marko <robimarko@gmail.com>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
This reverts commit 8cce00bc9d.
The confusion was real and this change cause regression on other
advanced devices that makes actual use of the first_phy_addr value.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Some local patches have been sent to upstream and they are slightly
different from the upstream version. So it's better to replace them
to avoid conflicts with the new mac80211 backport driver. The
different parts have been merged into patch 996.
This commit also includes some additional fixes:
* Fix watchdog function.
* Improve MT7620 register initialization.
* Introduce DMA busy watchdog for rt2800.
P.S.
Sometimes rt2800 series chips may fall into a DMA busy state. The
tx queues become very slow and the client cannot connect to the AP.
Usually, We can see a lot of hostapd warnings at this point:
'hostapd: IEEE 802.11: did not acknowledge authentication response'
The DMA busy watchdog can help the driver automatically recover
from this abnormal state. By the way, setting higer 'cell_density'
and disabling 'disassoc_low_ack' can significantly reduce the
probability of the DMA busy.
Signed-off-by: Shiji Yang <yangshiji66@qq.com>
Kernel 6.1 has introduced support for RTW8822BU network adapter, which
is an USB variant of the rtw8822b 802.11ac chipset family.
Build and install the corresponding module in the rtw88 package
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Current rtw88 build configuration builds modules for all chips supported by
rtw88 driver family. This brings the following issues:
- adding a chip with a different bus is not convenient (all chips currently
depends on PCI)
- some features requirements are not relevant for all chips in family (eg
802.11AC is enforced but RTW88-8723DE is only a 802.11b/g/n chip)
Remove those constraints/issues by adding one module build option per
supported chip, and add intermediate options to properly cascade
dependencies.
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
ath9k-htc USB-based adapters also support 5/10MHz channel bandwidth.
Move the code handling the features in debugfs to common-debug.c,
and create proper registration functions to use in debug.c and
htc_drv_debug.c, leaving only debugfs registration there.
While at that, refresh one patch that would conflict otherwise.
Tested on TP-Link Archer C7v2 (ath79) and TP-Link WN722Nv1 (AR9287)
and WN822Nv2 (AR7010+AR9287).
Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
ath9k-htc USB-based adapterssupport 5/10MHz channel bandwidth, the
same as standard ath9k ones.
Move the code handling the features in debugfs to common-debug.c,
and create proper registration functions to use in debug.c and
htc_drv_debug.c, leaving only debugfs registration there.
While at that, refresh one patch that would conflict otherwise.
Tested on TP-Link Archer C7v2 (ath79) and TP-Link WN722Nv1 (AR9287)
and WN822Nv2 (AR7010+AR9287).
Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
In-kernel driver for MCS7715 USB-serial bridge has a bool option,
enabling support for parallel port on that chip - which is tied to the
same kernel module. Enable it and select kmod-ppdev, as the image size
increase is minimal and the package isn't bundled in the images by
default.
Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
Adjust our local ath10k-ct patches to the change
from the -ct 6.2 version to 6.4.
This restores e.g. the LED functionality.
Fixes: 7d3651f1b9 ("ath10k-ct: switch to 6.4")
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Commit 947b44d9ae ("ipq807x: fix wrong define for LAN and WAN ess mask")
started fixing wrong switch_lan_bmp that defined lan there weren't
actually present. This displayed a fragility in the malibu phy init code
in qca-ssdk.
Add patch to fix this.
Quoting the patch detailed description:
I'm very confused by this and to me it's not clear the real usage of
this logic.
From what I can see the usage of this is EXTREMELY FRAGILE and results
in dangerous results if the OEM (or anyone that by chance try to
implement things in a logical manner) deviates from the default values
from the "magical template".
To be in more details. With QSDK 12.4, some tweaks were done to improve
autoneg and now on every call of port status, the phydev is tried to
add. This resulted in the call and log spam of an error with ports that
are actually not present on the system with qsdk reporting phydev is
NULL. This itself is not an error and printing the error is correct.
What is actually an error from ages is setting generic bitmap reporting
presence of port that are actually not present. This is very common on
OEM where the switch_lan_bmp is always a variant of 0x1e (that on bitmap
results in PORT1 PORT2 PORT3 PORT4 present) or 0x3e (PORT1 PORT2 PORT3
PORT4 PORT5). Reality is that many device are used as AP with one LAN
port or one WAN port. (or even exotic configuration with PORT1 not
present and PORT2 PORT3 PORT4 present (Xiaomi 3600)
With this finding one can say... ok nice, then lets update the DT and
set the correct bitmap...
Again world is a bad place and reality is that this cause wonderful
regression in some case of by extreme luck the first ever connected
port working and the rest of the switch dead.
The problem has been bisected to all the device that doesn't have the
PORT1 declared in any of the bitmap.
With this perfection in mind, on to the REAL problem.
malibu_phy_hw_init FOR SOME REASON, set a global variable first_phy_addr
to the first detected PHY addr that coincidentally is always PORT1.
PORT1 addr is 0x0. The entire code in malibu_phy use this variable to
derive the phy addrs in some function.
Declaring a bitmap where the PORT1 is missing (or worse PORT4 the only
one connected) result in first_phy_addr set to 1 or whatever phy addr is
detected first setting wrong value all over the init stage.
To fix this, just drop this variable and hardcode everything to assume
the first phy adrr is ALWAYS 0 and remove calculation and use define for
special case.
With the following change normal switch traffic is restored and ports
function is recovered.
Fixes: #13945
Fixes: 947b44d9ae ("ipq807x: fix wrong define for LAN and WAN ess mask")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
noscan option was changed to hostapd_noscan but the entry in
wpa_supplicant was never updated resulting in the noscan option actually
never set.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Update SSDK version to 12.4, this fixes weird SFP port link up/downs
while there is no SFP module plugged in.
Signed-off-by: Robert Marko <robimarko@gmail.com>
This fixes the download of the kernel 5.15 for the bpf-headers when
kernel 6.1 is build for the target.
Even if kernel 6.1 was selected for the target we still use kernel 5.15
for the bpf-headers. The download script tried to download the 5.15
kernel from the 6.x directory on kernel,org and this failed. Define
PKG_SOURCE_URL based on PKG_PATCHVER and not KERNEL_BASE like done in
kernel.mk.
Without this change it tries to download the kernel from this URL:
ttps://cdn.kernel.org/pub/linux/kernel/v6.x/linux-5.15.129.tar.xz
Fixes: #13190Fixes: #13671Fixes: #13814
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
ath10k-ct based on kernel 6.4 doesn't have a fix present in previous
kernel. Add patch that port the compilation error fix from previous
kernel in the new 6.4 kernel.
Fixes: 7d3651f1b9 ("ath10k-ct: switch to 6.4")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Switch to the latest version so we match as close as possible to
our own mac80211 version.
Run-time tested on hundreds of devices in the field for months now:
- qca988x (wave 1)
- qca4019 (wave 2)
Tested-by: Enrico Mioso <mrkiko.rs@gmail.com>
Signed-off-by: Koen Vandeputte <koen.vandeputte@citymesh.com>
Add a struct_group to around all members in struct ath_cycle_counters.
It can help the compiler detect the intended bounds of the memcpy() and
memset().
This patch fixes the following build warning:
In function 'fortify_memset_chk',
inlined from 'ath9k_ps_wakeup' at /home/db/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/backports-6.1.24/drivers/net/wireless/ath/ath9k/main.c:140:3:
./include/linux/fortify-string.h:314:25: error: call to '__write_overflow_field' declared with attribute warning: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror=attribute-warning]
314 | __write_overflow_field(p_size_field, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Shiji Yang <yangshiji66@qq.com>
Currently the zram default compressor choice is displayed whether or not
zram is activated. Since the default choice is lzo-rle, this adds a
false dependency on kmod-lib-lzo.
With this patch, the choice options appear only when activating zram.
Signed-off-by: Rani Hod <rani.hod@gmail.com>
Fix the building issue setting CC to KERNEL_CC in kernel.mk. The
kernel backports by default uses CC to compile kconf. A new patch is
added to mac80211 to compile kconf with host gcc.
Signed-off-by: Zeyu Dong <dzy201415@gmail.com>
[ refresh patches ]
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
CONFIG_USB_AUDIO is a "USB Audio Gadget" driver, not a usb device driver
CONFIG_USB_AUDIO is "USB Audio support" before linux 2.6
Signed-off-by: Liangbin Lian <jjm2473@gmail.com>
Expose the temperature sensors as hwmon, it improves the overall
user experience since on tiny boards the nvme can become a substantial
source of heat.
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
Reviewed-by: Robert Marko <robimarko@gmail.com>
The IXP4xx crypto module must be loaded after the rootfs is
up as it depends on loading some NPE microcode from the file
system.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Make use of new toolchain define. TOOLCHAIN_DIR should be used only for
toolchain related packages and for everything else TOOLCHAIN_ROOT_DIR
and other define should be used instead.
Switch to new entry where possible.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Some packages won't ever have something to patch as they normally
install files or are meta-packages.
For these special packages, disable QUILT refresh.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
To support the IXP42x platforms we need a kernel module
for the Epson R7301 RTC so we can load it as an optional
module.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Robert Marko <robimarko@gmail.com>
To support the IXP42x platforms we need a kernel module
for the X1205 RTC so we can load it as an optional module.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Robert Marko <robimarko@gmail.com>
The lib-chacha20 library is missing build rules for big endian
ARM, and since IXP4xx is big endian ARM we need those rules to
build for IXP4xx.
Suggested-by: Tomasz Maciej Nowak <tmn505@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Improve EEPROME load patches. Reorganize and rework them.
The current patch are bugged and with the case of MTD loading, leaks and
never free the EEPROM read values.
Also add support for loading EEPROM using NVMEM cells.
As a cleanup, change the binding to swap EEPROM read from mtd to
ralink,eeprom-swap and generilize it.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Synchronize the ath11k backports with the current ath-next tree.
All of the changes are various bugfixes, there is no new major feature.
Notable bugfixes are:
* WCN6855 board name fixes
* One MSI vector booting is working again
This is rather important for most of the older platforms.
* DFS CAC state in virtual interfaces was fixed
* TX power during CAC reporting
Signed-off-by: Robert Marko <robimarko@gmail.com>
DFS CAC time export is required for backport of a ath11k fix so lets
backport the required cfg80211 upstream commit as well.
Signed-off-by: Robert Marko <robimarko@gmail.com>
The PKG_MIRROR_HASH is wrong, fix it.
Found and fixed using this command:
make package download check FIXUP=1
Fixes: c123e4f053 ("rtl8812au-ct: bump to fix kernel 6.1 compile")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The mt76 driver usually reads the eeprom on the mtd partition at dts.
For emmc device we need to use caldata_extract script to read the
eeprom. However, the default eeprom file breaks the caldata script
execution, so remove it.
Signed-off-by: Chukun Pan <amadeus@jmu.edu.cn>
Package and add dependencies to drm-amdgpu and drm-radeon for 6.1
* package acpi-video, drm-display-helper and drm-buddy
* add acpi-video, drm-display-helper and drm-buddy as dependencies for
drm-amdgpu
* add acpi-video and drm-display-helper as dependencies for drm-radeon
Signed-off-by: Pascal Coudurier <coudu@gmx.com>
Synchronize the ath11k backports with the current ath-next tree.
All of the changes are various bugfixes, there is no new major feature.
Signed-off-by: Robert Marko <robimarko@gmail.com>
Commit ed3725e15a154ebebf44e0c34806c57525483f92
("wifi: ath11k: Fix qmi_msg_handler data structure initialization")
has been present upstream since 6.1.2 but it seems Quilt refreshed it
wrongly so it appeared like a completely different patch.
Commit 7c15430822e71e90203d87e6d0cfe83fa058b0dc
("wifi: ath11k: allow system suspend to survive ath11k")
has been present upstream since 6.1.16 but somehow quilt still happily
applied it.
So, drop both of them.
Signed-off-by: Robert Marko <robimarko@gmail.com>
General support for 9P is desirable on platforms without virtualization.
This decouples the 9P general networking support for use on more platforms.
Signed-off-by: Daniel Maslowski <info@orangecms.org>
MAC address and interface name assigned by mac80211.sh depend on the order in
which interfaces are brought up. This order changes when interfaces get added
or removed, which can cause unnecessary reload churn.
One part of the fix it making MAC address allocation more dynamic in both
wpa_supplicant and hostapd, by ignoring the provided MAC address using
the next available one, whenever the config does not explicitly specify one.
The other part is making use of support for renaming netdevs at runtime and
preserving the MAC address for renamed netdevs.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
When a device has more than one reserved mac address, they can be used for
virtual interfaces without the local bit in the first byte
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Mediatek EIP93 Crypto engine is a crypto accelerator which
is available in the Mediatek MT7621 SoC.
Signed-off-by: Aviana Cruz <gwencroft@proton.me>
Co-authored-by: Richard van Schagen <vschagen@icloud.com>
Co-authored-by: Chukun Pan <amadeus@jmu.edu.cn>
When switching from a STA-only configuration to AP+STA on the same phy, the
STA was previously restarted in order to notify hostapd of the new frequency,
which might not match the AP configuration.
Fix the STA restart by querying the operating frequency from within hostapd
when bringing up the AP.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
It seems that this was not functioning properly and was likely completely unused.
Keeping this out of tree also introduced some annoying churn when updating, because
of the iw nl80211.h sync patch.
If this is needed, it will be reintroduced when/if it is added upstream
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Arm platforms with the right hardware blocks (such as
GICv3.0+ interrupt controller and SMMU/IOMMU) are
able to use vfio-pci to pass through PCI devices
to a VM.
Signed-off-by: Mathew McBride <matt@traverse.com.au>
The vfio module only exposes the enable_unsafe_noiommu_mode parameter
if CONFIG_VFIO_NOIOMMU is enabled. When it isn't, the module
will complain about an unknown parameter:
vfio: unknown parameter 'enable_unsafe_noiommu_mode' ignored
As CONFIG_VFIO_NOIOMMU is disabled by the module package,
we can remove the module loading parameter.
Signed-off-by: Mathew McBride <matt@traverse.com.au>
