odhcpd, branch master

odhcpd: ensure zero padding on DNSSL

2026-03-16T17:50:03Z

From https://www.rfc-editor.org/rfc/rfc8106#section-5.2 regarding the DNSSL field: Because the size of this field MUST be a multiple of 8 octets, for the minimum multiple including the domain name representations, the remaining octets other than the encoding parts of the domain name representations MUST be padded with zeros. The current code leaves the trailing octets in uninitialized state, resulting in parsing errors on (at least) systemd RA clients. This commit restores the explicit zeroing of the trailing octets. Closes: https://github.com/openwrt/openwrt/issues/22351 Fixes: 0a54ce0d2b57 ("odhcpd: rename iface->search -> iface->dns_search") Signed-off-by: Eric Fahlgren Link: https://github.com/openwrt/odhcpd/pull/386 Signed-off-by: Álvaro Fernández Rojas

odhcpd: ignore NULL in dhcpv4_free_lease() stub

2026-03-16T17:47:36Z

When DHCPV4_SUPPORT is not defined, a stub implementation of dhcpv4_free_lease() is provided that logs an error when it's called. A couple sites in config.c unconditionally free the dhcpv4_lease field of struct lease_cfg, causing log spam like: | Trying to free IPv4 assignment 0x(nil) Silence by making the stub a noop if the argument is NULL, just like the actual implementation. Closes: https://github.com/openwrt/odhcpd/issues/382 Signed-off-by: Mark H. Spatz Link: https://github.com/openwrt/odhcpd/pull/384 Signed-off-by: Álvaro Fernández Rojas

ubus: fix truncated field in DHCPv6 lease query

2026-03-16T17:46:15Z

The assigned_subnet_id field in struct dhcpv6_lease is uint32_t, not uint16_t. Unlikely to matter in practice, since the number of meaningful bits in the field is (assignment size) - (interface assignemt size), so largest typically be 64 - 48 = 16. But it is possible to assign a /32 to an interface in openwrt. Signed-off-by: Mark H. Spatz Link: https://github.com/openwrt/odhcpd/pull/383 Signed-off-by: Álvaro Fernández Rojas

dhcpv4: fix avl_delete on leases not in avl tree

2026-03-16T17:33:48Z

dhcpv4_free_lease() unconditionally calls avl_delete() to remove the lease from the avl tree at lease->iface->dhcpv4_leases. This corrupts the tree if the lease is not in it, which happens if dhcpv4_assign() returns false during address reassignment (line ~611) or assignment (line ~630). The check for (lease->iface == NULL) before avl_delete() does not guard against this and is not necessary, because lease->iface is populated by dhcpv4_alloc_lease() and never cleared. Fix by explicitly checking that the lease is in the tree before deleting it. Fixes: aa6870b ("dhcpv4: use an AVL to store leases") Signed-off-by: Mark H. Spatz Link: https://github.com/openwrt/odhcpd/pull/381 Signed-off-by: Álvaro Fernández Rojas

dhcpv4: fix segfault when disabling interface

2026-03-16T17:33:44Z

dhcpv4_free_lease already calls avl_delete() on the node it's freeing, and thus shouldn't be called inside an avl_remove_all_elements() loop. Use avl_for_each_element_safe() instead. Its documentation says: "This loop can be used if the current element might be removed from the tree during the loop." Closes: https://github.com/openwrt/odhcpd/issues/371 Signed-off-by: Mark H. Spatz Link: https://github.com/openwrt/odhcpd/pull/381 Signed-off-by: Álvaro Fernández Rojas

dhcpv6-ia: dhcpv6_log_ia_addr(): parse return of snprintf

2026-03-12T22:26:23Z

snprintf() returns also in the failure case how many characters would have been required to print the full string. Check return of snprintf() and ensure the log_ctxt->buf_idx never points outside of the given log_ctxt->buf. Fixes: TOB-OWRT-5 Reported-by: Trail of Bits Signed-off-by: Alexander Couzens

dhcpv4: de-escalate error logging to debug

2026-01-19T18:56:28Z

Logging in dhcpv4_handle_msg() was changed from debug to error during code refactor. Signed-off-by: Paul Donald Link: https://github.com/openwrt/odhcpd/pull/375 Signed-off-by: Álvaro Fernández Rojas

netlink: clean up sockets, close files

2026-01-10T09:45:22Z

Sockets are handled for most usage paths, except for shutdown. Handle those at shutdown. Signed-off-by: Paul Donald Link: https://github.com/openwrt/odhcpd/pull/367 Signed-off-by: Álvaro Fernández Rojas

dhcpv6-ia: switch case refactor in dhcpv6_handle_ias()

2025-12-22T16:52:18Z

Rewrite core logic using switch cases for clarity and improved performance. This also simplifies the DHCPV6_MSG_REBIND cases: DHCPV6_MSG_REBIND and !a is parsed in the switch DHCPV6_MSG_REBIND and a is now parsed after the switch (5 lines) Signed-off-by: Paul Donald https://github.com/openwrt/odhcpd/pull/360 Signed-off-by: Álvaro Fernández Rojas

dhcpv6: set static defines for DHCPv6 option header size

2025-12-22T10:11:42Z

so the number 4 isn't so mysterious at first glance. Signed-off-by: Paul Donald Link: https://github.com/openwrt/odhcpd/pull/363 Signed-off-by: Álvaro Fernández Rojas