`ucv_array_set` releases the array's reference to the object being cleared.
If this is the last reference to the object, it will be freed, making our
pointer `val` invalid.
To avoid this, we need to obtain our own reference to the object so we
can safely return `val`.
Signed-off-by: Matthew Cather <mattbob4@gmail.com>
(cherry picked from commit 6a10da2934441a59283394a84917edf4983e3043)
Since `wpa_ucode_registry_add` collects its own reference to the values added, the
two functions `hostapd_ucode_bss_get_uval` and `hostapd_ucode_iface_get_uval` would
sometimes return a referenced object (from `uc_resource_new`) and sometimes return
an unreferenced object (from `wpa_ucode_registry_get`). Now, both functions always
return a referenced object.
This change also indirectly fixes `hostapd_ucode_bss_get_uval`, ensuring it now
always returns a referenced object.
Signed-off-by: Matthew Cather <mattbob4@gmail.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 7729f960933a03c9eebd2b65a695ea57718ccb77)
Remove extra ucv_get calls when passing a referenced value to an object
without using it further.
Signed-off-by: Matthew Cather <mattbob4@gmail.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 22eaf1864731dd29918357b26565dbd9420fd314)
This fixes a common reference counting bug typically along the lines of:
```
uc_value_push(ucv_get(ucv_string_new(...)));
```
This would leave our new string with a reference count of 2, one from
the construction of the string, the other from `ucv_get`. This would
prevent the strings from being correctly cleaned up when it goes out
of scope.
Signed-off-by: Matthew Cather <mattbob4@gmail.com>
(cherry picked from commit f79968ee0f76c41414fb0c08980b9c2a259964d3)
75a236be122a service: add missing null pointer check
f5341f327539 ubus: add api for generating and validating security tokens
3fab99eab4d5 add udebug support
28d86bd30e97 pex: only respond to update requests when we have network data
8e6f37cc361e pex-msg: ignore no-data responses if version is zero
12e6cf7f63e1 pex: create pex host from update responses
edc8fdae463a ubus: show the local addresses in network status
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit ce68f61cb67a4701a108f838f510e72ca63ed78e)
ethtool since version 6.9 introduced support for getting/setting RSS
input transformation supported in Linux since version 6.8.
The now changed kernel ioctl ABI, however, cannot be detected from
userland, and ethtool since version 6.9 simply assumes that a previously
reserved field is now used to set the input transformation.
Unfortunately the default value RXH_XFRM_NO_CHANGE (0xff) used by ethtool
userland creates an incompatibility with older kernels which cannot be
resolved easily without introducing even more ABI breakage.
Work-around the issue and fix support for --set-rxfh and --set-rxfh-indir
ethtool userland tool commands by making the support for input_xfrm
conditional on compile time, and keep it disabled for Linux 6.6.
Fixes: 8c2dcd1518 ("ethtool: update to 6.10")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Tested-by: Stijn Segers <foss@volatilesystems.org>
(cherry picked from commit 3a7467ffde)
Version 6.11 - October 8, 2024
* Feature: cmis: print active and inactive firmware versions
* Feature: flash transceiver module firmware (--flash-module-firmware)
* Feature: add T1BRR 10Mb/s mode to link mode tables
* Feature: support for disabling netlink from command line
* Fix: fix lanes parameter format specifier
* Fix: add missing clause 33 PSE manual description
* Fix: qsf: Better handling of Page A2h netlink read failure
* Fix: rss: retrieve ring count using ETHTOOL_GRXRINGS ioctl (-x)
* Misc: man page formatting fix
* changelog here: https://git.kernel.org/pub/scm/network/ethtool/ethtool.git/commit/NEWS?id=c0ea4b70c71334ef038f7a3416b228a50dada406
Tested on gl.inet MT6000, retrieve ring count is now working
Signed-off-by: Andrea Pesaresi <andreapesaresi82@gmail.com>
Link: https://github.com/openwrt/openwrt/pull/17607
Signed-off-by: Robert Marko <robimarko@gmail.com>
(cherry picked from commit 9454331b7f)
Depending on the config / circumstances, the get_psk call can be called
multiple times from differnt places, which can lead to wrong sta->psk_idx
values. The correct call is the one that is also interested in the vlan_id,
so use the vlan_id pointer as indication of when to set sta->psk_idx.
Also fix off-by-one error for secondary PSKs
Fixes: b2a2c28617 ("hostapd: add support for authenticating with multiple PSKs via ubus helper")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 8118b2dace)
The 00 address_mask needs to be inverted, otherwise the mac address
allocation will modify the last byte instead of the first one.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 1ee44825ad)
93458ac dns: fix response to TYPE_PTR query
68af311 fix unicast response port and timeout
a2b4979 service: announce all services in single dns answer
4537734 display announced services in ubus call umdns browse
0b50c29 display more srv attributes in output of ubus browse function
Signed-off-by: John Crispin <john@phrozen.org>
(cherry picked from commit d162fd5ed4)
322500403615 service: add default group @ to match all nodes
5f7860306200 ubus: rename unetd_ubus_notify to unetd_ubus_network_notify
d13752814651 enroll: add PEX sub-protocol to support enrolling new nodes into a network
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit c0f06cb6ac)
Makes it easier to enable MDNS on wan without having to edit the firewall
configuration for it
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 05138fe898)
Unless another toolchain is present (or selected), build the bpf toolchain
whenever a package is selected that needs it.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 6605e45676)
lldpd was updated, so reset PKG_RELEASE after the PKG_VERSION update.
Signed-off-by: Robert Marko <robimarko@gmail.com>
(cherry picked from commit abbec429b4)
Changes (breaking):
- Remove support for building 802.3bt TLVs (broken).
Fix:
- Fix memory leaks in EDP/FDP decoding when receiving some TLVs twice.
- Do not set interface description continuously.
- Use a different Netlink socket for changes and queries.
Signed-off-by: Sebastian Pflieger <sebastian@pflieger.email>
Link: https://github.com/openwrt/openwrt/pull/17570
Signed-off-by: Robert Marko <robimarko@gmail.com>
(cherry picked from commit a18385041e)
Some platforms a single ethernet device for all ports with multiple rx rings
and NAPI threading enabled. In this case, the steering script was limiting
performance by keeping all NAPI threads assigned to the same CPU.
Fix this by assigning each rx queue and the corresponding NAPI task separately.
Additionally, if the number of rx queues is at least as big as the number of
CPUs, skip weight based assignment and distribute the load across all CPUs
directly.
Fixes: https://github.com/openwrt/openwrt/issues/17611
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit acce25b789)
Bringing up a mesh interface using wpa_supplicant already supports a
per-VIF basic rate selection. Add the same ability when creating a mesh
VIF without wpa_supplicant.
Signed-off-by: David Bauer <mail@david-bauer.net>
(cherry picked from commit 3deeb7805f)
This can be used to configure the base mac address from which all
interface mac addresses are derived
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit 00860e485b)
Simplifies setting ifname to a different pattern for all affected
interfaces.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit a9ff3ba24b)
Basic rates were not set for mesh-interfaces, resulting in the undesired
behavior where 11s frames might be sent with a rate which was not
configured.
Depending on the driver, the basic rate might also be used to determine
the beacon rate configured to the chip. One such example are MediaTek
MT7915 platforms.
Signed-off-by: David Bauer <mail@david-bauer.net>
(cherry picked from commit 3ec7f3a512)
On some drivers, setting the tx power on the interface is not enough.
Set it for the phy as well.
Fixes: 04fb05914e ("wifi-scripts: add multi-radio config support")
Signed-off-by: Felix Fietkau <nbd@nbd.name>
(cherry picked from commit b795e5cbcf)
