Make the call deferred instead of blocking to avoid deadlock issues
Fixes: 3df9322771 ("hostapd: make ubus calls to wpa_supplicant asynchronous")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
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>
This fixes a deadlock issue where depending on the setup order, hostapd and
wpa_supplicant could end up waiting for each other
Reported-by: Michael-cy Lee (李峻宇) <Michael-cy.Lee@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Modems which are using qmi do not reply on the 1st sync but they do
on subsequent. Sometimes uqmi is hanging - even when using an early
dummy access to unlock the modem. To always guarantee a proper
initialisation, running or hanging uqmi processes must be stopped
before. All uqmi calls have now a timeout option -t to avoid hanging.
Signed-off-by: Uwe Niethammer <uwe@dr-niethammer.de>
Use postinst script to reload service instead of uci-defaults hack. It's
possible thanks to recent base-files change that executes postinst after
uci-defaults.
This fixes support for uhttpd customizations. It's possible (again) to
adjust uhttpd config with custom uci-defaults before it gets started.
Cc: Hauke Mehrtens <hauke@hauke-m.de>
Fixes: d25d281fd6 ("uhttpd: Reload config after uhttpd-mod-ubus was added")
Ref: b799dd3c70 ("base-files: execute package's "postinst" after executing uci-defaults")
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Increasing the receive window size improves throughout on higher-latency
links such as WAN connections. The current default of 24KB caps out at
around 500 KB/s.
Increasing the receive buffer to 256KB increases the throughput to at
least 11 MB/s.
Signed-off-by: David Bauer <mail@david-bauer.net>
730b4656e6b1 netifd: fix undefined va_list value which can cause crashes
c59457f69709 device: Log error message if device initialization failed
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Prior to this commit, "localuse" (which enables local resolving through
dnsmsasq) was off by "default". That default was in turn overridden when
"noresolv" was unset (which itself is the default for "noresolv") *and*
"resolvfile" was "/tmp/resolv.conf.d/resolv.conf.auto" (also the default
for this parameter).
In other words, the "default" unset value for "localuse" would only be
ever used in specific *non-default* configurations.
However, the problem with that logic is that a user who wants to ignore
their ISP-provided resolvers by setting "noresolv" to true ends up with
a device that will *only use* said resolvers for local DNS queries,
serving clients' queries via dnsmasq (which now ignores the ISP
resolvers). This can lead to confusion and break random setups as the
DNS lookup performed on clients behalf can differ in their replies from
DNS lookups performed locally on the router.
Furthermore, "localuse" is not configurable through Luci, contrary to
the other two involved settings, adding further confusion for the end
user.
To work around this situation, the logic that sets "localuse" is
inverted: "localuse" now defaults to on by default, and IFF "noresolv"
is unset (default) AND "resolvfile" is changed from default THEN
"localuse" gets turned back off, allowing for more sensible behaviour.
"localuse" value set in config/dhcp still overrides the logic in all
cases, as it did already.
Signed-off-by: Thibaut VARÈNE <hacks@slashdirt.org>
8f2806a37fe1 system-linux: set master early on apply settings
e3fc2b0026a5 system-linux: skip refreshing MAC on master change if custom MAC
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Update to the latest upstream release to include recent improvements and
bugfixes. Also refresh local patches.
Link: https://github.com/libbpf/bpftool/releases/tag/v7.3.0
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
If the dnsmasq process forks to handle TCP connections, it closes the ubus
context. But instead of changing the daemon wide pointer to NULL, only the
local variable was adjusted - and this portion of the code was even dropped
(dead store) by some optimizing compilers.
It makes more sense to change the daemon->ubus pointer because various
functions are already checking it for NULL. It is also the behavior which
ubus_destroy() implements.
Fixes: d8b33dad0b ("dnsmasq: add support for monitoring and modifying dns lookup results via ubus")
Signed-off-by: Sven Eckelmann <sven@narfation.org>
This is not activated by default and must be explicitly enabled via ubus
It supports reporting log messages and netlink packets
Signed-off-by: Felix Fietkau <nbd@nbd.name>
The new rewritten ipcalc.sh understands 3 notations:
ipaddr/prefix ...
ipaddr/dotted-netmask ...
ipaddr dotted-netmask ...
meaning that the previous 4th non-standard notation of "ipaddr prefix"
will be dropped, alas that's the notation that dnsmasq currently uses.
This change has us using the first notation which is the most common.
This behavior came in as
eda27e8382
a long time ago.
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
eee02ccca8c8 device: add support to configure eee
bb28f6a291d9 wireless: fix sign comparison warning
35facc8306f5 wireless: fix premature removal of hotplug devices due to down state
Signed-off-by: Felix Fietkau <nbd@nbd.name>
841b05fbb91e system-linux: fix compilation error if IFLA_DSA_MASTER is not supported
5c9ecc1ff74f system-linux: make system_if_get_master_ifindex static
2dc7f450f3a2 system-linux: add option to configure DSA conduit device
838f815db5ef system-linux: add support for configurable GRO option
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Currently for 802.1s only, for wifi 2.4GHz in g/n mode, 40MHz is never
permitted.
This is probably due to the complexity of setting periodic check for the
intolerant bit. When noscan option is set, we ignore the presence of the
intoleran bit in near AP, so we can enable 40MHz and ignore any complex
logic for checking.
Fixes: #13112
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Also channel 7 for 2.4GHz can be set to HT40PLUS. Permit this and add it
to the list of the channels.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
noscan option for mesh was broken and actually never applied.
This is caused by a typo where ssid->noscan value is check instead of
conf->noscan resulting in the logic swapped and broken.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
383753dd65ae device/bridge: support passing extra vlans in the device_set_state call
b6e75eafc1af device: send notifications for device events via ubus
cab415c7aefd bridge: add auth-required bridge members with auth_status=0 if vlan is enabled
827a02f0343c bridge: add support for configuring vlans for auth=1,auth_status=false
40ed7363caf2 device: fix build error on 32 bit systems
516ab774cc16 system-linux: fix race condition on bringing up wireless devices
Signed-off-by: Felix Fietkau <nbd@nbd.name>
4101dd4 fw4: perform strict validation of zone and set names
a923c88 fw4: pass zone to templates whenever possible
597dc90 fw4: add support for zone log_limit
1874050 fw4: add log_limit to rules and redirects
19a8caf ruleset: dispatch ct states using verdict map
a5553da ruleset: reduce ksoftirqd load by refering to looopback by numeric id
de3483c tests: adjust zone log limit testcases
7392792 ruleset: do not emit redundant drop invalid rules
698a533 ruleset: apply egress MSS fixup later to apply final MTU before wire
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
It's already pulled in from /etc/rc.common.
Fixes: #13758
Fixes: 6b23836071 ("package: avoid the use of eval to parse ipcalc.sh output")
Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
Upgrading wpa_supplicant from 2.9 to 2.10 breaks broadcom-wl/ath11k
based adapters. The reason for it is hostapd tries to install additional
IEs for scanning while the driver does not support this.
The kernel indicates the maximum number of bytes for additional scan IEs
using the NL80211_ATTR_MAX_SCAN_IE_LEN attribute. Save this value and
only add additional scan IEs in case the driver can accommodate these
additional IEs.
Bug: http://lists.infradead.org/pipermail/hostap/2022-January/040178.html
Bug-Debian: https://bugs.debian.org/1004524
Bug-ArchLinux: https://bugs.archlinux.org/task/73495
Upstream-Status: Changes Requested [https://patchwork.ozlabs.org/project/hostap/patch/20220130192200.10883-1-mail@david-bauer.net]
Reported-by: Étienne Morice <neon.emorice@mail.com>
Tested-by: Étienne Morice <neon.emorice@mail.com>
Signed-off-by: David Bauer <mail@david-bauer.net>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
The code for hostapd-mbedtls did not work when used for OWE association.
When handling association requests, the buffer offsets and length
assumptions were incorrect, leading to never calculating the y point,
thus denying association.
Also when crafting the association response, the buffer contained the
trailing key-type.
Fix up both issues to adhere to the specification and make
hostapd-mbedtls work with the OWE security type.
Signed-off-by: David Bauer <mail@david-bauer.net>
Configure the PLMN and APN to the modem. This is required in cases,
where either the SGSN or GGSN does not permit the selection of IPv4v6
pdp type.
Previously, the modem always tried to establish a dual-stacked PDP
context regardless of the configured PDP type in uci. As this setting
can not be parameterized when creating a WDS context, configure it to
the modems internal list of profiles. This way, the PDP type is taken
into account when creating the WDS context.
Signed-off-by: David Bauer <mail@david-bauer.net>
The PLMN selection was reset when calling network-register, thus
rendering the sepcific selection of a carrier unapplied.
Set the PLMN selection after executing network-register. This seems to
cause the modem to re-select the carrier eventually.
That being said, qmi does allow the parameterization of the
network-register to include dpecific PLMN settings, however this is
currently not implemented in uqmi.
Signed-off-by: David Bauer <mail@david-bauer.net>
Set the RAT preference before attaching. This handles cases better,
where a network might be available but not with the preferred RAT.
If RAT is changed to a non-available RAT after attach, QMI does not fail
with missing registration but with failing to establish a PDP session.
Signed-off-by: David Bauer <mail@david-bauer.net>
Increase the wait time before polling the connection state for the first
time.
Depending on the prior state of the modem, the first poll might still
return a connected state. The script then tries to establish a PDP
session, which subsequently fails as the modem by then is in scan state.
Increasing the wait-time to 3 seconds mitigates this from happening.
Signed-off-by: David Bauer <mail@david-bauer.net>
On some network-triggered disconnections the UIM state might end up in
"illegal". This prevents the modem from attaching to any network in
non-restricted service modes.
Detect this state and reset the SIM card. This way, the modem can attach
to networks again.
Signed-off-by: David Bauer <mail@david-bauer.net>
