Vulnerabilities

This page collects all of the vulnerabilities that are discovered and fixed in each release. It will also often have more details than is available in the releases. Some vulnerabilities are deemed to be sensitive, and will not be publicly discussed until there is sufficient time to fix them. Because the release notes are locked to a version, the information here can be updated after the embargo is lifted.

Vulnerabilities from previous years are collected on separate pages:

CVE-2026

CVE 2026-0849

crypto: ATAES132A response length allows stack buffer overflow

Malformed ATAES132A responses with an oversized length field overflow a 52-byte stack buffer in the Zephyr crypto driver, allowing a compromised device or bus attacker to corrupt kernel memory and potentially hijack execution.

This has been fixed in main for v4.4.0

CVE 2026-1677

net: TLS 1.2 connections allowed on TLS 1.3 sockets

Zephyr sockets created with IPPROTO_TLS_1_3 can still negotiate a TLS 1.2 connection when both TLS versions are enabled in Kconfig, because the socket-level protocol selection is not propagated to mbedTLS (e.g. via mbedtls_ssl_conf_min_tls_version). The ClientHello advertises both versions and the peer can establish TLS 1.2, so applications that assumed IPPROTO_TLS_1_3 enforces TLS 1.3 may silently use TLS 1.2 and remain exposed to TLS 1.2-specific weaknesses.

This has been fixed in main for v4.4.0

CVE 2026-1678

dns: memory‑safety issue in the DNS name parser

dns_unpack_name() caches the buffer tailroom once and reuses it while appending DNS labels. As the buffer grows, the cached size becomes incorrect, and the final null terminator can be written past the buffer. With assertions disabled (default), a malicious DNS response can trigger an out-of-bounds write when CONFIG_DNS_RESOLVER is enabled.

This has been fixed in main for v4.4.0

CVE 2026-1679

The eswifi socket offload driver copies user-provided payloads into a fixed buffer without checking available space; oversized sends overflow eswifi->buf, corrupting kernel memory (CWE-120). Exploit requires local code that can call the socket send API; no remote attacker can reach it directly.

This has been fixed in main for v4.4.0

CVE 2026-1681

net: Stack Overflow with Ping (to own IP Address) via Shell

Issuing an ICMP ping via the net ping shell command to a device’s own IPv4 address causes the network stack to recursively re-enter the input path on the same system work-queue stack. Because the destination is recognized as a local address, both the echo request and the resulting echo reply are processed inline before the current frame returns. The nested input-path frames exceed the work-queue stack and trigger a stack overflow.

This has been fixed in main for v4.4.0

CVE 2026-4179

stm32: usb: Infinite while loop in Interrupt Handler

Issues in stm32 USB device driver can lead to an infinite while loop.

This has been fixed in main for v4.4.0

CVE 2026-5066

net: sockets: tls: Potential out-of-bounds write/read in socket_op_vtable::connect function

A potential out-of-bounds write/read exists in the TLS socket connect path of the network sockets subsystem (subsys/net/lib/sockets/sockets_tls.c). When the TLS session cache is enabled, tls_session_store() and tls_session_restore() memcpy the caller-supplied address into a fixed-size buffer using the caller-controlled addrlen value without validating it against the destination size. Since struct net_sockaddr is an opaque type, an application can pass an addrlen larger than sizeof(struct net_sockaddr) (for example 128 bytes into a 24-byte stack buffer), causing the memcpy to read and write past the end of the address memory used by the TLS session cache. This can lead to a crash and denial of service, and potentially to arbitrary code execution.

This has been fixed in main for v4.4.0

CVE 2026-5067

Out-of-bounds read/write in HTTP WebSocket upgrade via non-null-terminated Sec-WebSocket-Key

A remote, unauthenticated attacker can trigger memory corruption in Zephyr’s HTTP server WebSocket upgrade path by sending a crafted Sec-WebSocket-Key header that is copied without guaranteed NUL termination and then passed to strlen(). This can cause out-of-bounds read and out-of-bounds write on stack memory, leading to a crash (denial of service) and potentially code execution. The path is reachable when CONFIG_HTTP_SERVER_WEBSOCKET is enabled.

This has been fixed in main for v4.4.0

CVE 2026-5068

Bluetooth: L2CAP LE CoC: remote out-of-bounds write via segmentation counter stored in net_buf user_data

A remote, unauthenticated BLE peer can trigger a 2-byte out-of-bounds write in the Bluetooth host during L2CAP LE CoC SDU reassembly. When the application enables segmentation (via chan_ops.alloc_buf) and the chosen RX pool has a user_data_size smaller than 2 bytes, the segmentation counter stored in the net_buf user_data area is written out of bounds in l2cap_chan_le_recv_seg (subsys/bluetooth/host/l2cap.c). This can lead to heap corruption and a fatal error.

This has been fixed in main for v4.4.0

CVE 2026-5071

can: Local Denial of Service via SocketCAN Send

The SocketCAN send path (zcan_sendto_ctx) validated the caller-supplied buffer length with a NET_ASSERT instead of a real runtime check. In production builds where assertions are compiled out, a userspace app could pass a buffer shorter than struct socketcan_frame, and socketcan_to_can_frame() would dereference fields past the end of that buffer — an out-of-bounds read that can crash the system (local DoS) or, because the parsed frame is then transmitted, potentially leak adjacent memory.

This has been fixed in main for v4.4.0

CVE 2026-5072

net: ptp: Potential Denial of Service via PTP Interval Shift

A bitwise shift vulnerability allows a remote attacker to cause undefined behavior and potential crashes in the PTP subsystem by sending a crafted PTP Management or Delay Response packet containing a large, unvalidated, negative log_announce_interval used in the bitwise shift operation.

This has been fixed in main for v4.4.0

CVE 2026-5589

Bluetooth: Mesh: Out-of-bounds write caused by an integer underflow

An integer underflow in bt_mesh_sol_recv() in the Bluetooth Mesh solicitation handling (subsys/bluetooth/mesh/solicitation.c) leads to an out-of-bounds write. When CONFIG_BT_MESH_OD_PRIV_PROXY_SRV is enabled, the function parses solicitation PDUs from raw BLE advertising payloads. The AD parsing loop reads an attacker-controlled length byte and computes reported_len - 3 without checking that reported_len is at least 3. When the value is smaller, the signed subtraction yields a negative number that bypasses the length guard and is then implicitly converted to a very large size_t, advancing the buffer pointer far out of bounds so that subsequent reads dereference invalid memory. A nearby BLE device can trigger this with a non-connectable advertisement carrying a UUID16 AD structure and a crafted length byte, with no pairing or prior association required, potentially leading to denial of service or arbitrary code execution.

This has been fixed in main for v4.4.0

CVE 2026-5590

net: ip/tcp: Null pointer dereference can be triggered by a race condition

A race condition during TCP connection teardown can cause tcp_recv() to operate on a connection that has already been released. If tcp_conn_search() returns NULL while processing a SYN packet, a NULL pointer derived from stale context data is passed to tcp_backlog_is_full() and dereferenced without validation, leading to a crash.

This has been fixed in main for v4.4.0

CVE 2026-8718

Out-of-bounds write in DTLS peer Connection ID getsockopt (TLS_DTLS_PEER_CID_VALUE) in Zephyr net sockets/TLS

tls_opt_dtls_peer_connection_id_value_get() in subsys/net/lib/sockets/sockets_tls.c, which handles getsockopt(SOL_TLS, TLS_DTLS_PEER_CID_VALUE), passed the caller-supplied optval directly to mbedtls_ssl_get_peer_cid() without verifying the buffer was at least MBEDTLS_SSL_CID_OUT_LEN_MAX (default 32) bytes. mbedtls_ssl_get_peer_cid() copies the peer-negotiated DTLS Connection ID (length 1..MBEDTLS_SSL_CID_OUT_LEN_MAX) into that buffer without a destination-size parameter, so a caller-supplied optlen smaller than the CID causes a write of up to 31 bytes past the buffer end.

In CONFIG_USERSPACE builds the getsockopt syscall verifier (z_vrfy_zsock_getsockopt) bounce-buffers the user’s optval into a kernel allocation of exactly optlen bytes (k_usermode_alloc_from_copy -> z_thread_malloc), so an unprivileged user thread that passes a small optlen on a connected DTLS socket with Connection ID enabled induces a kernel-heap buffer overflow, with the overflowing content being the remote peer’s CID.

The defect requires CONFIG_MBEDTLS_SSL_DTLS_CONNECTION_ID, an established DTLS session with a negotiated peer CID, and (for the kernel-crossing case) CONFIG_USERSPACE. Introduced when the TLS_DTLS_CID option was added (v3.5.0).

The fix rejects callers whose optlen is below MBEDTLS_SSL_CID_OUT_LEN_MAX with -EINVAL.

This has been fixed in main for v4.5.0

CVE 2026-9263

Out-of-bounds read in Bluetooth Controller ISOAL framed RX reassembly leaks adjacent memory into host HCI ISO packets

The Zephyr Bluetooth controller ISO Adaptation Layer (subsys/bluetooth/controller/ll_sw/isoal.c) fails to validate the length field of a framed ISO PDU start segment. Per the Bluetooth specification a start segment (sc=0) always carries a 3-byte time_offset, so its segment-header len must be at least PDU_ISO_SEG_TIMEOFFSET_SIZE (3). isoal_check_seg_header() accepted start segments with len < 3 as valid, and isoal_rx_framed_consume() then computed length = seg_hdr->len - 3 in a uint8_t, underflowing to 253-255 when len is 0-2. That oversized length is passed to isoal_rx_append_to_sdu(), whose copy is clamped only against the destination SDU buffer size, not the source PDU length, so up to ~255 bytes of controller memory beyond the received PDU are copied (via sink_sdu_write_hci()/net_buf_add_mem) into an HCI ISO data packet and delivered to the host. The PDU and its segment headers are entirely attacker-controlled and arrive over the air, reachable through both the CIS and BIS-sync HCI data paths (hci_driver.c) and the vendor data path (ull_iso.c), so a remote CIS peer or a broadcaster the device is synced to can trigger an out-of-bounds read causing information disclosure to the host and potential denial of service (faults or malformed oversized HCI ISO packets). The flaw affects all Zephyr releases since framed ISO reception was introduced in v3.0.0. The fix rejects sc=0 segments with len < 3 in isoal_check_seg_header() and adds a guard before the subtraction in isoal_rx_framed_consume().

This has been fixed in main for v4.5.0

CVE 2026-10593

Remotely triggerable NULL-pointer dereference in Bluetooth LE Audio BAP unicast client QoS-state handling

The Zephyr Bluetooth LE Audio Basic Audio Profile (BAP) unicast client mishandles peer-supplied ASE state notifications. In unicast_client_ep_qos_state() (subsys/bluetooth/audio/bap_unicast_client.c), the handler writes attacker-controlled QoS fields (interval, framing, phy, sdu, rtn, latency, pd) through the stream->qos pointer with only a stream != NULL guard. stream->qos is NULL for any stream that has been codec-configured via bt_bap_stream_config() but not yet added to a unicast group (it is set only by unicast_group_add_stream()).

A malicious or buggy remote ASCS server, to which the local device is connected as a BAP unicast client, can send a GATT notification announcing the ASE has entered the QoS Configured state while the local endpoint is still in the Codec Configured state — a transition the dispatcher explicitly permits — during that window, causing a write through a NULL pointer and a crash (denial of service). The data written is itself remote-controlled.

The defect shipped in v4.3.0 and v4.4.0 (and earlier). The fix re-points all BAP QoS storage to the always-valid embedded ep->qos struct, eliminating the NULL dereference.

This has been fixed in main for v4.5.0

CVE 2026-10634

Use-after-free in Zephyr native TCP net_tcp_foreach() due to dropping tcp_lock during the callback

Zephyr’s native TCP stack iterates the global connection list in net_tcp_foreach() (subsys/net/ip/tcp.c) using the SYS_SLIST_FOR_EACH_CONTAINER_SAFE macro, which caches a pointer to the next list node. Prior to this fix the function released tcp_lock while invoking the per-connection callback and re-acquired it afterwards. During that window a concurrent tcp_conn_release(), running on the dedicated TCP work-queue thread when a connection’s reference count drops to zero (e.g. a remote peer closing or resetting the connection), can remove and k_mem_slab_free() the cached next connection. When the iterator advances it dereferences the freed (and possibly reallocated) slab memory — a use-after-free that can crash the system (denial of service) and, if the slot has been reused, cause the callback to operate on an attacker-influenced object (potential information disclosure or further fault). net_tcp_foreach() is reached in production via the net conn network shell command and via net_tcp_close_all_for_iface() on interface-down; the freeing side is driven by ordinary TCP traffic. The fix moves the connection/context teardown in tcp_conn_release() inside the tcp_lock critical section and keeps tcp_lock held across the callback in net_tcp_foreach(). The defect was introduced with the modern (TCP2) stack in 2020 and affects releases up to and including v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10635

Dangling memory-domain pointer (use-after-free) in Xtensa MMU page-table code on memory-domain de-init

On Xtensa targets with CONFIG_USERSPACE and CONFIG_XTENSA_MMU, the page-table code (arch/xtensa/core/ptables.c) maintains a global list, xtensa_domain_list, of active memory domains using a list node embedded inside the caller-owned struct k_mem_domain. When a domain is destroyed via k_mem_domain_deinit() -> arch_mem_domain_deinit(), the page tables are torn down and domain->arch.ptables is set to NULL, but the domain’s node was not removed from xtensa_domain_list. The freed/deinitialized domain therefore remained linked into the global list as a dangling pointer into caller-owned storage that may then be freed or reused.

Any subsequent arch_mem_map()/arch_mem_unmap() operation (widely invoked by kernel memory-mapping and demand-paging code) traverses the stale node and dereferences domain->ptables: at minimum a NULL pointer dereference causing a fatal MMU exception (denial of service), and if the k_mem_domain storage has been freed or reused, a use-after-free in which a stale/controlled ptables value is dereferenced and written through during the page-table walk (l2_page_table_map writes l1_table[...] and l2_table[...], and xtensa_mmu_compute_domain_regs writes into the domain struct and the L1 table), yielding page-table memory corruption that can undermine userspace isolation.

The vulnerable path is reachable only from privileged kernel/supervisor code (k_mem_domain_deinit is not a syscall), not directly from unprivileged user threads or remotely. Affected: Zephyr v4.4.0 (the Xtensa memory-domain de-initialization feature was introduced in commit 3032b58f52d and first shipped in v4.4.0); fixed on main by adding sys_slist_find_and_remove() in arch_mem_domain_deinit(). The Xtensa MPU path is unaffected.

This has been fixed in main for v4.5.0

CVE 2026-10636

Use-after-free in Zephyr IPv4 IGMP send path (igmp_send)

In Zephyr’s IPv4 IGMP implementation, igmp_send() in subsys/net/ip/igmp.c read the network interface back out of the packet via net_pkt_iface(pkt) after the packet had been handed to net_send_data(). On the successful-send path the packet’s last reference may already have been released by the L2 driver or by the network stack’s TX handling (synchronously in the default NET_TC_TX_COUNT=0 immediate-transmit configuration), returning the net_pkt slab block to its free list. The subsequent net_pkt_iface(pkt) dereferences the freed packet, a use-after-free read; with CONFIG_NET_STATISTICS_PER_INTERFACE the resulting dangling interface pointer is further dereferenced for a statistics-counter write. The IGMP send path is reachable without authentication from inbound IPv4 IGMP membership queries addressed to 224.0.0.1 (net_ipv4_igmp_input -> send_igmp_report/send_igmp_v3_report -> igmp_send), as well as from local multicast join/leave/rejoin operations. Realistic impact is undefined behavior and potential denial of service (sporadic crash or stats corruption); a controllable write requires the asynchronous TX path plus a concurrent slab reuse. The flaw was introduced with IGMPv2 support and affects releases from v2.6.0 through v4.4.0. The fix caches the interface pointer before sending. Note the analogous IPv6 MLD path (mld_send in subsys/net/ip/ipv6_mld.c) retains the same unfixed pattern.

This has been fixed in main for v4.5.0

CVE 2026-10637

Use-after-free of net_pkt in IPv6 MLD send path triggerable by a link-local MLD Query

subsys/net/ip/ipv6_mld.c:mld_send() read the packet interface via net_pkt_iface(pkt) after net_send_data(pkt) returned successfully. Per the network stack’s ownership contract (include/zephyr/net/net_core.h, and the explicit warning in subsys/net/ip/net_core.c:453-460 ‘do not use pkt after that call’), a successful send transfers ownership of the net_pkt and the L2 driver frees it (e.g. ethernet_send() unrefs the packet on success, subsys/net/l2/ethernet/ethernet.c:790), returning it to its k_mem_slab. The subsequent net_pkt_iface(pkt) is therefore a read of a freed object; the recovered interface pointer is then dereferenced and incremented by the per-interface statistics path (net_stats.h UPDATE_STAT/SET_STAT) when CONFIG_NET_STATISTICS_PER_INTERFACE is enabled. If the freed slot is concurrently reallocated, pkt->iface may read back as NULL (NULL-pointer dereference / crash) or as a stale/garbage pointer (stray increment write / memory corruption). The path is reachable remotely on the local link without authentication: handle_mld_query() (registered for NET_ICMPV6_MLD_QUERY) responds to a valid MLDv2 General Query (unspecified multicast address, hop limit 1) by calling send_mld_report() -> mld_send(). The result is a remotely triggerable denial of service of the networking stack, with a narrow possibility of memory corruption. The fix caches the interface in a local before sending and no longer touches the packet after net_send_data(). The IPv4/IGMP sibling (igmp_send) already used the corrected pattern.

This has been fixed in main for v4.5.0

CVE 2026-10638

Use-after-free in Zephyr ICMPv6 RX path when updating statistics after sending an echo reply or error

subsys/net/ip/icmpv6.c reads the network interface from a net_pkt after that packet has been handed to net_try_send_data(). In icmpv6_handle_echo_request() and net_icmpv6_send_error(), the post-send statistics update calls net_pkt_iface(reply)/net_pkt_iface(pkt) on the just-sent packet. The send path (net_try_send_data -> net_if_tx) unreferences and may free the packet back to its memory slab before returning — synchronously in the RX thread when no TX queue is configured (CONFIG_NET_TC_TX_COUNT == 0), and asynchronously the driver/L2 may already have freed it otherwise. net_pkt_iface() therefore dereferences a freed (and possibly reused) net_pkt; with CONFIG_NET_STATISTICS_PER_INTERFACE the stale iface pointer is further dereferenced and written through (iface->stats.icmp.sent++), turning the use-after-free read into a write through an attacker-influenceable pointer. The core stack already documents this hazard in net_core.c (“do not use pkt after that call”) and caches iface before sending; the ICMPv6 callers did not. An unauthenticated remote attacker triggers the flaw simply by sending an ICMPv6 Echo Request (ping) or an IPv6 packet that elicits an ICMPv6 error (unknown next header, fragment reassembly timeout, destination unreachable), leading to denial of service via crash and potential memory corruption. Affected: Zephyr networking with CONFIG_NET_NATIVE_IPV6, roughly v4.2.0 through v4.4.0. The fix caches the interface pointer before sending and uses it for all statistics updates; the sibling commit 86e21665d46 fixes the identical bug in ICMPv4.

This has been fixed in main for v4.5.0

CVE 2026-10639

Use-after-free reading net_pkt_iface() of a sent ICMPv4 echo-reply packet in icmpv4_handle_echo_request()

In Zephyr’s native IPv4 stack, icmpv4_handle_echo_request() in subsys/net/ip/icmpv4.c builds an echo-reply packet (reply), hands it to net_try_send_data(), and then, on success, calls net_stats_update_icmp_sent(net_pkt_iface(reply)). net_try_send_data() transfers ownership of reply to the TX path (net_if_try_queue_tx -> net_if_tx -> L2/driver send, or the asynchronous net_if_tx_thread), which can unref it to refcount 0 and return the struct net_pkt to its slab (net_pkt_unref -> k_mem_slab_free) before the stats line runs. net_core.c documents this exact contract (‘the pkt might contain garbage already … do not use pkt after that call’).

The post-send net_pkt_iface(reply) therefore reads reply->iface out of a freed (and possibly already reallocated) net_pkt, a use-after-free read; with CONFIG_NET_STATISTICS_PER_INTERFACE the stats macro additionally increments a counter through that value, i.e. a dereference/write through a stale or recycled-slot pointer.

The path is reached unauthenticated by any remote host that pings the device (net_icmpv4_input -> net_icmp_call_ipv4_handlers -> icmpv4_handle_echo_request) and is gated on CONFIG_NET_STATISTICS_ICMP. Impact is a probabilistic read of recycled packet memory plus a possible wild-pointer write under a timing race, leading most likely to corrupted interface statistics or a remotely triggerable crash (DoS).

The defect was introduced in 2019 (v1.14) and is present through v4.4.0. The companion change in net_icmpv4_send_error() is not a use-after-free because it reads net_pkt_iface(orig), the caller-owned received packet, which stays alive across the send. The fix caches the interface pointer from the live received packet before sending and uses it for the post-send stats updates.

This has been fixed in main for v4.5.0

CVE 2026-10640

Use-after-free reading net_pkt iface after send in IPv6 Neighbor Discovery (ipv6_nbr.c)

Zephyr’s IPv6 Neighbor Discovery send paths (net_ipv6_send_na, net_ipv6_send_ns, net_ipv6_send_rs in subsys/net/ip/ipv6_nbr.c) updated the per-interface ICMP- sent statistics by calling net_pkt_iface(pkt) after net_send_data(pkt) had already returned successfully. On the success path the network stack owns and releases the packet’s reference (the L2/driver send unrefs it, e.g. ethernet_send -> net_pkt_unref), so for a freshly allocated packet with refcount 1 the net_pkt slab block can be freed before the statistics line runs (synchronously when no TX queue thread is configured, or via a concurrent TX thread otherwise).

The subsequent net_pkt_iface(pkt) reads pkt->iface from the freed slab block, and with CONFIG_NET_STATISTICS_PER_INTERFACE enabled that loaded pointer is dereferenced to increment iface->stats.icmp.sent, a use-after-free (CWE-416). If the slab block was reallocated in the meantime the read/increment targets unrelated or attacker-influenced memory, yielding corrupted statistics, a fault/crash (denial of service), or potential limited memory corruption.

The vulnerable Neighbor Advertisement path is reachable by any unauthenticated on-link node simply by sending ICMPv6 Neighbor Solicitations to a Zephyr node with native IPv6 enabled (handle_ns_input -> net_ipv6_send_na).

Affected from v3.3.0 through v4.4.0; the fix uses the already-available iface argument instead of touching the sent packet. Configurations without per-interface statistics dereference only a global counter and are not affected by the memory-safety aspect.

This has been fixed in main for v4.5.0

CVE 2026-10641

Out-of-bounds write in Bluetooth HFP Hands-Free CIND indicator parsing (cind_handle_values)

Zephyr’s Bluetooth Classic Hands-Free Profile (HFP) Hands-Free role parser (subsys/bluetooth/host/classic/hfp_hf.c) contains an out-of-bounds write. During Service Level Connection setup the HF sends AT+CIND=? and parses the AG’s +CIND: response in cind_handle(), which assigns a per-entry counter index and calls cind_handle_values() for each list element. cind_handle_values() then wrote hf->ind_table[index] = i without verifying that index is within the 20-element int8_t ind_table[] array of struct bt_hfp_hf. Because the parser places no cap on the number of +CIND: list entries, a remote Attendant Gateway (a malicious, compromised, or spoofed peer the device connects to over Bluetooth) can send a response with more than 20 recognized indicator entries and drive index arbitrarily large, writing a small attacker-positioned value past the array into adjacent struct fields (feature masks, SDP/version state, the calls[] array, work/atomic bookkeeping) and potentially beyond the static connection pool slot. This yields memory corruption and at least denial of service of the Bluetooth host, triggered by a single malformed AT response with no user interaction. The sibling consumer ag_indicator_handle_values() already performed the equivalent bounds check; this commit adds the same index >= ARRAY_SIZE(hf->ind_table) guard to close the gap. Affects builds with CONFIG_BT_HFP_HF enabled; introduced with the original HFP HF CIND parser (~v1.7) and present through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10642

Unbounded TX busy-loop DoS in Zephyr PL011 UART driver under CTS hardware flow control

The Zephyr PL011 UART driver (drivers/serial/uart_pl011.c) contains an unbounded software loop in pl011_irq_tx_enable() that repeatedly invokes the interrupt-driven application callback while the TX interrupt mask bit (PL011_IMSC_TXIM) is set, to work around the controller’s level-transition TX-interrupt behavior.

When CTS hardware flow control is enabled (devicetree hw-flow-control or runtime UART_CFG_FLOW_CTRL_RTS_CTS) and the wired serial peer de-asserts CTS, the controller stops draining the TX FIFO; pl011_fifo_fill() then returns 0 on every call while the application still has pending data and therefore never disables the TX interrupt. The loop condition never clears, so the thread that called uart_irq_tx_enable() (e.g. h4_send() in the Bluetooth HCI H4 driver) spins indefinitely, hanging the executing context and stalling the transport — a denial of service (CWE-835).

An attacker controlling the device attached to the UART’s CTS line can trigger the hang by withholding CTS during transmission. Because that peer is the device wired to the UART — which may be a removable or external module (e.g. an off-board Bluetooth controller on the HCI H4 link) rather than a permanently-bonded on-PCB part — the attack vector is scored Adjacent (AV:A) rather than Physical; the security subcommittee should confirm the vector against the specific deployment. Impact is availability only; there is no memory-safety, confidentiality, or integrity consequence.

The vulnerable loop was introduced in commit b783bc8448ef (Feb 2025) and shipped in releases v4.1.0 through v4.4.0. The fix breaks out of the loop when CTS is blocking and arms the CTS modem-status interrupt to resume transmission when CTS re-asserts.

This has been fixed in main for v4.5.0

CVE 2026-10643

Out-of-bounds heap write in Zephyr recvmsg() ancillary-data path (insert_pktinfo undersizes the control-buffer capacity check)

Zephyr’s IP socket recvmsg() implementation (subsys/net/lib/sockets/sockets_inet.c, insert_pktinfo()) validated the user-supplied ancillary (msg_control) buffer using only the payload length (msg->msg_controllen < pktinfo_len) before writing a full control message consisting of an aligned cmsg header plus the payload. Because the check omitted the cmsg header size, a control buffer whose length falls in the under-checked window (e.g. 16-27 bytes for IPv4 IP_PKTINFO on a 64-bit target, where a single element actually occupies 28 bytes) passes the guard yet causes a fixed-size out-of-bounds write of up to one cmsg header (~12 bytes) past the end of the buffer.

Under CONFIG_USERSPACE the recvmsg verifier allocates a kernel-heap copy of the control buffer sized to msg_controllen and runs the implementation against it, so the overflow corrupts kernel heap memory and is triggerable from an unprivileged userspace thread; in supervisor mode it corrupts the caller’s buffer.

The path is reachable on a UDP/IP socket with IP_PKTINFO/IPV6_RECVPKTINFO (or hoplimit/timestamping) enabled when the application calls recvmsg() with an undersized control buffer and a datagram is received; part of the overwritten bytes (the destination IP in ipi_addr) is influenced by the received packet.

The fix makes the capacity check use NET_CMSG_SPACE(pktinfo_len) (aligned header + aligned data) and returns -ENOMEM when the buffer is too small. Affected: v3.6.0 through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10644

Out-of-bounds write in Microchip SERCOM-G1 (PIC32CM-JH) async UART RX with 1-byte buffer

The Microchip SERCOM-G1 UART driver (drivers/serial/uart_mchp_sercom_g1.c), used by the PIC32CM-JH SoC family, contains an out-of-bounds write in its asynchronous (DMA) receive path. When uart_rx_enable() is invoked with a one-byte receive buffer (len == 1) and CONFIG_UART_MCHP_ASYNC is enabled, the RX-complete ISR starts a single-beat DMA transfer while a received byte is already pending in the SERCOM DATA register. On this SoC the peripheral-triggered DMA start sequencing then writes one byte past the end of the caller-supplied buffer (CWE-787).

The overflowed byte’s value is the UART RX data supplied by the connected serial peer (adjacent attacker), while its size and location are fixed at one byte immediately after the buffer.

Exploitation requires the async UART config (not enabled by default on the in-tree PIC32CM-JH boards) and a consumer that enables RX with a one-byte buffer; impact is limited single-byte memory corruption adjacent to the RX buffer (possible crash / denial of service).

The defect shipped in v4.4.0. The fix reads the first byte with the CPU and, for one-byte buffers, performs no DMA at all; for larger buffers it sizes the DMA for the remaining len-1 bytes.

This has been fixed in main for v4.5.0

CVE 2026-10645

Out-of-bounds read in Zephyr ext2 directory entry traversal from a crafted filesystem image

The Zephyr ext2 filesystem driver (subsys/fs/ext2) trusted the on-disk directory entry fields de_rec_len and de_name_len when walking a directory block. ext2_fetch_direntry() guarded only with de_name_len > EXT2_MAX_FILE_NAME, but de_name_len is a uint8_t and EXT2_MAX_FILE_NAME is 255, so the check is always false; the function then memcpy’d up to 255 name bytes and the lookup/readdir paths advanced traversal by an unvalidated de_rec_len. Each directory block is read into a block_size-sized slab buffer, and block_off can be driven near the block end by preceding entries’ rec_len, so the 8-byte header read and the subsequent name memcpy can read up to ~263 bytes past the end of the block buffer into adjacent heap/slab memory. On the readdir path those bytes are returned to the caller in fs_dirent.name, leaking adjacent kernel heap memory; a de_rec_len of 0 also causes a zero-progress infinite loop (denial of service), and the unlink path’s memmove(de, next, next_reclen) over unvalidated records is an additional OOB read/write source. The defect is reached by any path-based operation (open, stat, unlink, rename, mkdir) or directory listing on a mounted ext2 volume, so a crafted or corrupted ext2 image on attacker-supplied storage (SD card, USB mass storage, or otherwise mounted image) triggers it. Affected: Zephyr ext2 from its introduction in v3.5.0 through v4.4.0. The fix validates rec_len and name_len in the parser and rejects entries whose header does not fit the remaining block or whose rec_len crosses the block boundary in every traversal caller.

This has been fixed in main for v4.5.0

CVE 2026-10646

Use-after-return in zsock_getaddrinfo() when a timed-out DNS query is retried without cancellation

Zephyr’s BSD-sockets getaddrinfo() implementation (subsys/net/lib/sockets/getaddrinfo.c) passes a pointer to a stack-allocated state object (struct getaddrinfo_state ai_state) as the user_data of an asynchronous DNS resolver query. The socket layer waits on a semaphore with a timeout deliberately set slightly longer than the resolver’s own per-query timeout. When that semaphore wait nonetheless times out (-EAGAIN) - which can occur when the resolver’s timeout work is delayed by workqueue contention, or in the documented multi-retry configuration where CONFIG_NET_SOCKETS_DNS_TIMEOUT exceeds CONFIG_NET_SOCKETS_DNS_BACKOFF_INTERVAL - the pre-fix code retries the query (goto again) without cancelling the previous one and without resetting the semaphore.

The previous query slot remains active in the resolver with its callback and the stack pointer as user_data, and ai_state->dns_id is overwritten so the stale query can no longer be cancelled. A subsequent DNS response delivered over UDP and matched by its 16-bit transaction id (in dispatcher_cb()/dns_read()), or the resolver’s own delayed query-timeout work, then invokes dns_resolve_cb() against the now out-of-scope stack frame, writing through the stale pointer (state->status, state->idx, state->ai_arr[], and k_sem_give()).

Because the triggering response is network-delivered and its 16-bit id is spoofable/replayable by an on- or off-path attacker, this is a network-influenceable use-after-return that can corrupt reused stack memory, leading to crashes/denial of service or memory corruption.

The fix cancels the timed-out query by name and type before retrying and resets the local semaphore, eliminating the stale callback path. Affected: Zephyr v4.0.0 through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10647

Deadlock denial of service in USB CDC-NCM device class on TX enqueue failure

The USB CDC-NCM device class (subsys/usb/device_next/class/usbd_cdc_ncm.c) ignores the return value of usbd_ep_enqueue() in its ethernet transmit callback cdc_ncm_send(). When the enqueue fails, the function still calls k_sem_take(&data->sync_sem, K_FOREVER), blocking on a completion semaphore that is only ever signaled from the bulk-IN transfer-completion callback. Because nothing was enqueued, that callback never fires and the calling thread — a shared network traffic-class TX thread — deadlocks permanently while holding the interface TX lock, halting transmission until reboot (and leaking the transmit buffer).

The enqueue fails under conditions controlled by the attached USB host: usbd_ep_enqueue() returns -EPERM whenever the bus is suspended (a standard, persistent host operation), and the underlying udc_ep_enqueue() returns -EPERM/-ENODEV on disconnect, bus reset, or endpoint disable. The cdc_ncm_send() guard only checks the DATA_IFACE_ENABLED and IFACE_UP flags, not the suspended state, so a packet transmitted while the host holds the bus suspended reaches the failing enqueue and deadlocks the TX path.

The realistic trigger is a bus suspend that occurs while the exported network interface is active and has traffic to send — host sleep, USB selective/auto-suspend, or hub power management — after which any device-originated packet deadlocks the path, recoverable only by reboot. The impact is a persistent loss of the virtual network connection between the host’s NCM interface and the Zephyr device; because the deadlocked thread is a shared traffic-class TX thread, egress on other network interfaces can stall as well. There is no memory corruption or information disclosure.

The defect was introduced with the CDC-NCM driver and shipped in releases through v4.4.0; it is fixed by checking the usbd_ep_enqueue() return value and freeing the buffer before the blocking wait.

This has been fixed in main for v4.5.0

CVE 2026-10648

NULL-pointer dereference in MCUmgr serial/console SMP transport on buffer-pool exhaustion

mcumgr_serial_process_frag() in subsys/mgmt/mcumgr/transport/src/serial_util.c calls net_buf_reset() on the result of smp_packet_alloc() before checking it for NULL. smp_packet_alloc() uses net_buf_alloc(K_NO_WAIT) against the shared MCUmgr packet pool (CONFIG_MCUMGR_TRANSPORT_NETBUF_COUNT, default 4), which returns NULL when the pool is exhausted. In default builds the __ASSERT_NO_MSG in net_buf_reset is a no-op, so net_buf_simple_reset writes through the NULL pointer (buf->len = 0; buf->data = buf->__buf), causing a fault/crash.

The fragment data reaches this code from attacker-controlled bytes on the MCUmgr serial/UART/shell-console transports (smp_uart.c, smp_raw_uart.c, smp_shell.c), and a fresh buffer is allocated at the start of essentially every new packet. An attacker on the serial/console link can flood the transport to drive the 4-entry buffer pool to exhaustion and induce the NULL dereference, crashing the device (denial of service).

The defect was introduced after the original MCUmgr rework and shipped in Zephyr v4.4.0. The fix moves the NULL check ahead of net_buf_reset.

This has been fixed in main for v4.5.0

CVE 2026-10651

Out-of-bounds read in Bluetooth Classic SDP attribute parsing (bt_sdp_parse_attribute)

bt_sdp_parse_attribute() in subsys/bluetooth/host/classic/sdp.c validated only that the SDP record buffer held the type-marker byte plus the 2-byte attribute ID (a check of buf->len < 3) but then read a fourth byte, the data-element descriptor (type), via net_buf_simple_pull_u8(). Because net_buf_simple_pull_u8() dereferences buf->data[0] before its only bounds guard (an __ASSERT_NO_MSG that compiles out when CONFIG_ASSERT is disabled, the production default), a record of exactly three bytes (0x09 followed by a 2-byte attribute ID) causes a one-byte read past the end of the logical buffer. The parser is reachable from inbound, remote-controlled data: a Bluetooth BR/EDR peer acting as an SDP server returns discovery-response records that are stored verbatim in the client receive buffer and parsed via the public bt_sdp_get_attr()/bt_sdp_has_attr()/bt_sdp_record_parse() helpers. The over-read is bounded to a single byte that is used only as an internal length selector and is never leaked to the attacker; subsequent length checks then reject the malformed record. Realistic impact is therefore limited to an edge-case denial of service (a fault only if the record ends exactly at a mapped-memory boundary, or a deterministic assert panic when CONFIG_ASSERT=y). Affects Zephyr v4.3.0 and v4.4.0; fixed by adding sizeof(type) to the length check.

This has been fixed in main for v4.5.0

CVE 2026-10652

Out-of-bounds read in Zephyr DNS resolver TXT/SRV record parsing (unvalidated rdlength)

Zephyr’s DNS resolver (subsys/net/lib/dns) parses resource records from DNS responses in dns_unpack_answer(), which validated only the fixed RR header (type, class, TTL, rdlength) and accepted any attacker-declared rdlength, including one extending past the end of the received datagram. The TXT and SRV consumers in dns_validate_record() (resolve.c) then read up to rdlength bytes (clamped only to a record-type maximum such as DNS_MAX_TEXT_SIZE, default 64, not to the packet) from the receive buffer via memcpy without their own bounds check, and pass the result to the application’s resolve callback. A malicious or spoofed DNS server, an on-path attacker forging UDP DNS replies, or (with mDNS/LLMNR enabled) any LAN node can craft a truncated TXT or SRV response that causes an out-of-bounds read of adjacent receive-pool memory; the disclosed stale bytes (residual contents of prior DNS packets / uninitialized pool memory) are returned to the application as TXT/SRV record contents, an information leak, and may in some configurations cross the allocation boundary and fault, causing a denial of service. The read is bounded (~64 bytes for TXT, ~6 for SRV) and read-only (no write). The fix rejects any record whose declared rdata extends past dns_msg->msg_size at the single chokepoint in dns_unpack_answer(). Affected: v4.3.0 and v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10653

Non-atomic net_buf reference counts cause double-free / free-list corruption under concurrent unref

The Zephyr net_buf library (lib/net_buf/buf.c) manipulated both of its reference counts – the per-header buf->ref and the per-data-block ref_count at the start of each variable/heap data allocation – with plain non-atomic C operators (buf->ref++, if (--buf->ref > 0), if (--(*ref_count))).

The API is documented as self-synchronizing: callers may share one buffer across threads (e.g. via k_fifo) and each holder independently calls net_buf_unref() with no surrounding lock. Under true concurrency (SMP, or single-core preemption between the non-atomic load and store while another context unrefs the same buffer), two holders can both observe the same prior reference value and both conclude they are the last reference.

For heap/variable-data pools (mem_pool_data_unref/heap_data_unref, used by zbus message subscribers, the IP stack RX/TX buffers when CONFIG_NET_BUF_FIXED_DATA_SIZE=n, capture, wireguard, ISO-TP and usbip) this produces a double k_heap_free()/k_free() of the same block – heap-metadata corruption and a use-after-free on the heap-hardening poison pattern.

For the per-header refcount the buffer is returned to the pool free LIFO twice for any pool type (including fixed-data pools used by Bluetooth and networking), corrupting the free list so a later allocation hands the same buffer to two owners.

The fix converts both refcounts to atomic_inc/atomic_dec (overlaying buf->ref in an atomic_t-sized union and changing the data-block refcount from uint8_t to atomic_t).

Impact is gated on genuine concurrency and on an application architecture that shares one buffer among multiple independent unref’ers; the trigger is a refcount/timing race rather than packet content, so an external attacker has at most weak indirect influence over the race window. Affects all Zephyr releases through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10654

RFCOMM session-disconnect race leaks session/L2CAP and denies further RFCOMM service in Zephyr Bluetooth Classic

A race condition in the Zephyr Bluetooth Classic RFCOMM host stack (subsys/bluetooth/host/classic/rfcomm.c) mishandles a simultaneous bidirectional session disconnect. When the local device has initiated a session teardown (state BT_RFCOMM_STATE_DISCONNECTING, DISC sent, RTX timer armed) and the connected peer concurrently sends its own DISC frame for dlci 0, rfcomm_handle_disc() invokes rfcomm_session_disconnected(), which unconditionally forced the session to BT_RFCOMM_STATE_DISCONNECTED without ever calling bt_l2cap_chan_disconnect().

Because the recovery timer was also cancelled and a later UA is ignored in the DISCONNECTED state, the session becomes permanently wedged: the underlying L2CAP channel is never released and the session slot in the fixed bt_rfcomm_pool[CONFIG_BT_MAX_CONN] array is never reclaimed (its conn pointer stays set).

Subsequent bt_rfcomm_dlc_connect() calls on that connection fail with -EINVAL due to the invalid session state, so RFCOMM service is denied for that peer, and repeated occurrences can exhaust the session pool. The DISC frame is peer-controlled over the air, but exploitation requires the peer’s DISC to collide with a local-initiated disconnect (a high-complexity timing race). Impact is availability/resource-leak only; there is no memory-safety, confidentiality, or integrity consequence. The defect shipped in released versions (present in v4.4.0 and earlier).

The fix only transitions to DISCONNECTED when the session is not already in DISCONNECTING, preserving the proper L2CAP teardown path.

This has been fixed in main for v4.5.0

CVE 2026-10655

Use-after-free race in SNTP async client when closing the socket while the socket service is still polling it

The asynchronous SNTP client in Zephyr (subsys/net/lib/sntp/sntp.c, sntp_close_async) closed the UDP socket file descriptor directly from the calling thread immediately after detaching it from the network socket service, without synchronizing with the socket-service poll thread.

The socket service thread polls each socket via zvfs_poll, which (in zsock_poll_prepare_ctx) registers a k_poll_event pointing into the socket’s net_context (&ctx->recv_q) and then blocks in k_poll without holding a reference or lock. net_context objects are allocated from a fixed pool (contexts[CONFIG_NET_MAX_CONTEXTS]) and reused after close.

When sntp_close_async is invoked from a different thread than the poll thread (in the in-tree consumer subsys/net/lib/config/init_clock_sntp.c, the SNTP timeout handler runs on the system workqueue while the socket service thread is blocked in poll on the same fd), the close frees and may reuse the net_context while the poll thread still has a poller node linked into the freed object, resulting in a use-after-free / object confusion of kernel poll structures.

The SNTP timeout path is the normal no-response failure mode, so a network peer or off-path attacker who drops or delays the SNTP/NTP response can drive the racing close repeatedly (and periodically with NET_CONFIG_SNTP_INIT_RESYNC). The most likely consequence is a crash of the networking thread (denial of service), with potential memory corruption when the freed context slot is reallocated.

The fix defers the close to the socket service thread itself via net_socket_service_close (NET_SOCKET_SERVICE_CLOSE_SOCKETS), so the same thread that polls performs the close, eliminating the race. Affected releases: v4.2.0 through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10656

NULL-pointer dereference DoS in MAX32 USB device controller transfer-completion handlers

The MAX32xxx USB device controller driver (drivers/usb/udc/udc_max32.c, compatible adi_max32_usbhs) dereferenced an endpoint buffer in its OUT and IN transfer-completion handlers without checking it for NULL. udc_event_xfer_out_done() called net_buf_add(buf, ep_request->actlen) immediately after buf = udc_buf_get(ep_cfg), where udc_buf_get() returns NULL when the endpoint FIFO is empty.

A transfer-completion event is queued from interrupt context and processed asynchronously by the driver thread; between queuing and processing, the endpoint FIFO can be drained by host-controlled control flow — in particular udc_setup_received() drains the EP0 OUT/IN FIFOs whenever a new SETUP packet arrives, and dequeue/disable/purge paths drain it likewise.

A USB host that aborts an in-flight EP0 control transfer with a new SETUP packet (legal USB behavior) can therefore cause a stale XFER_OUT_DONE event to be processed against an empty FIFO, producing net_buf_add(NULL, ...), a near-NULL pointer dereference that faults and crashes the device. No authentication is required; the attacker is the USB host the device is connected to (physical bus access). Impact is denial of service (device crash).

The defect was introduced when the MAX32 UDC driver was added and shipped in Zephyr v4.4.0. The fix adds NULL-buffer checks that return early with UDC_EVT_ERROR/-ENOBUFS in both the OUT-done and IN-done handlers.

This has been fixed in main for v4.5.0

CVE 2026-10657

Out-of-bounds read in Zephyr DNS resolver mDNS suffix check (memcmp past string NUL)

Zephyr’s DNS resolver detects mDNS (.local) queries in dns_resolve_name_internal() (subsys/net/lib/dns/resolve.c) with memcmp(strrchr(query, '.'), ".local", 7), which always reads a fixed 7 bytes from the suffix pointer. When the resolved hostname’s final label is shorter than 7 bytes (e.g. names ending in .org, .com, .net, .io, or a trailing dot), the comparison reads 1-2 bytes past the string’s NUL terminator.

The hostname (query) is the caller-supplied name passed through the standard getaddrinfo()/dns_get_addr_info()/dns_resolve_name() path and is influenceable by operators or remote inputs (server names from configuration, parsed URLs, or app-facing interfaces).

On a tightly-sized buffer with no slack (for example a userspace getaddrinfo call where the hostname is copied with k_usermode_string_alloc_copy to exactly strlen+1 bytes), the over-read crosses the allocation boundary; if that boundary is unmapped (guard page, memory-domain boundary under MPU, or an address sanitizer) the over-read faults, causing a denial of service. The over-read bytes are never returned, so there is no information disclosure.

The flaw is compiled only when CONFIG_MDNS_RESOLVER is enabled, exists since v1.10.0, and is fixed by replacing the fixed-length memcmp with a NUL-safe strcmp(ptr, ".local").

This has been fixed in main for v4.5.0

CVE 2026-10658

Out-of-bounds access in Bluetooth ISO receive (bt_iso_recv) due to missing SDU-header length validation

bt_iso_recv() in subsys/bluetooth/host/iso.c pulled the ISO SDU header (4 bytes) or, when the timestamp flag is set, the timestamped SDU header (8 bytes) from the inbound HCI ISO Data buffer via net_buf_pull_mem() without first checking buf->len. The upstream hci_iso() handler enforces buf->len == the controller-declared ISO Data_Load length, so a malicious or buggy controller / adjacent BLE peer on an established CIS/BIS can present a first-fragment (BT_ISO_START) or single (BT_ISO_SINGLE) PDU shorter than the SDU header. Because net_buf_simple_pull_mem only guards length with __ASSERT_NO_MSG (compiled out when CONFIG_ASSERT is disabled, the production default), the pull underflows buf->len (uint16_t, e.g. 0 - 8 = 0xFFF8) and advances buf->data past valid data: the subsequent reads of hdr->slen and hdr->sn are out-of-bounds reads of adjacent pool memory. For the multi-fragment (START) case the corrupted buffer is retained as iso->rx, and a following CONT/END fragment’s net_buf_tailroom() guard underflows to a near-SIZE_MAX value, defeating the bounds check and causing net_buf_add_mem() to memcpy attacker-supplied fragment data far past the RX pool buffer (out-of-bounds write). The flaw affects ISO receive builds (CONFIG_BT_ISO_RX, selected by the default-off LE Audio options BT_ISO_PERIPHERAL/BT_ISO_CENTRAL/BT_ISO_SYNC_RECEIVER) and has existed since the ISO subsystem was introduced (v2.6.0) through v4.4.0. The fix adds explicit buf->len < sizeof(*ts_hdr) and buf->len < sizeof(*hdr) checks that drop the buffer before pulling.

This has been fixed in main for v4.5.0

CVE 2026-10659

NULL pointer dereference in Zephyr Dhara FTL disk driver on flash read error during journal resume

The Dhara flash translation layer disk driver (drivers/disk/ftl_dhara.c) implemented the dhara_nand_* callbacks so that, on a flash error, the error code was written unconditionally through the caller-supplied dhara_error_t *err pointer (e.g. *err = DHARA_E_ECC in dhara_nand_read, and similar in dhara_nand_erase/prog/copy).

The upstream Dhara library calls these callbacks with err == NULL along its journal-resume binary search: find_last_checkblock() invokes find_checkblock(j, mid, &found, NULL), which forwards the NULL pointer into dhara_nand_read(). This path runs during disk_ftl_access_init() -> dhara_map_resume() whenever the FTL disk is mounted/initialised.

If a flash read error (uncorrectable ECC, bad block, controller error) occurs on one of the probed checkpoint pages, the driver dereferences and writes to NULL, faulting the kernel (denial of service). The trigger is conditioned on the NAND medium content/health, which can be influenced by media wear, induced faults, or a corrupted/crafted on-flash image.

The fix routes all error assignments through the library’s NULL-safe dhara_set_error() helper. Affects Zephyr v4.4.0, where the driver was introduced.

This has been fixed in main for v4.5.0

CVE 2026-10660

Shared reassembly buffer in Bluetooth BAP Broadcast Assistant enables cross-connection memory corruption

The Bluetooth BAP Broadcast Assistant GATT client in subsys/bluetooth/audio/bap_broadcast_assistant.c reassembled remote Broadcast Receive State data into a single file-static net_buf_simple (att_buf, BT_ATT_MAX_ATTRIBUTE_LEN = 512 bytes) shared by all connection instances, while the BUSY flag, long-read handle, and reset/offset state were per-connection.

When the device acts as a Broadcast Assistant connected to multiple Scan Delegator peripherals, notification and long-read callbacks from different connections interleave on the shared buffer: the append in notify_handler (net_buf_simple_add_mem at the not-busy branch) performs no tailroom check, so receive-state notifications from two or more delegators accumulate on the same 512-byte buffer and, with a sufficiently large configured ATT MTU (BT_L2CAP_TX_MTU up to 2000) and two-to-three concurrent connections, write past the buffer into adjacent .bss (net_buf_simple_add only asserts in debug builds).

Even below the overflow threshold, one connection’s net_buf_simple_reset zeroes the shared length while another connection’s reassembly and GATT read offset are in flight, mixing one peer’s data into another’s parse. A malicious or compromised Scan Delegator (or two colluding peers) over BLE can trigger this, causing out-of-bounds writes (memory corruption / denial of service) and cross-connection data corruption.

The fix moves the buffer into the per-connection instance struct so each connection reassembles into its own buffer. Affects Zephyr releases shipping the Broadcast Assistant with the shared buffer, including v4.4.0 and earlier.

This has been fixed in main for v4.5.0

CVE 2026-10663

Use-after-free / double-free of the root USB device in the experimental USB host stack

In Zephyr’s experimental USB host stack (CONFIG_USB_HOST_STACK), usbh_device_disconnect() (subsys/usb/host/usbh_device.c) freed the root usb_device slab object without clearing the cached pointer ctx->root. The bus removal handler dev_removed_handler() (subsys/usb/host/usbh_core.c) decides what to tear down solely from ctx->root, checking only that it is non-NULL.

Because UHC controller drivers (e.g. uhc_max3421e, uhc_mcux_common) synthesize UHC_EVT_DEV_REMOVED directly from physical bus line state with no debounce or state guard, an attacker with physical USB access (or a rogue device that bounces its connection) can deliver a second device-removed event after a root device disconnect. The handler then re-enters usbh_device_disconnect() with the dangling pointer, locking a mutex inside the freed object (use-after-free), removing the freed node from the device list, and calling k_mem_slab_free() on the already-freed block (double-free). If the slab block has been reissued to a newly attached device in between, this corrupts a live object.

Impact is denial of service (crash) and memory corruption; the attack vector is physical/local. The flaw was introduced in v4.4.0 by the connect/disconnect refactor and is fixed by clearing ctx->root in usbh_device_disconnect() before freeing.

This has been fixed in main for v4.5.0

CVE 2026-10664

Out-of-bounds write in nRF70 Wi-Fi driver power-save event handler (unbounded TWT flow count)

The nRF70 Wi-Fi driver’s power-save event handler nrf_wifi_event_proc_get_power_save_info() in drivers/wifi/nrf_wifi/src/wifi_mgmt.c copied TWT (Target Wake Time) flow entries from an nrf_wifi_umac_event_power_save_info event into the fixed-size twt_flows[WIFI_MAX_TWT_FLOWS] (8-element) array of a caller-supplied struct wifi_ps_config, looping over event-provided num_twt_flows without validating it against WIFI_MAX_TWT_FLOWS or checking event_len. When num_twt_flows exceeds 8, the handler writes past the destination array (which is typically on the caller’s stack, e.g. the wifi ps shell command) – an out-of-bounds write of ~40-byte TWT entries – and reads twt_flow_info[i] past the event buffer. The event is delivered by the nRF70 co-processor firmware in response to a host-initiated power-save GET, so reaching the overflow requires the firmware to emit a malformed or out-of-range event; the trust boundary is host-to-trusted-coprocessor rather than a direct remote-AP write, with over-the-air influence on the flow count being indirect and bounded by the 3-bit TWT flow-id space. Affected: builds with CONFIG_NRF70_STA_MODE on releases through v4.4.0. The fix rejects events with num_twt_flows > WIFI_MAX_TWT_FLOWS or with event_len shorter than the claimed entries, and adds a NULL check on the caller buffer.

This has been fixed in main for v4.5.0

CVE 2026-10665

Heap buffer overflow on WireGuard receive path via unbounded incoming packet length

In Zephyr’s WireGuard subsystem (subsys/net/lib/wireguard), wg_process_data_message() in wg_crypto.c linearizes an inbound transport-data payload into a fixed pool buffer of CONFIG_WIREGUARD_BUF_LEN bytes before decryption. The call net_buf_linearize(buf->data, data_len, pkt->buffer, ..., data_len) passed the attacker-derived data_len as both the destination capacity and the copy length, defeating the function’s internal len = min(len, dst_len) bound. data_len is derived from the received UDP datagram length and is only lower-bounded by wg_ctrl_recv() (no upper bound). When data_len exceeds CONFIG_WIREGUARD_BUF_LEN — e.g. when the buffer length is lowered below the link MTU, on links with MTU above the buffer size, or via reassembled IPv4/IPv6 fragments that exceed it — the underlying memcpy writes past the end of the pool buffer, an out-of-bounds write (CWE-787). The overflow occurs before the Poly1305 authentication check, so it requires only a valid receiver session index rather than a valid authenticator, and is reachable by a malicious or compromised peer (or an on-path attacker driving an established session) over the network, yielding remote memory corruption and at minimum a reliable denial of service. The defect was present in the WireGuard implementation shipped in Zephyr 4.4.0. The fix adds an explicit data_len > CONFIG_WIREGUARD_BUF_LEN rejection and corrects the linearize call to pass net_buf_max_len(buf) as the destination capacity.

This has been fixed in main for v4.5.0

CVE 2026-10667

SMP use-after-free in Zephyr CONFIG_USERSPACE dynamic kernel-object tracking, reachable from unprivileged user threads

Zephyr’s dynamic kernel-object tracking (kernel/userspace/userspace.c, formerly kernel/userspace.c) maintains a doubly-linked list (obj_list) of dynamically allocated kernel objects. Iteration over this list in k_object_wordlist_foreach() was performed under lists_lock using the SAFE iterator (which caches the next node), but list removal and freeing of nodes was performed under different, disjoint spinlocks: objfree_lock in k_object_free() and obj_lock in unref_check(). On an SMP system, while one CPU iterated obj_list under lists_lock, another CPU could unlink and k_free() the dyn_obj node that the iterator had cached as its next pointer, causing the iterator to dereference freed kernel memory (use-after-free / dangling list traversal). All of the racing operations are reachable from unprivileged user-mode threads via system calls: k_object_alloc/k_object_alloc_size and k_object_release drive removals through unref_check() (under obj_lock), while k_thread_abort and thread creation drive the iteration through k_thread_perms_all_clear()/k_thread_perms_inherit() (under lists_lock). A deprivileged user thread on a CONFIG_SMP + CONFIG_USERSPACE build can therefore corrupt the kernel’s object-tracking structures across the userspace security boundary, yielding kernel memory corruption (potential privilege escalation) or a kernel crash (denial of service). The fix removes objfree_lock and serializes every obj_list modification under lists_lock, including holding it across find+remove in k_object_free() and around unref_check() in k_thread_perms_clear(). Affects CONFIG_SMP``+``CONFIG_USERSPACE``+``CONFIG_DYNAMIC_OBJECTS configurations; the defect dates to the 2019 spinlockification (commit 8a3d57b6cc6, first released in v1.14.0) and shipped through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10668

Host-triggerable control-endpoint wedge (DoS) in Nuvoton NuMaker HSUSBD UDC driver

The Nuvoton NuMaker HSUSBD USB device-controller driver (drivers/usb/udc/udc_numaker.c) armed the control Data IN stage unconditionally (base->CEPTXCNT = len in numaker_hsusbd_ep_trigger). Because the HSUSBD hardware cannot disarm a control Data IN already armed for a previous transfer, a USB host that cancels an in-flight control transfer (timeout) and then issues a new SETUP packet can drive the driver out of sync: stale data may be transmitted in the new transfer and the control endpoint can become permanently stuck NAK’ing every subsequent control transfer.

A malicious or buggy host (physical/adjacent attacker driving the bus) can repeatedly cancel-and-re-SETUP to wedge the device’s USB control endpoint, denying service to the device’s USB function (the device stops enumerating/responding on the control pipe) until a USB reset or re-plug. The flaw is an availability-only denial of service; the FIFO copy loops (bounded by net_buf length and the hardware BUFFULL flag) and the net_buf lifecycle are independent of the arming desync, so there is no out-of-bounds access, use-after-free, or information leak.

The fix monitors the IN-token and new-SETUP events (k_event) and only arms control Data IN when an IN token is present and no new SETUP has arrived, cancelling the current transfer on a new SETUP. Affects boards using the Nuvoton NuMaker HSUSBD controller (CONFIG_UDC_NUMAKER with DT_HAS_NUVOTON_NUMAKER_HSUSBD_ENABLED); shipped in v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10669

Xtensa MPU arch_buffer_validate() integer-overflow lets a user thread bypass syscall pointer validation

On Xtensa SoCs built with CONFIG_XTENSA_MPU and CONFIG_USERSPACE, arch_buffer_validate() in arch/xtensa/core/mpu.c — the architecture hook that verifies a user-mode-supplied buffer is accessible to the calling user thread with the requested permission — defaulted its return value to 0 (access permitted) and only set a denial result inside its per-MPU-region probe loop. When the rounded extent of the buffer wraps the 32-bit address space (size + alignment offset near SIZE_MAX, or ROUND_UP(size + offset) overflowing to 0), the loop executes zero iterations and the function returns 0 = permitted without probing any MPU region.

The syscall-layer pre-checks (K_SYSCALL_MEMORY_SIZE_CHECK / Z_DETECT_POINTER_OVERFLOW) only catch a raw addr+size wrap and do not cover the ROUND_UP-induced wrap, and the string path (arch_user_string_nlen -> arch_buffer_validate) has no syscall-layer guard at all.

An unprivileged user-mode thread can therefore pass a crafted (addr, size) to any syscall that validates user buffers via k_usermode_from_copy/to_copy or k_usermode_string_copy and have validation succeed for memory it must not access; the kernel then reads from (disclosure) or, with write=1, writes to (corruption) attacker-chosen kernel or other-partition memory on the thread’s behalf, enabling information disclosure, memory corruption, privilege escalation, and denial of service.

Affected from v3.7.0 (when Xtensa MPU userspace support was added) through v4.4.0. The fix changes the default to -EINVAL (deny by default), adds an explicit size_add_overflow check, and sets the success value only after the full range has been validated.

This has been fixed in main for v4.5.0

CVE 2026-10670

User-triggerable kernel NULL-pointer dereference (DoS) in k_thread_name_copy() syscall verifier

The CONFIG_USERSPACE verification handler for the k_thread_name_copy() system call (z_vrfy_k_thread_name_copy() in kernel/thread.c) calls k_object_find() on the caller-supplied thread pointer and then dereferences the returned struct k_object without checking it for NULL. k_object_find() returns NULL whenever the supplied pointer is not a registered (static or dynamic) kernel object.

The pre-fix guard tested thread == NULL instead of ko == NULL, so an unprivileged user-mode thread that invokes k_thread_name_copy() with any non-NULL but unregistered pointer (e.g. an arbitrary address) passes the NULL test, after which the verifier reads ko->type through a NULL pointer.

Because the syscall verifier runs in supervisor mode, this NULL dereference is a kernel-mode fault that halts or reboots the system, allowing untrusted user code to crash the kernel across the userspace security boundary (denial of service). The marshaller passes the thread argument to the verifier without any prior K_SYSCALL_OBJ validation, so the bad pointer reaches the defect directly.

The flaw affects builds with CONFIG_USERSPACE and CONFIG_THREAD_NAME enabled and has been present since the special-case lookup was introduced around v2.0.0; it is present in v4.4.0 and earlier. The fix changes the guard to check the k_object_find() return value (ko == NULL) before dereferencing it.

This has been fixed in main for v4.5.0

CVE 2026-10671

User thread can re-initialize an in-use k_pipe, corrupting kernel wait queues (CONFIG_USERSPACE)

In Zephyr’s kernel pipe implementation, the userspace syscall verifier z_vrfy_k_pipe_init() in kernel/pipe.c used K_SYSCALL_OBJ() (which requires the kernel object to already be initialized) instead of K_SYSCALL_OBJ_NEVER_INIT() (which rejects an already-initialized object). As a result, on CONFIG_USERSPACE builds an unprivileged user thread that has been granted access to a k_pipe object can invoke the k_pipe_init syscall to re-initialize a pipe that is already in use.

z_impl_k_pipe_init() unconditionally resets the ring buffer, sets pipe->waiting to 0, and re-initializes both wait queues (z_waitq_init on pipe->data and pipe->space) without waking or accounting for threads currently blocked on the pipe. Any thread already pended in k_pipe_read()/k_pipe_write() is left orphaned: still marked pending with pended_on pointing at the cleared wait queue and with stale qnode_dlist links into the (now re-initialized) embedded list head.

When such an orphaned waiter is later timed out or woken, the scheduler calls sys_dlist_remove() on its stale node, writing through dangling prev/next pointers into kernel wait-queue/scheduler structures, causing list corruption (an attacker-driven invalid kernel write), lost wakeups, indefinitely blocked threads, and silent data loss. The flaw lets a deprivileged user thread corrupt the state of a kernel object shared with other threads/partitions.

The fix switches the verifier to K_SYSCALL_OBJ_NEVER_INIT(), matching the existing k_msgq_init verifier, so a user thread can no longer re-initialize a live pipe. The vulnerable code shipped in v4.1.0 and remained through v4.4.0.

This has been fixed in main for v4.5.0

CVE 2026-10672

Unterminated URI buffer causes out-of-bounds read in LwM2M firmware pull (Package URI)

subsys/net/lib/lwm2m/lwm2m_pull_context.c copied the firmware-update Package URI into a fixed static buffer (context.uri, size CONFIG_LWM2M_SWMGMT_PACKAGE_URI_LEN, default 128) with memcpy(context.uri, uri, LWM2M_PACKAGE_URI_LEN), copying exactly the destination size with no length validation. The Firmware-Update object stores the server-supplied Package URI (/5/0/1) in a 255-byte buffer, so a LwM2M management server (or an on-path attacker on a session lacking strong DTLS) can WRITE a URI of 128-254 characters; only the first 128 bytes are then copied into context.uri with no NUL terminator. That buffer is subsequently consumed as a C string by http_parser_parse_url(context.uri, strlen(context.uri), ...), strlen-based CoAP URI-path/PROXY-URI option appends, and lwm2m_parse_peerinfo(), causing an out-of-bounds read of adjacent static memory. The over-read bytes are appended to outbound CoAP requests (information disclosure of adjacent device memory to the server/proxy) and can crash the device (denial of service). The vulnerable copy was introduced by the pull-context refactor (first released in v3.0.0) and is present through v4.4.0; the default-on CONFIG_LWM2M_FIRMWARE_UPDATE_PULL_SUPPORT path is affected. The fix adds a strlen(uri) >= sizeof(context.uri) check returning -ENOMEM and switches to strcpy(), guaranteeing a bounded, NUL-terminated buffer.

This has been fixed in main for v4.5.0

CVE 2026-10674

DoS (hard fault) in NXP LPUART driver: unsupported runtime UART config leaves clocks disabled

The NXP LPUART serial driver (drivers/serial/uart_mcux_lpuart.c), when CONFIG_UART_USE_RUNTIME_CONFIGURE is enabled, called LPUART_Deinit() at the start of mcux_lpuart_configure(), which disables the LPUART peripheral clocks. The requested configuration is validated only afterwards (in mcux_lpuart_configure_basic), and unsupported parity/data-bit/stop-bit/flow-control values return -ENOTSUP before the clock is re-enabled.

As a result, a uart_configure() request with an unsupported configuration left the LPUART in a clock-disabled state; any subsequent access to LPUART registers (poll_out/poll_in, interrupt handling, or a later reconfigure) faults on the gated peripheral and escalates to a hard fault, crashing the system.

uart_configure() is a Zephyr syscall whose verifier (z_vrfy_uart_configure) only checks that cfg is readable user memory and forwards the caller-supplied configuration unchanged, so an unprivileged userspace thread with access to an LPUART device can deterministically trigger the fault, a persistent system-wide denial of service.

Introduced in v2.5.0 and present in all subsequent releases until this fix, which removes the LPUART_Deinit() call and instead only disables the transmitter/receiver, leaving the clock running.

This has been fixed in main for v4.5.0

CVE 2026-10675

Bluetooth Mesh PB-ADV: invalidated provisioning link kept alive indefinitely, blocking (re)provisioning (DoS)

In Zephyr’s Bluetooth Mesh PB-ADV provisioning bearer (subsys/bluetooth/mesh/pb_adv.c), prov_msg_recv() rescheduled the provisioning protocol watchdog timer unconditionally at the top of the function, before the FCS check and before the ADV_LINK_INVALID check. Once a provisioning attempt fails, prov_failed() sets ADV_LINK_INVALID and the only recovery path is the protocol timer firing (protocol_timeout -> prov_link_close -> close_link -> reset_adv_link and re-enabling of scanning and the unprovisioned device beacon).

A remote, unauthenticated attacker on the BLE advertising channel can first induce a provisioning failure (e.g. with a malformed generic-provisioning PDU) and then transmit any FCS-valid PB-ADV transaction PDU on the same link ID more often than once per protocol timeout (60 s, or 120 s for OOB input/output). Because each such packet reset the timer even on an invalidated link, protocol_timeout never fired, the dead link was never torn down, and the device remained pinned in an un-provisionable state with its unprovisioned beacon disabled and new Link Open requests rejected.

PB-ADV PDUs are processed without authentication and the FCS is a keyless CRC, so no pairing or prior trust is required and the attacker chooses the link ID itself. The impact is a persistent denial of provisioning/re-provisioning service; there is no memory-safety, confidentiality, or integrity impact.

The vulnerable code shipped in releases through v4.4.1. The fix moves the timer reschedule to after the ADV_LINK_INVALID check (and the FCS check before the reset) so an invalidated link can no longer be kept alive by incoming packets.

This has been fixed in main for v4.5.0

CVE 2026-10677

Kernel heap memory leak in z_vrfy_k_poll() lets an unprivileged user thread exhaust the kernel resource pool

The CONFIG_USERSPACE syscall verifier z_vrfy_k_poll() in kernel/poll.c allocates a kernel-side copy of the user-supplied k_poll_event[] via z_thread_malloc() and then validates each event’s object handle. Before this fix, validation used K_OOPS(K_SYSCALL_OBJ(...)) inline inside the loop, which kills the calling thread without freeing events_copy.

A user thread can pass num_events >= 1 with a forged object handle to leak the allocation; because newly spawned user threads inherit the parent’s resource_pool (kernel/thread.c), an attacker spawns sacrificial threads to repeat the leak until the shared kernel heap is exhausted. Once depleted, legitimate kernel allocations from that pool (k_queue alloc nodes, k_msgq buffers, future k_poll calls, etc.) fail, causing a system-level denial of service.

The fix replaces each inline K_OOPS with a conditional goto oops_free so the buffer is freed before the thread is killed. Affects Zephyr releases from v1.12.0 (when k_poll was first exposed to user mode) through v4.4.1.

This has been fixed in main for v4.5.0

CVE 2026-10678

NULL-pointer / out-of-bounds write in Zephyr MCTP I2C+GPIO target binding driven by an unauthenticated I2C controller

The MCTP-over-I2C+GPIO target binding in Zephyr (subsys/pmci/mctp/mctp_i2c_gpio_target.c) processes pseudo-register writes from an I2C bus master byte-by-byte in mctp_i2c_gpio_target_write_received() without validating the order or the receive buffer. In the affected versions the MCTP_I2C_GPIO_RX_MSG_ADDR (data) handler dereferences and writes through b->rx_pkt without checking that the receive buffer was allocated: a controller that selects the data register and writes a byte without first sending the length register (which is what allocates the buffer) causes a write of an attacker-chosen byte through a NULL/unallocated mctp_pktbuf pointer (i.e. into a small attacker-advanceable offset above address 0), producing memory corruption or a hard fault.

The same handler also performs a write-then-check bounds test, allowing a one-byte heap overflow at data[255] when more than 255 data bytes are sent.

Because the I2C target callback is invoked with raw bytes supplied by whatever device is the bus master and the binding performs no authentication, a malicious or malfunctioning controller on the bus can trigger these without any prior protocol state, leading to memory corruption and/or denial of service on the target device.

The vulnerable code was introduced when the I2C+GPIO target binding was added and shipped in Zephyr v4.3.0 and v4.4.0. The fix defers allocation to the first data byte with a NULL check, treats a missing length as a zero-sized packet rejected by libmctp, and moves the bounds check before the store.

This has been fixed in main for v4.5.0

CVE 2026-10679

Divide-by-zero in DesignWare SPI driver reachable from spi_transceive syscall (local DoS)

The DesignWare SPI driver (drivers/spi/spi_dw.c) computed the SPI BAUDR clock divider as info->clock_frequency / config->frequency without validating config->frequency.

spi_transceive is a Zephyr __syscall and its verify handler (drivers/spi/spi_handlers.c) copies the caller-supplied spi_config from userspace without checking the frequency field, so a userspace thread that has been granted access to a DesignWare SPI device kernel object can pass frequency = 0 and trigger an unsigned integer divide-by-zero in spi_dw_configure().

On Cortex-M Mainline (SCB->CCR.DIV_0_TRP is set in z_arm_fault_init()) and on ARC (a dedicated __ev_div_zero vector) this raises a CPU exception, resulting in a kernel fault and local denial of service.

The fix rejects zero frequency and frequencies above clock_frequency / 2 (the DesignWare SSI databook minimum SCKDIV of 2) with -EINVAL. The defect affects all Zephyr releases up to and including v4.4.0; exploitation requires CONFIG_USERSPACE=y and an unprivileged thread already granted SPI driver permission. There is no memory-corruption or information-disclosure impact.

This has been fixed in main for v4.5.0

CVE 2026-10680

Out-of-bounds access in Zephyr BR/EDR L2CAP configuration request handling via uint16_t length underflow

The Classic (BR/EDR) L2CAP signaling handlers l2cap_br_conf_req() and l2cap_br_conf_rsp() in subsys/bluetooth/host/classic/l2cap_br.c validated the minimum command size against buf->len (the bytes remaining in the whole received PDU) instead of len (the per-command data length from the L2CAP signaling header). Because multiple signaling commands can be packed into one PDU, buf->len may exceed a command’s len. An attacker can send a CONF_REQ command with a header length smaller than the configuration-request structure (e.g. 0), followed by another command so that buf->len still satisfies the check. The check then passes incorrectly and opt_len = len - sizeof(*req) underflows the uint16_t to a near-0xFFFF value. The configuration-option loop, which lacks an opt_len-versus-buf->len guard, then walks far past the end of the pooled ACL receive buffer using net_buf pull primitives that perform no runtime bounds check, producing an out-of-bounds read of host memory and, when the out-of-bounds option bytes encode an MTU or flush-timeout option, an out-of-bounds write. The BR/EDR signaling channel is processed before pairing/encryption and an L2CAP channel to an L0 service such as SDP can be opened without pairing, so an unauthenticated peer within radio range that can establish an ACL connection can trigger the flaw, leading to memory corruption and denial of service (host/device crash). The defect is present in released versions including v4.4.0. The fix validates against len instead of buf->len in both handlers.

This has been fixed in main for v4.5.0

CVE 2026-10681

SMP race in thread_idx_alloc() lets concurrent k_object_alloc(K_OBJ_THREAD) callers share a kernel-object permission slot

In Zephyr’s userspace dynamic-objects subsystem, thread_idx_alloc() in kernel/userspace/userspace.c allocated a new thread permission index from the global _thread_idx_map[] bitmap without holding lists_lock.

On SMP systems, two user-mode threads invoking the k_object_alloc(K_OBJ_THREAD) syscall concurrently can both observe the same low free bit, perform the same non-atomic RMW to clear it, and return the identical tidx.

The two newly created K_OBJ_THREAD objects are then assigned the same thread_id, so the two user threads alias a single bit position in every kernel object’s perms[] bitfield: any subsequent grant of access on a kernel object to one thread is implicitly a grant to the other, defeating userspace ACL isolation. A secondary lost-update window between the unlocked &=~BIT() in alloc and the locked |= BIT() in thread_idx_free() can also leak entries from the thread-index pool.

The defect is reachable from any user-mode thread via the unrestricted __syscall k_object_alloc and is gated on CONFIG_USERSPACE, CONFIG_DYNAMIC_OBJECTS, and CONFIG_SMP. The flaw was introduced when the per-thread permission index was added in 2018 and is present in every release up to and including v4.4.0. Fixed by holding lists_lock across the bitmap RMW and the permissions clear (and inlining the obj_list traversal that previously took the lock itself).

This has been fixed in main for v4.5.0

CVE 2026-10682

Out-of-bounds write in Zephyr log_filter_set syscall verifier reachable from userspace

The userspace verifier z_vrfy_log_filter_set() for the log_filter_set syscall in subsys/logging/log_mgmt.c performed a signed comparison against the int16_t src_id parameter: src_id < (int16_t)log_src_cnt_get(domain_id). Any negative value for src_id (e.g. -1) trivially satisfied this check and was forwarded into z_impl_log_filter_set, where it propagated to filter_set() and ultimately to get_dynamic_filter(), which uses source_id as an unsigned index into the linker-section array &TYPE_SECTION_START(log_dynamic)[source_id].filters.

After implicit conversion through uint32_t, an int16_t -1 becomes 0xFFFFFFFF, indexing log_dynamic far out of bounds and causing the kernel to perform an OOB read and an OOB read-modify-write (LOG_FILTER_SLOT_GET/SET) against memory adjacent to the log_dynamic section.

The written value is a constrained 3-bit log level slot within the targeted 32-bit word, but the target address is attacker-chosen (a small negative offset from log_dynamic) and the write occurs in supervisor mode following a syscall from an unprivileged user thread, providing a kernel memory-corruption / privilege-escalation primitive.

The defect is reachable on any build with CONFIG_USERSPACE=y and CONFIG_LOG_RUNTIME_FILTERING=y. Present from Zephyr v3.3.0 through v4.4.1. The fix replaces the signed bound check with an unsigned comparison: (uint32_t)src_id < log_src_cnt_get(domain_id), which correctly rejects negative inputs.

This has been fixed in main for v4.5.0

CVE 2026-10683

DesignWare I2C target driver can be wedged into a permanent stuck state by an on-bus master (DoS)

In the Synopsys DesignWare I2C driver (drivers/i2c/i2c_dw.c) operating in target/slave mode, the rx_full interrupt handler gates the write_requested() callback on dw->state != CMD_SEND, and dw->state is only reset to READY on a STOP interrupt. The START_DET interrupt, whose handler in i2c_dw_slave_read_clear_intr_bits() would reset the state on every (re)START, was never added to the enabled interrupt mask in i2c_dw_slave_register(), so that recovery path was dead code.

As a result, if the STOP interrupt is lost (bus glitch/reset, or a concurrent master driving STOP) or the bus master issues a legal WRITE-repeated-START-WRITE sequence with the same direction, the driver remains in CMD_SEND permanently and never invokes write_requested() again for the life of the target.

An I2C master on the same physical bus can deliberately trigger this, causing the I2C target function to malfunction for all subsequent write transactions and desynchronizing consumer framing state (e.g. MCTP-over-I2C), a recoverable-by-reset denial of service of the target peripheral.

The fix unmasks START_DET so the state is reset at every bus (re)START. Impact is availability-only over a local board-level bus; no memory corruption results in the in-tree consumer, whose per-byte buffer write is independently bounds-checked.

This has been fixed in main for v4.5.0

CVE 2026-10684

Out-of-bounds read in coredump shell when printing stored-dump target code

In subsys/debug/coredump/coredump_shell.c, print_coredump_hdr() used the 16-bit tgt_code field of a stored Zephyr coredump header directly as an index into coredump_target_code2str[], a fixed 7-element array of string pointers, with no bounds check.

A stored coredump whose tgt_code is >= 7 causes an out-of-bounds read of a char* up to ~64K entries past the array; that value is passed as the %s argument to shell_print, which dereferences and walks it as a string. The result is either disclosure of device memory contents to the shell user or a crash when the out-of-bounds pointer is unmapped.

The defect is reached via the coredump print shell command (cmd_coredump_print_stored_dump -> pretty_print_coredump -> parse_and_print_coredump -> print_coredump_hdr). The tgt_code field is device-generated and in-range during normal crash handling, so triggering requires local shell access plus the ability to stage or corrupt the stored coredump in the flash/in-memory backend.

Introduced in v4.2.0 (commit 13abd7fe730) and present through v4.4.0; fixed by clamping out-of-range codes to the ‘unknown’ (index 0) entry.

This has been fixed in main for v4.5.0

CVE 2026-10685

Use-after-free of GATT subscribe params in Bluetooth host CCC-write response handler

The Zephyr Bluetooth GATT client CCC-write response handler gatt_write_ccc_rsp() in subsys/bluetooth/host/gatt.c invoked the application’s params->subscribe() callback after it had already called params->notify(conn, params, NULL, 0).

Per the public GATT API, a notify callback with NULL data is the documented signal that the subscription has terminated and the bt_gatt_subscribe_params struct may be freed or reused by the application; calling subscribe() on the struct afterwards is a use-after-free, including an indirect call through the freed params->subscribe function pointer.

The error branch is remotely (adjacent) reachable: a Zephyr device acting as a GATT client that calls bt_gatt_subscribe() can be driven into this ordering when a connected GATT server peer answers the CCC write with an ATT Error Response (the peer-supplied error code flows through att_error_rsp -> att_handle_rsp into gatt_write_ccc_rsp).

For applications that free or recycle subscription parameters in their notification-termination handler, this results in memory corruption, a crash (denial of service), or potentially attacker-influenced control flow. The fix reorders the handler so the subscribe() callback runs before the terminating notify(NULL) in both the error and unsubscribe paths.

This has been fixed in main for v4.5.0

CVE 2026-10686

Missing hop-limit decrement on IPv6 forwarding path allows unbounded packet looping (DoS) in Zephyr routers

Zephyr’s IPv6 forwarding path re-sent routed unicast packets without ever decrementing the IPv6 hop limit. Both routing branches of ipv6_route_packet() (subsys/net/ip) were affected: the explicit-route path (net_route_packet()) and the on-link cross-interface path (net_route_packet_if()). Each set the packet forwarding flag and called net_send_data() with the hop limit untouched and no expiry check.

Per RFC 8200 the hop-limit decrement is the mechanism that bounds packet lifetime and terminates routing loops; without it, a device acting as an IPv6 router relays looping packets indefinitely. An on-path attacker who can induce or exploit a transient L3 loop turns it into a permanent forwarding storm, causing CPU/bandwidth resource exhaustion (availability DoS) on the forwarder and adjacent links; path-discovery and loop diagnostics that rely on hop-limit expiry are also defeated.

Affected configurations. In every affected release the forwarding path is reached via CONFIG_NET_ROUTE (enabled by default when CONFIG_NET_IPV6_NBR_CACHE is set), together with CONFIG_NET_ROUTING for cross-interface routing. Note that CONFIG_NET_IPV6_FORWARDING and CONFIG_NET_IPV4_FORWARDING — which appear in the fix and in this advisory’s evidence notes — were introduced after v4.4.0, when the routing options were split and renamed; they do not exist in any affected release. When auditing a v4.4.1-or-earlier configuration, look for CONFIG_NET_ROUTE and CONFIG_NET_ROUTING.

IPv4 is not affected in any release. The IPv4 forwarding path (net_route_ipv4_packet() in route_ipv4.c) was added after v4.4.0 and has never shipped in a release. Its TTL decrement and IPv4 header-checksum recomputation landed on main as part of the same fix, so the evidence notes below discuss it, but no released version is reachable by way of IPv4.

Affected releases are v1.8.0 through v4.4.1: v1.8.0 introduced net_route_packet() and v2.2.0 added net_route_packet_if(), and neither decremented the hop limit. v4.3.1 carries the explicit-route fix but not the on-link one, so it is affected as well. Fixed on main by 7d8f1afa7345 (explicit-route path) and 589eadc74efa (on-link path).

This has been fixed in main for v4.5.0

CVE 2026-10687

Under embargo until 2026-08-01

CVE 2026-10772

Under embargo until 2026-08-01

CVE 2026-10773

Out-of-bounds read in DHCPv4 client message-type name lookup (net_dhcpv4_msg_type_name)

The DHCPv4 client helper net_dhcpv4_msg_type_name() in subsys/net/lib/dhcpv4/dhcpv4.c indexes a static 8-element const char * name table after a faulty bounds check. The guard used msg_type <= sizeof(name) instead of msg_type <= ARRAY_SIZE(name); sizeof returns the byte size of the pointer array (32 on 32-bit, 64 on 64-bit targets) rather than the element count of 8, so message-type values from 9 up to that byte size pass the check and cause name[msg_type - 1] to read past the end of the array.

The msg_type value originates from the DHCP MESSAGE TYPE option, which is read as an unchecked raw byte from a received packet (net_pkt_read_u8) and passed unmodified into the lookup. A DHCP server, or any host able to inject a spoofed DHCP reply onto the client’s link, can therefore drive the index out of bounds. The out-of-range slot yields a garbage const char * that is then dereferenced by a %s log conversion.

The lookup is reached only from a debug log statement (NET_DBG / LOG_DBG), so the out-of-bounds read is triggerable only when the DHCPv4 log module is built at DEBUG level (CONFIG_NET_DHCPV4_LOG_LEVEL_DBG), which is not the default configuration. When that condition holds, the result is an out-of-bounds read and a wild-pointer dereference: most likely a crash of the DHCP client (denial of service) and potentially disclosure of an adjacent pointer’s contents through the log output. The fix replaces sizeof with ARRAY_SIZE, restoring the correct 1..8 acceptance window.

This has been fixed in main for v4.5.0

CVE 2026-10774

PSA key-slot leak in Bluetooth Mesh subnet deletion leading to resource-exhaustion DoS

Zephyr’s Bluetooth Mesh subnet key management leaks one PSA Crypto key slot on every subnet-key teardown. In subsys/bluetooth/mesh/subnet.c, net_keys_create() imports the Private Beacon Key into a PSA key slot under CONFIG_BT_MESH_PRIV_BEACONS (enabled by default), but subnet_keys_destroy() guarded the matching psa_destroy_key() with CONFIG_BT_MESH_V1d1. That Kconfig symbol was removed when explicit Mesh 1.0.1 support was dropped, so the destroy branch became permanently dead code and the import is never balanced by a destroy.

The imbalanced teardown is reached every time subnet keys are destroyed: deleting a subnet (Config Server NetKey Delete), completing a Key Refresh Procedure (which retires the old key set), and resetting/re-provisioning the node. The over-the-air triggers are processed only under the node’s device key, so they are exercisable by the provisioner or network administrator that owns the node, reachable over the Bluetooth Mesh network.

With the default CONFIG_MBEDTLS_PSA_KEY_SLOT_COUNT of 16, repeated add/delete or key-refresh cycles exhaust the shared PSA key-slot pool after roughly a dozen rounds. Once exhausted, bt_mesh_private_beacon_key() and thus subnet creation fail: the node can no longer add subnets or complete key refresh, and other PSA crypto consumers on the device may be starved, until the device is rebooted. The fix aligns the destroy guard with the import guard (CONFIG_BT_MESH_PRIV_BEACONS) so each slot is freed.

This has been fixed in main for v4.5.0

CVE 2026-10848

Out-of-bounds read in Zephyr OCPP 1.6 RPC message parser (parse_rpc_msg)

The OCPP 1.6 client in subsys/net/lib/ocpp parsed inbound WAMP RPC frames in parse_rpc_msg() (subsys/net/lib/ocpp/ocpp_j.c) using a hand-rolled helper, extract_string_field(), that copied the message’s uid and action fields with strncpy(out_buf, token + 1, outlen - 1) and then scanned the result with strchr(out_buf, '"'). Because strncpy does not NUL-terminate the destination when the source is at least outlen - 1 (127) bytes long, the subsequent strchr reads past the 128-byte destination buffer into adjacent stack memory; if a " byte is found beyond the buffer, a one-byte out-of-bounds NUL write also occurs. A related defect in extract_payload() runs strchr/strrchr over the receive buffer, which may not be NUL-terminated when a maximal-length frame fills it.

The parsed bytes come directly from the OCPP central-system server over a websocket: the reader thread fills recv_buf via websocket_recv_msg() and calls parse_rpc_msg() on each inbound DATA frame (subsys/net/lib/ocpp/ocpp.c). A malicious or compromised central server, or an on-path attacker (OCPP is commonly deployed over plain ws://), can send an RPC frame whose uid or action field is 127+ bytes with no closing quote, triggering the out-of-bounds access.

The primary impact is a remotely triggerable denial of service: the unbounded scan can fault on an unmapped page, and the stray NUL write can corrupt adjacent stack state. The over-read data is not reflected to the peer, so disclosure is limited. The feature is EXPERIMENTAL and must be explicitly enabled (CONFIG_OCPP). The fix replaces the manual parser with the bounds-respecting json_mixed_arr_parse() and copies the extracted uid with an explicitly NUL-terminated buffer, eliminating both over-reads.

This has been fixed in main for v4.5.0

CVE 2026-10849

Heap out-of-bounds write in Zephyr hawkBit OTA client when terminating server response body

The hawkBit device management client in subsys/mgmt/hawkbit accumulates the body of an HTTP response from the update server into a heap buffer in response_json_cb() (subsys/mgmt/hawkbit/hawkbit.c). The buffer is sized to hold the received body bytes but reserves no space for a terminating NUL. When the full response has arrived, the code writes response_data[downloaded_size] = '\0' — and whenever the accumulated body length equals the allocation, that terminator lands one byte past the end of the heap object (a heap-based out-of-bounds write, CWE-122 / CWE-787).

The body length and fragmentation are taken directly from the parsed HTTP response (rsp->body_frag_start / rsp->body_frag_len) and are fully controlled by the remote hawkBit server, which chooses its own response length. The precise trigger depends on how the buffer grows, and both forms are remotely reachable. Since v4.0.0 the reallocation is sized to exactly downloaded_size + body_len, so any response body larger than the 1100-byte initial buffer makes the out-of-bounds write deterministic; such response sizes are normal for hawkBit deployment metadata. Before v4.0.0 the buffer grew by doubling and the growth check ((downloaded_size + body_len) > response_buffer_size) is false at equality, so a response body whose length is exactly the current allocation — 1100 bytes with the default initial buffer — skips the reallocation entirely and writes the terminator at response_data[1100] of an 1100-byte object. The HTTP length-mismatch check does not catch this, because the declared and received lengths genuinely agree. Either form is reachable by a malicious, compromised, or man-in-the-middle update server (TLS is optional and, when enabled, does not protect against a hostile server), with no authentication of response content and no client-side length cap protecting the write.

The out-of-bounds write is a fixed single NUL byte immediately following the allocation, corrupting adjacent allocator metadata or the next allocation. The practical impact is heap corruption leading to denial of service (fault on a subsequent allocation or free), with the bounded, allocator-dependent possibility of further corruption. The fix sizes the buffer to the body length plus one and copies with memcpy, ensuring the terminator always lands within the allocation.

This has been fixed in main for v4.5.0

CVE 2026-11368

Use-after-free in Bluetooth host ATT TX completion on disconnect mid-transfer

The Bluetooth host ATT layer (subsys/bluetooth/host/att.c) associates each in-flight ATT TX buffer with its owning channel via the static tx_meta_data_storage[] array (data->att_chan = chan). When a buffer’s last reference is dropped, its net-buf destroy callback defers the completion handling to the system workqueue (att_tx_destroy -> att_tx_destroy_work_handler -> att_on_sent_cb -> bt_att_sent), where bt_att_sent dereferences the channel and its ATT context (sys_slist_get(&att->reqs)).

When a peer disconnects while an ATT PDU (a server notification/indication or any response) is still in flight in the controller TX path, L2CAP tears the channel down in l2cap_chan_del(): it runs the disconnected callback and then the released callback (bt_att_released), which frees the channel slab slot. Because the in-flight buffer is held by the connection TX path rather than the channel’s own queue, its deferred destroy work can run after the channel has been freed. The att_on_sent_cb guard intended to drop the stale callback itself dereferences meta->att_chan, which is now a dangling pointer into a freed (and possibly reused) slab slot.

A remote peer with an ATT connection can drive this by disconnecting during routine ATT traffic; no pairing or user interaction is required to reach the ATT bearer. The result is a use-after-free read/write of freed channel memory, reliably crashing the Bluetooth host (denial of service) and, because the channel slab slot may be reused, potentially corrupting live memory.

The fix makes bt_att_released() NULL the att_chan field of every tx_meta_data_storage[] entry still referencing the channel before freeing it, so the deferred guard observes a NULL pointer and drops the callback. Teardown and the destroy work both run on the cooperative system workqueue, so the array update is serialized and needs no lock.

This has been fixed in main for v4.5.0

CVE 2026-11742

Use-after-free race in kernel k_queue_peek_head/tail due to missing spinlock

The kernel queue helper z_queue_node_peek() in kernel/queue.c dereferences a node taken from a queue’s data_q list, reading the node’s flag byte and, for items enqueued via k_queue_alloc_append/alloc_prepend, the data pointer of an internally allocated alloc_node struct. The implementations of z_impl_k_queue_peek_head() and z_impl_k_queue_peek_tail() performed this read-and-dereference without holding the queue’s spinlock, while every other accessor of the same list — including k_queue_get(), which unlinks a node and k_free()s its backing alloc_node — operates under that lock.

Because peek was unsynchronized, a concurrent k_queue_get() on the same queue (on an SMP build, or under preemption/ISR concurrency) can free the node between the moment peek obtains the node pointer and the moment it dereferences it. The peek then reads flag bits and a data pointer out of freed, potentially re-allocated heap memory and returns a stale or dangling pointer to its caller. k_fifo and k_lifo are thin wrappers over k_queue, so this affects buffer queues used throughout the net_buf, Bluetooth, USB, and networking subsystems; the peek operations are also system calls reachable from CONFIG_USERSPACE threads.

The consequences are a use-after-free read that can leak stale heap contents (one pointer word) and, when the returned dangling pointer is subsequently consumed as a live buffer, a dereference that can crash the system or corrupt memory. Exploitation requires winning a small race window with local access (e.g. a userspace process racing k_queue_peek_* against k_queue_get on a shared queue, or two CPUs), so practical impact is bounded and of low severity.

The fix wraps both peek implementations with k_spin_lock/k_spin_unlock on the queue lock, making the read-and-dereference atomic with respect to the concurrent unlink-and-free and bringing peek into line with the rest of the queue’s locking discipline.

This has been fixed in main for v4.5.0

CVE 2026-11743

Missing negative-offset/overflow check in SF32LB MPI QSPI NOR flash driver allows out-of-bounds read and write

The SF32LB MPI QSPI NOR flash driver (drivers/flash/flash_sf32lb_mpi_qspi_nor.c) validated the flash offset and length on its read and write paths with the test (offset + size) > data->size. Because offset is a signed off_t while size is unsigned, a negative offset is converted to a large unsigned value and the addition can wrap to a small result that passes the check. The read path then performs memcpy(dst, (void *)(data->base + offset), size) and the write path programs flash at offset and cache-invalidates data->base + offset, in both cases accessing memory outside the mapped flash window. The driver’s erase path already rejected negative offsets, but read and write did not.

In builds with CONFIG_USERSPACE, flash_read and flash_write are syscalls whose verifiers validate the device object and the caller’s buffer but deliberately delegate offset bounds checking to the driver. An unprivileged thread that has been granted access to this flash device can therefore call the syscall with a crafted negative offset and a buffer valid in its own memory domain, and reach the unchecked access.

The most direct impact is on the read path: by choosing a negative offset and matching size, an attacker slides the memcpy source below the flash base and copies arbitrary CPU-addressable memory into its own buffer, disclosing memory it is not authorized to read. The write path additionally allows programming flash at an out-of-range address and invalidating an attacker-chosen cache range, affecting integrity and availability. Reachability requires userspace to be enabled and the raw flash device object to be granted to an untrusted thread.

The fix replaces the check with qspi_nor_range_is_valid(), which rejects negative offsets and performs the bound comparison in overflow-safe 64-bit arithmetic on both paths, and additionally adds an SRAM DMA bounce buffer plus source/destination overlap rejection to prevent a separate DMA bus-hang condition.

This has been fixed in main for v4.5.0

CVE 2026-11809

UpdateHub probe: uninitialized-heap out-of-bounds read of network-supplied metadata

The UpdateHub OTA client in subsys/mgmt/updatehub/updatehub.c contains an out-of-bounds / uninitialized-memory read in z_impl_updatehub_probe(). The probe response from the UpdateHub server is copied into a heap buffer (metadata) that is correctly NUL-terminated, but a second buffer (metadata_copy) is allocated with k_malloc (unzeroed) and filled with memcpy(metadata_copy, metadata, strlen(metadata)), which omits the terminating NUL. Everything after the copied content remains uninitialized heap.

When the first json_obj_parse() over the array descriptor fails, the code falls back to json_obj_parse(metadata_copy, strlen(metadata_copy), ...). The strlen() call scans past the copied bytes through uninitialized heap and, if no zero byte is found before the end of the allocation, reads beyond the buffer; the resulting over-long length is then parsed as JSON. The probe payload is fully controlled by the (malicious, compromised, or — without the optional CONFIG_UPDATEHUB_DTLS — on-path) UpdateHub server, which can craft a large payload that fails the first parse to drive this path.

The consequence is a read of uninitialized heap, with a worst case of an out-of-bounds read past the metadata_copy allocation that can fault and crash the update thread/device, producing a network-triggerable denial of service. The over-read data is consumed only internally to evaluate the update and is not returned to the attacker, so there is no direct information disclosure and no out-of-bounds write.

The fix zeroes metadata_copy with memset before the copy, guaranteeing NUL termination and bounding strlen() within the allocation.

This has been fixed in main for v4.5.0

CVE 2026-11810

NULL-pointer dereference in UpdateHub OTA agent on empty inner metadata array (remote DoS)

The UpdateHub firmware-update agent’s probe handler (z_impl_updatehub_probe() in subsys/mgmt/updatehub/updatehub.c) parses the JSON metadata returned by the update server into a fixed two-level nested-array struct. After parsing it validates only the outer array length (objects_len != 2) and then dereferences objects[1].objects[0].objects.sha256sum via strlen() without checking that the inner object array of element [1] is non-empty.

The metadata is attacker-influenceable network input: the agent fetches it over CoAP from the configured UpdateHub server during its routine OTA probe. A malicious or compromised update server (or, when DTLS is disabled, a network man-in-the-middle) can return a response whose second outer object array is empty. Because the parse target is zero-initialised, the corresponding objects[1].objects[0].objects.sha256sum pointer is NULL, and the subsequent strlen() dereferences address zero. The same defect exists in both the ‘any boards’ and ‘some boards’ metadata layouts.

The resulting CPU fault is fatal under Zephyr’s default error handling, halting or resetting the device, so the flaw is a remotely triggerable denial of service. Impact is limited to availability; it is a read from NULL with no out-of-bounds write, memory corruption, or information disclosure. The fix rejects metadata whose inner object array is empty before any dereference, on both layouts.

This has been fixed in main for v4.5.0

CVE 2026-11811

Socket file-descriptor leak in UpdateHub OTA client start_coap_client() leading to resource-exhaustion DoS

The UpdateHub over-the-air update client’s start_coap_client() in subsys/mgmt/updatehub/updatehub.c leaks the CoAP/DTLS socket descriptor on its connection-setup failure paths. The shared error: cleanup gated socket closing on a ret > 0 flag, but ret was set to -1 immediately after the socket was created, so when zsock_setsockopt() (DTLS) or zsock_connect() subsequently failed the gate was false and cleanup_connection() was never called. The open descriptor in the global ctx.sock was then overwritten by the next attempt, permanently leaking it from the socket / net_context pool until reboot.

The failing setup path is reached every time the OTA client tries to contact the UpdateHub server and the connection cannot be established — driven automatically by the periodic autohandler() poll (and on demand via the updatehub_probe()/updatehub_update() API or the updatehub run shell command). The DTLS handshake/connect outcome is influenceable by a network or on-path attacker who drops, resets, or otherwise disrupts traffic to the server, and also fails naturally whenever the server is unreachable.

Each failed attempt permanently leaks one descriptor; once the shared socket pool is exhausted, networking degrades device-wide until the device is rebooted, a denial-of-service condition. Severity is low because the leak rate is bounded by the configured OTA poll interval (default once per 24 hours), the effect is gradual and recovered by reboot, and only builds with the UpdateHub client enabled are affected. There is no memory-corruption, information-disclosure, or authentication impact.

This has been fixed in main for v4.5.0

CVE 2026-11812

UpdateHub: race condition on shared context causes out-of-bounds write and DoS

The UpdateHub management subsystem (subsys/mgmt/updatehub/updatehub.c) drives every update operation through a single file-scope ctx structure that holds the CoAP block context, payload buffer, status code, socket, and a one-element poll-fd array fds[1]. Access to ctx was not serialized, and prepare_fds() wrote ctx.fds[ctx.nfds] and incremented ctx.nfds with no bounds check.

Two independent paths mutate ctx concurrently: the background autohandler running on the system workqueue, and user-triggered operations reached through the updatehub run shell command, direct API calls, or — since the operations are exposed as syscalls — userspace threads. When a second flow enters prepare_fds() while ctx.nfds is already 1, the write lands one element past the array; by struct layout it overlaps the adjacent ctx.sock/ctx.nfds members. More broadly, the unsynchronized sharing lets two flows interleave connection setup and teardown, double-closing a socket descriptor or scribbling the shared buffers.

The result is corruption of the update subsystem’s internal state and denial of service of the firmware-update path; the out-of-bounds write is contained within the ctx structure and there is no demonstrated path to memory outside it or to code execution. Triggering requires a local actor able to invoke update operations (or, with CONFIG_USERSPACE, an unprivileged userspace thread) and to win a timing race against the background handler; remote peers cannot control the race timing. The fix serializes the entry points with a mutex and adds a bounds check to prepare_fds().

This has been fixed in main for v4.5.0

CVE 2026-11893

Double free / use-after-free in Bouffalo Lab HCI driver send() error paths (hci_bflb)

The Bluetooth HCI driver for Bouffalo Lab on-chip BLE controllers (BL60x/BL70x/BL61x), bt_bflb_send() in drivers/bluetooth/hci/hci_bflb.c, violates the bt_hci_driver_api.send() buffer-ownership contract. That contract (documented at include/zephyr/drivers/bluetooth.h) requires the buffer reference to be consumed only on success; on error the caller still owns the reference and unrefs it. The driver instead routed all error paths through a shared label that unconditionally called net_buf_unref(buf) before returning the error code, consuming the buffer on failure as well.

When send() returns an error, the host TX path (send_buf() in subsys/bluetooth/host/conn.c) unrefs the same buffer again, believing it still owns it. This double-unref over-decrements the net_buf reference count. Because the buffer is a TX fragment whose destroy callback also decrements its still-queued parent buffer, the parent is freed prematurely while reachable on the connection TX queue, producing a use-after-free and corruption of the shared net_buf pool rather than a benign leak.

The error conditions are on the host-to-controller transmit path (controller send failure, or an unsupported H:4 packet type), so they are not driven directly by attacker-supplied radio bytes; a remote/adjacent peer can influence them only indirectly, e.g. by inducing controller TX failures under heavy link load. The consequence when reached is BLE-stack denial of service (crash / pool corruption) with possible further memory corruption, bounded to devices using one of these Bouffalo Lab on-chip controllers.

This has been fixed in main for v4.5.0

CVE 2026-11894

Double-free / use-after-free in Realtek BEE Bluetooth HCI driver send() error paths

The Realtek BEE Bluetooth HCI driver’s send callback, bt_hci_bee_send() in drivers/bluetooth/hci/hci_bee.c, violated the bt_hci_driver_api buffer-ownership contract. That contract requires the driver to consume (unref) the transmit net_buf only on success; on an error return the host caller retains ownership and unrefs the buffer itself. The pre-fix code routed all error paths through a shared cleanup label that unconditionally called net_buf_unref(buf) before returning the error code.

Because the host TX paths (in subsys/bluetooth/host/hci_core.c) unref the buffer again after send() returns an error, the buffer is freed twice: the driver returns it to its net_buf pool and the host then unrefs the already-freed buffer, corrupting the shared pool / underflowing the reference count (CWE-415). The same error branch additionally dereferenced buf->len inside a LOG_ERR call after the buffer had already been unref’d, a read of freed memory (CWE-416) that is compiled in at the default error log level.

The failing edges are reached when the controller’s host-to-controller buffer allocation fails or the controller send fails (resource-exhaustion / IO conditions). A remote Bluetooth peer can push the device toward these conditions indirectly by driving heavy host transmit activity, at which point the double-free corrupts the host net_buf pool and most likely crashes the device, with residual potential for further memory corruption. The impact is confined to builds using this specific Realtek BEE HCI driver.

The fix returns early from each error path without unreffing and unrefs the buffer only on the success path, restoring the ownership contract and eliminating both the double-free and the use-after-free read.

This has been fixed in main for v4.5.0

CVE 2026-11985

Cross-thread FPU register leak on ARM when FPU enabled without register sharing

On the Zephyr ARM port, enabling the hardware FPU (CONFIG_FPU) forces the “Floating point ABI” choice, which defaults to CONFIG_FP_HARDABI. Both FP_HARDABI and FP_SOFTABI permit the compiler to emit hardware FP instructions in any function, even code that never uses floating-point types. However, the callee-saved FP registers (s16-s31 / d8-d15) are only saved and restored across a context switch when CONFIG_FPU_SHARING is enabled (arch/arm/core/cortex_m/swap_helper.S and arch/arm/core/cortex_a_r/swap_helper.S), and prior to this fix selecting an ABI did not enable FPU register sharing, which defaults off.

In a build that enables the FPU with the default ABI but leaves CONFIG_FPU_SHARING disabled, the kernel preserves no callee-saved FP register state across thread switches. The documented precondition for this “unshared” mode — that only a single thread ever executes FP instructions — is silently violated because the compiler may generate FP instructions in every thread.

Under CONFIG_USERSPACE, where threads are mutually isolated, this becomes an information-disclosure boundary crossing: a victim thread can leave secret-derived values in s16-s31, and a co-resident unprivileged thread can read those registers directly (FP register access is not privilege-gated), recovering data left behind by another thread. Without userspace the same defect causes cross-thread FP state corruption (a correctness fault). The leak is bounded to the 16 callee-saved single-precision registers and is opportunistic, so impact is low.

The fix makes FP_HARDABI and FP_SOFTABI select CONFIG_FPU_SHARING and tags every thread with K_FP_REGS at creation, so callee-saved FP state is always preserved across context switches whenever the compiler may emit FP instructions.

This has been fixed in main for v4.5.0

CVE 2026-12051

NULL pointer dereference in USB DFU device_next download handler (handle_download)

The USB DFU class implementation in Zephyr’s new (experimental) device_next USB device stack contains a NULL pointer dereference in handle_download() (subsys/usb/device_next/class/usbd_dfu.c). The handler computes MIN(setup->wLength, buf->len) and passes buf->data to the image write callback without checking that the buf net_buf pointer is non-NULL.

The handler is reached over the USB control endpoint, driven by the USB host. For a DFU_DNLOAD (download) request with no Data OUT stage — notably the zero-length terminating download that the DFU protocol uses to end a firmware transfer — the USB core invokes the class handler with a NULL buffer. After the device has been advanced to the DFU_DNLOAD_IDLE state (by sending one valid download block and a GET_STATUS), a zero-length DFU_DNLOAD reaches handle_download() with buf == NULL, dereferencing it.

The result is a NULL+offset read that triggers a fatal CPU fault, i.e. a denial of service (device crash/reset). The attacker is whatever controls the USB host the device is attached to; DFU download support must be enabled with a registered image. There is no memory corruption or information disclosure — impact is limited to availability. The fix adds an explicit if (buf != NULL) guard so the callback receives a zero-length, NULL-data transfer instead of crashing.

This has been fixed in main for v4.5.0

CVE 2026-12052

Out-of-bounds write in USB CDC NCM control handler when host wLength is smaller than the response

The USB device-side CDC NCM class control-to-host handler usbd_cdc_ncm_cth in subsys/usb/device_next/class/usbd_cdc_ncm.c builds a fixed-size response for the GET_NTB_PARAMETERS (28-byte struct ntb_parameters) and GET_NTB_INPUT_SIZE (8-byte struct ntb_input_size) class requests and copies the whole structure into the control DATA IN buffer with net_buf_add_mem(buf, ..., sizeof(...)), ignoring the host-supplied wLength.

The control DATA IN buffer is allocated by the USB stack with a capacity of exactly wLength bytes (usbd_ep_ctrl_data_in_alloc -> udc_ctrl_data_alloc -> net_buf_alloc_len(&udc_ep_pool, wLength); no round-up is applied for the IN endpoint). Because net_buf_add_mem/net_buf_simple_add only bounds the copy with an __ASSERT_NO_MSG, which is compiled out in production builds, a host that issues one of these standard CDC NCM control requests with a wLength smaller than the response structure (e.g. wLength = 1) causes the handler to memcpy up to 27 bytes past the end of the allocated pool buffer.

The request fields come straight from the USB SETUP packet, so any host (or USB interposer) the Zephyr device enumerates against can trigger the overflow with no authentication once an image built with the device_next USB stack and the CDC NCM class is connected. The out-of-bounds write corrupts adjacent allocations and metadata in the shared udc_ep_pool, primarily causing memory corruption and denial of service of the USB stack; the overflow length is bounded (<= 27 bytes) and the written content is fixed device constants, and the bug reads nothing back so there is no information disclosure. The fix clamps the copy with MIN(sizeof(...), setup->wLength), matching the existing CDC ACM handler.

This has been fixed in main for v4.5.0

CVE 2026-12232

Out-of-bounds read via unvalidated stream_id in Intel ALH DAI get_properties

The Intel ALH digital-audio-interface driver function dai_alh_get_properties() in drivers/dai/intel/alh/alh.c used a caller-supplied int stream_id with no range validation. The value indexes the fixed-size static const uint8_t alh_handshake_map[64] array and scales a FIFO register address, so an out-of-range stream_id produces an out-of-bounds read of one byte at an attacker-chosen signed offset from the array. That byte is written into prop->dma_hs_id and the resulting struct dai_properties is copied back to the caller, leaking it.

dai_get_properties_copy() is a Zephyr __syscall, and its verifier z_vrfy_dai_get_properties_copy() (drivers/dai/dai_handlers.c) validates only the device-object permission and the destination buffer, not stream_id. A user-mode thread that has been granted access to the ALH DAI device object can therefore call the syscall with an arbitrary stream_id, crossing the userspace/kernel sandbox boundary.

The impact is a one-byte-per-call arbitrary-offset kernel information disclosure (and leakage of a computed kernel address via fifo_address); a stream_id that resolves to an unmapped page faults in kernel context, giving a local denial of service. Exploitation requires CONFIG_USERSPACE and device access, making this a local, moderate-severity issue. The fix rejects negative and too-large stream_id values up front and returns NULL, which the copy wrapper maps to -ENOENT.

This has been fixed in main for v4.5.0

CVE 2026-12233

Uninitialized mutex in TLS trusted-credential backend causes kernel NULL-deref DoS under contention

The PSA Protected Storage credential backend (subsys/net/lib/tls_credentials/tls_credentials_trusted.c) declared its credential-store mutex as a plain zero-filled static struct k_mutex credential_lock; and never called k_mutex_init() on it. A statically zero-filled k_mutex has an uninitialized wait queue (its dlist head/tail are NULL instead of the self-referential sentinels that k_mutex_init/K_MUTEX_DEFINE install). The uncontended lock path does not touch the wait queue, so the defect is latent and serialized use behaves correctly.

When two execution contexts contend on the lock, k_mutex_lock() pends the blocking thread on the wait queue via z_pend_curr(), which calls sys_dlist_append() on the zeroed list and dereferences a NULL tail pointer (tail->next = node), faulting the kernel. The lock is held during TLS handshake credential loading and by all credential add/get/delete operations, so a deployment performing concurrent TLS handshakes (for example a server handling multiple simultaneous connections from a remote peer) or a credential-management operation concurrent with a handshake can trigger the dereference.

The impact is a denial of service: a deterministic kernel panic / device reset on the first contention. There is no memory corruption beyond the NULL dereference and no confidentiality or integrity impact; mutual exclusion on the fast path remains correct. Exposure is limited to builds with CONFIG_TLS_CREDENTIALS_BACKEND_PROTECTED_STORAGE enabled (PSA Protected Storage / TF-M platforms); the default volatile RAM backend initializes its lock correctly and is unaffected.

The fix initializes the mutex statically with K_MUTEX_DEFINE(credential_lock), providing a valid wait queue so the contended path no longer touches a NULL list.

This has been fixed in main for v4.5.0

CVE 2026-12234

TOCTOU double-fetch in zsock_sendmsg/recvmsg userspace verifiers allows kernel-heap out-of-bounds write

The userspace syscall verifiers z_vrfy_zsock_sendmsg() and z_vrfy_zsock_recvmsg() in subsys/net/lib/sockets/sockets.c snapshot the caller-supplied struct net_msghdr into a kernel-side copy with k_usermode_from_copy(), but then re-read the still-live user struct for subsequent decisions. The kernel iovec shadow buffer is sized from one read of msg->msg_iovlen, while the population loop is bounded by a second, live read of the same field.

Because msg points into ordinary user memory, a cooperating second thread in the same memory domain can inflate msg->msg_iovlen in the window between the sizing read and the loop test (a classic double-fetch / TOCTOU). The population loop then iterates past the number of net_iovec slots actually allocated, writing attacker-influenced iov_base/iov_len values beyond the end of the kernel-heap shadow buffer. The recvmsg verifier has the same defect on both its inbound and result write-back loops.

The code is reachable from an unprivileged user thread whenever CONFIG_USERSPACE is enabled and the zsock_sendmsg/zsock_recvmsg syscalls are available. A successful race corrupts kernel-managed heap memory across the user-to-kernel privilege boundary, yielding a local privilege-escalation primitive or, at minimum, a kernel-fault denial of service. The fix copies the header once and derives every size, bound, and gate from the snapshot, copying each iovec entry atomically so its base and length can no longer be raced apart.

This has been fixed in main for v4.5.0

CVE 2026-12235

Out-of-bounds write in Xtensa llext PLT relocation from malformed ELF (CWE-787)

The Linkable Loadable Extensions (llext) subsystem mis-handles PLT/RELA relocation entries when linking a relocatable (partially-linked) ELF extension. In llext_link_plt() (subsys/llext/llext_link.c), the relocatable branch (tgt != NULL, the path used for Xtensa relocatable objects) computed the patch address as ext->mem[LLEXT_MEM_TEXT] - text.sh_offset + rela.r_offset + tgt->sh_offset and then performed the relocation write there without validating rela.r_offset. Its sibling shared/dynamic branch already rejected out-of-range offsets via llext_file_offset().

rela.r_offset is read directly from the ELF’s RELA table, so a crafted entry with an offset larger than the target section makes the write land arbitrarily far outside the extension’s text buffer. The result is an attacker-influenced out-of-bounds write (the location via r_offset, the written value being the resolved symbol address) performed in supervisor context at link time, before any extension code runs.

The path is reached from llext_load() whenever an application loads an attacker-influenced ELF extension on Xtensa with writable storage; llext is documented to accept extensions of untrusted origin. Impact is supervisor-context memory corruption (integrity and availability loss, and a sandbox-boundary escape for user-mode extensions). Exploitation is gated by the Xtensa relocatable PLT path and writable storage, and turning the out-of-range write into a useful primitive is non-trivial.

The fix adds a bound check rejecting any RELA entry whose r_offset >= tgt->sh_size, mirroring the existing validation in the shared branch.

This has been fixed in main for v4.5.0

CVE 2026-12236

Infinite loop (DoS) in Bluetooth GATT client parsing of Read-By-Type responses with zero data length

The Bluetooth host GATT client function parse_read_std_char_desc() in subsys/bluetooth/host/gatt.c parses an ATT Read By Type Response received from a remote GATT server during BT_GATT_DISCOVER_STD_CHAR_DESC discovery. The per-entry stride rsp->len is taken directly from the peer’s PDU, and the parse loop both tests its exit condition (length >= rsp->len) and advances (length -= rsp->len, pdu += rsp->len) using that value. The minimum value of rsp->len was never validated before the loop.

A malicious or malfunctioning peer can reply with rsp->len = 0. Because length is unsigned and never decreases, the loop condition stays true forever and the read pointer never advances; as long as the body is at least a few bytes with a non-zero handle and a matching descriptor UUID, the host repeatedly re-parses the same bytes and invokes the discovery callback, never terminating. This hangs the Bluetooth host processing thread (CWE-835, loop with unreachable exit condition).

The condition is reachable by any connected peer once the local device initiates standard-descriptor-value discovery; GATT discovery does not require bonding or encryption, so an unauthenticated adjacent attacker that the device connects to can trigger it. The impact is denial of service of the Bluetooth subsystem (and likely a watchdog reset on constrained targets); there is no memory disclosure or corruption.

The fix adds a rsp->len < sizeof(struct bt_att_data) check before the loop, rejecting under-length responses so the stride is always non-zero and the loop terminates. The sibling parsers parse_include() and parse_characteristic() already validated rsp->len and are unaffected.

This has been fixed in main for v4.5.0

CVE 2026-7007

Division by zero in Zephyr ext2 superblock parsing allows DoS via crafted filesystem image

The Zephyr ext2 file system validates the on-disk superblock in ext2_verify_disk_superblock() (subsys/fs/ext2/ext2_impl.c) before completing a mount. The validator checked the magic number, block size, revision and feature flags, but did not verify that the on-disk fields s_blocks_per_group and s_inodes_per_group are non-zero. Both fields are read directly from the image and are later used as divisors during mount-time initialization.

During mount, get_ngroups() divides and modulos s_blocks_count by s_blocks_per_group (reached via ext2_fetch_block_group() from ext2_init_fs()), and get_itable_entry() divides (ino - 1) by s_inodes_per_group when fetching the root inode (both in subsys/fs/ext2/ext2_diskops.c). A superblock with either field set to zero therefore causes an integer division by zero during the mount sequence.

An attacker who can present a crafted ext2 image to a device that mounts ext2 — removable media such as an SD card or a USB mass-storage device — can trigger this. On ARMv7-M / ARMv8-M-mainline Cortex-M targets, divide-by-zero trapping is enabled (SCB_CCR_DIV_0_TRP), so the division raises a UsageFault that Zephyr treats as a fatal error, producing a denial of service. The impact is limited to availability; the malformed value is consumed only as a divisor.

The fix rejects a zero s_blocks_per_group or s_inodes_per_group in the superblock validator, returning -EINVAL so the mount fails before any block-group or inode I/O occurs.

This has been fixed in main for v4.5.0

CVE 2026-8023

Path traversal in Zephyr HTTP server static-filesystem resource handler allows unauthenticated remote arbitrary file read

Zephyr’s HTTP server (subsys/net/lib/http) provides a static-filesystem resource type (HTTP_RESOURCE_TYPE_STATIC_FS, available when CONFIG_FILE_SYSTEM is enabled) that serves files from a configured root directory. Before this fix, both the HTTP/1 and HTTP/2 front-ends placed the raw, attacker-controlled request path into client->url_buffer (assembled in on_url() for HTTP/1 and copied verbatim from the :path pseudo-header for HTTP/2) without resolving ./.. segments. The static-FS handler then built the on-disk filename by directly concatenating the configured root with that raw URL (snprintk(fname, ..., "%s%s", static_fs_detail->fs_path, client->url_buffer) at http_server_http1.c:603 and http_server_http2.c:490) and opened it with fs_open(fname, FS_O_READ). Because the handler is reached via wildcard/leading-dir (fnmatch FNM_LEADING_DIR) or fallback resource matching, a request such as GET /<prefix>/../../<file> is dispatched to the handler and, after the underlying filesystem (e.g. LittleFS/FAT) resolves the .. segments, escapes the configured web root, letting an unauthenticated remote client read arbitrary readable files on the mounted volume (information disclosure). The HTTP server requires no TLS or authentication to reach this path. The fix adds http_server_remove_dot_segments(), which canonicalizes the path portion of the URL before resource lookup in both protocol handlers, neutralizing the traversal. Affects releases v4.0.0 through v4.4.0 for deployments that register a static-filesystem resource.

This has been fixed in main for v4.5.0

CVE 2026-9728

TOCTOU race in mbox_send syscall verifier allows userspace to leak kernel memory

The userspace syscall verifier z_vrfy_mbox_send() in drivers/mbox/mbox_handlers.c validated the nested msg->data/msg->size fields by reading them directly out of live userspace memory, and then forwarded the original, still-mutable userspace struct mbox_msg * pointer to z_impl_mbox_send() and the underlying driver. Between the access check and the driver’s use of msg->data, the validated pointer could be replaced, leaving a time-of-check/time-of-use window.

On a system built with CONFIG_USERSPACE, any unprivileged userspace thread may invoke the mbox_send() system call. A second thread sharing the caller’s address space can race to overwrite msg->data with a supervisor (kernel) address after the verifier’s bounds check has passed but before the driver dereferences it. The driver then reads from the attacker-chosen address in supervisor context (for example memcpy(&data32, msg->data, msg->size) in the NXP mailbox driver, whose bytes are subsequently emitted to the peer mailbox endpoint).

The impact is a userspace-to-supervisor access-control bypass: disclosure of kernel memory contents (high confidentiality impact), or, for an invalid/unmapped target address, a faulting kernel read causing denial of service. The fix snapshots the entire struct mbox_msg into a kernel-stack copy with k_usermode_from_copy() and validates and forwards that immutable copy, closing the race.

This has been fixed in main for v4.5.0

CVE 2026-9771

Missing device-pointer validation in flash_copy() syscall allows userspace privilege escalation

The flash_copy() system call is verified by z_vrfy_flash_copy() in drivers/flash/flash_util.c. On builds with CONFIG_USERSPACE enabled, this handler is the kernel-side trust boundary for a user-mode caller. Prior to the fix it validated only the output buffer (K_SYSCALL_MEMORY_WRITE) and passed the two struct device * arguments, src_dev and dst_dev, directly into the implementation without any object validation — unlike every sibling flash syscall, which guards its device pointer with K_SYSCALL_DRIVER_FLASH.

A user-mode thread fully controls the values of src_dev/dst_dev and the contents of its own address space. The implementation z_impl_flash_copy() dereferences these pointers and calls through their driver-API function tables (e.g. api->get_parameters(dst_dev), flash_read(src_dev, ...), flash_write(dst_dev, ...)). By supplying a pointer to a forged struct device whose api table contains attacker-chosen function pointers, an unprivileged thread can cause the kernel to call arbitrary code in supervisor mode; passing any arbitrary or invalid address otherwise yields a kernel crash or out-of-bounds read.

The result is a local privilege escalation out of the userspace sandbox (with kernel denial-of-service and information disclosure as lesser outcomes). The fix adds K_SYSCALL_DRIVER_FLASH(src_dev, read) and K_SYSCALL_DRIVER_FLASH(dst_dev, write) to z_vrfy_flash_copy(), which verify each device is a registered flash-driver kernel object the calling thread is permitted to use before any dereference, closing the path completely.

This has been fixed in main for v4.5.0

CVE 2026-12363

Out-of-bounds write in LoRaWAN fragmented transport from a fragment index of 0

The LoRaWAN Fragmented Data Block Transport service (subsys/lorawan/services/frag_transport.c) does not validate the fragment counter in a received DATA_FRAGMENT command before forwarding it to the configured decoder. In frag_transport_package_callback() the value frag_counter = hdr->frag_index_n & 0x3FFF is taken directly from the downlink payload and passed to the decoder, which derives an array index and flash offset as frag_counter - 1. DataFragment fragments are 1-indexed, so a frag_counter of 0 underflows that arithmetic.

With the default Semtech/LoRaMAC-node decoder, this reaches FragDecoder.FragNbMissingIndex[fragCounter - 1] = 0; in FragDecoderProcess(), where fragCounter - 1 evaluates to -1 and writes a uint16_t zero out of bounds, just before the array and into the adjacent MatrixM2B recovery-matrix state of the static decoder object (CWE-787). A companion write derives a wild flash offset, but that path is rejected by the flash_area_write() bounds check. The in-tree low-memory decoder (frag_dec()) is not corrupted: its out-of-range bit-array and flash accesses are caught by sys_bitarray_* and flash_area_* bounds checks.

The handler is the registered downlink callback for the fragmentation transport port, reachable whenever an active fragmentation session exists, so the triggering byte is attacker-influenceable LoRaWAN/FUOTA network input. Triggering it requires authenticated downlinks (LoRaWAN MAC session keys or a malicious/compromised network or FUOTA server) and an active fragmentation session. The impact is contained: corruption of decoder state and denial of the firmware-update (FUOTA) session rather than controllable memory corruption or code execution. The fix adds a transport-layer check that rejects frag_counter == 0, closing the defect for both decoder backends.

This has been fixed in main for v4.5.0

CVE 2026-12364

Missing user-space pointer validation in logging syscall z_log_msg_static_create allows kernel memory disclosure and denial of service

The user-space system-call verifier z_vrfy_z_log_msg_static_create() in subsys/logging/log_msg.c was a pure pass-through: it forwarded the caller-supplied source, desc, package, and data arguments directly to the kernel-mode implementation z_impl_z_log_msg_static_create() without performing any of the mandatory K_SYSCALL_* checks. Because z_log_msg_static_create() is declared __syscall, under CONFIG_USERSPACE any unprivileged user-mode thread can invoke it directly with fully attacker-controlled arguments.

The kernel-mode handler dereferences each of these untrusted values: frontend_runtime_filtering() reads through the source pointer as a struct log_source_dynamic_data, cbprintf_package_copy() reads desc.package_len bytes from the package pointer, and z_log_msg_finalize() performs a memcpy() of desc.data_len bytes from the data pointer. With no verification, a user thread can supply arbitrary kernel addresses and arbitrary lengths, and the kernel will read from them.

The impact is a kernel-mode denial of service (the kernel faults dereferencing an attacker-chosen pointer) and, where a log backend output is observable to the attacker, disclosure of arbitrary kernel memory copied into the emitted log message — a confidentiality breach across the user/kernel boundary that the userspace sandbox is meant to enforce. The reads do not corrupt kernel memory, so there is no out-of-bounds write primitive.

The fix adds the required validation to the verifier: it bounds desc.package_len against Z_LOG_MSG_MAX_PACKAGE, rejects non-NULL/length mismatches, and applies K_SYSCALL_MEMORY_READ() to package, data, and (when runtime filtering with a frontend is enabled) source, so any out-of-bounds or kernel pointer now raises K_OOPS instead of being honored.

This has been fixed in main for v4.5.0

CVE 2026-12365

Use-after-free in Zephyr delayable work-queue cancellation under SMP timing race

A use-after-free exists in the Zephyr second-generation work queue (kernel/work.c) in the handling of delayable work timeouts. When a delayable work item’s timeout has been dequeued and its handler work_timeout() is in flight (blocked acquiring the work-queue spinlock), a concurrent cancellation does not wait for that handler to finish. In unschedule_locked() the pre-fix code called z_abort_timeout(), which for an already-announcing record returns -EINVAL without removing it; cancel_async_locked() then observes the work as idle, so even k_work_cancel_delayable_sync() and k_work_flush_delayable() return without blocking on the in-flight handler.

Because those are the APIs the kernel header documents as the safe way to cancel before freeing a k_work_delayable, a caller that frees the object immediately after a successful sync cancel can race the still-pending handler. work_timeout() subsequently dereferences the freed record: it reads to->dticks via z_is_timeout_handler_canceled() and, if the freed slot has been reused so the bail check fails, performs a read-modify-write of wp->flags (K_WORK_DELAYED_BIT) and submits work against a stale dw->queue pointer — a use-after-free read and write.

The k_work API is kernel-mode only (no __syscall entry point), so this is a kernel-internal concurrency defect rather than a userspace privilege escalation. Triggering it requires an SMP build and a subsystem that schedules and then frees (or reschedules) a delayable work item in the narrow window while its timeout is announcing; an attacker able to influence the timing of such teardown (for example via connection churn driving subsystem timers) has a plausible but probabilistic path. The impact is kernel memory corruption or crash (denial of service).

The fix makes unschedule_locked() wait, by spinning on z_try_abort_timeout() returning -EAGAIN while releasing and re-acquiring the work spinlock, until any in-flight handler completes before returning, and switches work_timeout() to atomic K_WORK_DELAYED_BIT ownership. This closes both the free-then-handler use-after-free and the related reschedule early-fire race.

This has been fixed in main for v4.5.0

CVE 2026-12366

Use-after-free freeing an armed dynamically-allocated k_timer in Zephyr userspace object disposal

Zephyr’s dynamic kernel-object disposal path unref_check() in kernel/userspace/userspace.c frees an object’s storage (k_free(dyn->data)) once its reference count reaches zero, after running a per-object-type cleanup. The cleanup switch handled only K_OBJ_MSGQ and K_OBJ_STACK; there was no K_OBJ_TIMER case. A dynamically-allocated, initialized, and armed k_timer keeps its embedded struct _timeout dnode linked in the global timeout queue (_timeout_q), so freeing the timer storage without cancelling the timeout leaves a dangling node in that queue.

When the timer next expires, the timeout machinery walks _timeout_q and invokes z_timer_expiration_handler() on the freed node, dereferencing and writing freed (and reusable) kernel heap in kernel/ISR context. This is a deterministic use-after-free that does not depend on SMP: the queued node is simply never unlinked at free time.

The disposal is reachable from an unprivileged user thread under CONFIG_USERSPACE + CONFIG_DYNAMIC_OBJECTS: a thread that holds the last permission on such a timer drops it via the k_object_release() syscall (or by exiting, through k_thread_perms_all_clear()), and can arm the timer itself via the k_timer_start() syscall. The free and the expiration handler run at kernel privilege while the actor is a user thread, so the bug is a sandbox-escape memory-corruption primitive usable for privilege escalation. The fix adds k_timer_cleanup() (cancel the timeout and wait for any in-flight handler) and calls it for K_OBJ_TIMER before freeing.

This has been fixed in main for v4.5.0

CVE 2026-12519

Out-of-bounds stack read and write in Zephyr WNC-M14A2A modem socket-notify parsing

The WNC-M14A2A LTE-M modem driver mishandles unsolicited %NOTIFYEV: events in on_cmd_socknotifyev() (drivers/modem/vendor_standalone/wncm14a2a.c). The response line is linearized into a fixed 40-byte stack buffer via net_buf_linearize(), which caps the copy at 39 bytes and returns out_len <= 39. The two quote-delimiter scanning loops, however, were bounded by len — the full CR/LF-delimited frame length returned by net_buf_findcrlf() — rather than by out_len.

When a %NOTIFYEV: line longer than 39 bytes contains no " within the linearized region, the loop indices p1/p2 walk past value[39] and read adjacent stack memory until a stray quote byte is found or the index reaches len. The over-read string is then passed to strncmp()/atoi()/LOG_*, and if a quote byte is found out of bounds the subsequent value[p2] = '\0' performs a single-NUL out-of-bounds stack write at an attacker-influenced offset.

The %NOTIFYEV: payload carries network-derived content (LTIME network time, SIB1 base-station system information, CSPS/RRCSTATE), so a rogue cellular base station, a malicious or compromised modem module, or RF manipulation that induces an over-long notify line reaches the defect without any application interaction; the handler runs automatically on the unsolicited event in the modem RX thread.

The impact is out-of-bounds stack disclosure (into logs and parsing) and stack corruption that can crash the modem RX thread (denial of service). The write offset is only weakly controlled, so memory-safe code execution is not demonstrated. The fix bounds both scanning loops by out_len, keeping all accesses within the linearized buffer.

This has been fixed in main for v4.5.0

CVE 2026-12520

Stack buffer overflow and off-by-one writes in Zephyr HL7800 modem AT response handlers

The Sierra Wireless HL7800 cellular modem driver (drivers/modem/vendor_standalone/hl7800.c, located at drivers/modem/hl7800.c in v4.4.0 and earlier) parses AT responses with roughly twenty handlers that call net_buf_linearize(value, sizeof(value), *buf, 0, len) into a 128-byte stack buffer and then write value[out_len] = 0. Because net_buf_linearize() (lib/net_buf/buf.c) can return a count equal to its destination-length argument, a field that exactly fills the buffer makes the terminating NUL land one byte past the end, a single-byte out-of-bounds write into adjacent stack memory.

The +KCELLMEAS cell-measurement handler on_cmd_atcmdinfo_rssi() is worse: it passed the wire length len as the destination size (net_buf_linearize(value, len, *buf, 0, len)), so a response line longer than 128 bytes overflows the value stack buffer with attacker-influenceable content. The line length comes from net_buf_findcrlf(), which accumulates bytes across the whole net_buf fragment chain and is not bounded to 128, so an over-long line reaches the defect.

The data originates from the cellular modem over UART, driven by the network: operator-scan results, +CGCONTRDP IP/DNS info, socket indications, and +KCELLMEAS neighbour-cell reports. An attacker able to shape what the modem emits — a rogue base station, a compromised modem baseband, or a remote peer feeding oversized response framing — can drive a line past 128 bytes. The handlers run in the driver’s RX thread in kernel context, so the corruption is kernel-side.

The +KCELLMEAS path is a full stack buffer overflow whose worst case is code execution in kernel context and whose floor is a reliable crash; the remaining sites are single-byte NUL out-of-bounds writes. Exploitation requires the modem to emit an over-long AT response line, giving high attack complexity over an adjacent (cellular radio) vector. The fix passes sizeof(dst) - 1 (and correct explicit bounds for the IMSI and +KCELLMEAS sites) so the terminator always stays in bounds.

This has been fixed in main for v4.5.0

CVE 2026-12521

Zephyr HTTP server kernel timeout-list corruption on spurious zsock_poll() return

The HTTP server core loop http_server_run() in subsys/net/lib/http/http_server_core.c polls the listening, stop, and client sockets with an infinite timeout and treats a zsock_poll() return of 0 as impossible, executing break and returning 0 without reaching its closing: cleanup. However, zsock_poll() (zvfs_poll_internal() in lib/os/zvfs/zvfs_poll.c) can legitimately return 0 even with an infinite timeout: after k_poll() wakes, the ZFD_IOCTL_POLL_UPDATE pass may find every fd reporting no revents (a spurious wakeup), yielding ret == 0.

On that path close_all_sockets() is skipped, so the per-client inactivity_timer (struct k_work_delayable) cancellations in close_client_connection() never run. Control returns to http_server_thread() with server_running still true, which re-enters http_server_init(). http_server_init() then memset()s the ctx->clients array — including the k_work_delayable timeout nodes — while one or more of those timers are still armed and linked in the kernel sys-timeout list.

When the kernel later services such a timeout it operates on a reinitialized object and follows the now-zeroed list links, corrupting the kernel timeout list and producing a delayed fault. The HTTP server loop is driven by unauthenticated remote network peers (CONFIG_HTTP_SERVER, all of HTTP/1/2/3), and connection churn raises the probability of the spurious-wakeup race, so a remote peer can influence the trigger. The realistic impact is a denial of service (kernel crash/hang) from the corrupted timeout list.

The fix replaces break with continue, re-polling on a spurious 0 return, which leaves the sockets and their armed timers intact and never re-initializes the context over live timers.

This has been fixed in main for v4.5.0

CVE 2026-12522

Stack buffer overflow in Zephyr hl7800 modem driver parsing network-supplied +CGCONTRDP address fields

The HL7800 cellular modem driver’s +CGCONTRDP: response handler on_cmd_atcmdinfo_ipaddr() in drivers/modem/vendor_standalone/hl7800.c parses the PDP-context dynamic parameters (local address, subnet mask, gateway, and DNS servers) that the cellular network assigns to the device. The response is linearized into a 256-byte stack buffer, after which each address field length is computed from comma/. delimiter positions in the network-supplied data and used directly as the length argument to strncpy() into the fixed 64-byte stack buffer temp_addr_str (and the 16-byte iface_ctx.dns_v4_string).

Because the field length is derived from attacker-controlled delimiter positions and was not bounded against the destination buffer, a single field can be far larger than 64 bytes. A malicious or impersonated cellular network (for example a rogue base station) can return a crafted +CGCONTRDP response with an overlong address field, causing strncpy() to write past temp_addr_str on the modem worker thread’s stack, plus an out-of-bounds NUL write at temp_addr_str[addr_len].

No device-side privileges or user interaction are required: the device itself issues the AT+CGCONTRDP=1 query during normal network attach and parses whatever the network returns. The overflow corrupts adjacent stack memory in supervisor context, yielding at minimum a remotely triggerable crash and potentially control-flow hijacking on targets without stack protection.

The fix bounds every field length against its destination buffer (temp_addr_str and dns_v4_string) before each copy, rejecting overlong fields.

This has been fixed in main for v4.5.0

CVE 2026-7656

Broken IPv6 Neighbor Discovery input validation allows spoofed RA/NS/NA acceptance in Zephyr net stack

The IPv6 Neighbor Discovery handlers in subsys/net/ip/ipv6_nbr.c (handle_ra_input, handle_ns_input, handle_na_input) used an incorrect boolean expression that combined the RFC 4861 validity checks with the ICMPv6 code check using the wrong operator precedence: the form was ((length/hop/source/target checks) && (icmp_hdr->code != 0)). Because every legitimate ND message carries ICMPv6 code 0, an attacker setting code == 0 (the normal value) caused the entire predicate to evaluate false, so the packet was never dropped and all of the other checks were silently skipped. The bypassed checks include the mandatory Hop Limit == 255 verification (which proves an ND packet originated on-link and was not forwarded) and, for Router Advertisements, the requirement that the source be a link-local address, as well as multicast-target sanity checks. As a result, an adjacent on-link attacker — and, because the Hop-Limit-255 guard is bypassed, potentially a remote/off-link attacker whose packets would otherwise be rejected — can have forged Router Advertisement, Neighbor Solicitation, and Neighbor Advertisement messages accepted. A forged RA lets the attacker reconfigure the victim’s default router, on-link prefixes (SLAAC), MTU, reachable/retransmit timers, and (with CONFIG_NET_IPV6_RA_RDNSS) DNS servers, while forged NS/NA enable neighbor-cache poisoning, enabling man-in-the-middle, traffic redirection, and denial of service. The flaw is an input-validation/authentication weakness rather than a memory-safety issue: the underlying packet-parsing primitives (net_pkt_get_data, net_pkt_read, net_pkt_skip) are independently bounds-safe and the validated length is the true buffer length, so skipping the length check causes no out-of-bounds access. The defect has existed since the logic was introduced in 2018 and shipped in all releases through v4.4.0; it is fixed by splitting the condition so any failing check drops the packet.

This has been fixed in main for v4.5.0

CVE 2026-10666

Stack buffer overflow in net_ipaddr_parse() IPv4 address-with-port parsing in subsys/net/ip/utils.c

parse_ipv4() in subsys/net/ip/utils.c (reached via net_ipaddr_parse() for strings of the form “a.b.c.d:port”) copies the port substring into a fixed 17-byte stack buffer (char ipaddr[NET_IPV4_ADDR_LEN + 1]) using a length of str_len - end - 1, where str_len is the full, unbounded input length and end is only the (<=15-byte) offset of the ‘:’ delimiter. Because the destination size is never consulted, a crafted address string with a long suffix after the colon (e.g. “1.2.3.4:” followed by hundreds of bytes) causes an out-of-bounds stack write whose length and contents are fully attacker-controlled (memcpy of the suffix plus a trailing NUL), enabling memory corruption and at minimum a denial of service, and potentially control-flow hijack. The parser is reached from the standard socket API (zsock_getaddrinfo / literal-address resolution), DNS server-string configuration, and the eswifi Wi-Fi co-processor DNS-response path, so an application that resolves a network-influenced address string is exposed. The bug was introduced when the parser was added (Zephyr v1.9.0) and shipped in all releases through v4.4.0. The fix removes the unbounded copy and validates the port length before copying into a small dedicated buffer. Note: the equivalent IPv6 “[addr]:port” path in parse_ipv6() retains the same unbounded copy at this commit and remains a separate, still-reachable instance of the defect.

This has been fixed in main for v4.5.0

CVE 2026-10673

Out-of-bounds write in ADIN2111/ADIN1110 OA SPI Ethernet RX frame reassembly

The Zephyr ADIN2111/ADIN1110 10BASE-T1S/T1L Ethernet driver (drivers/ethernet/eth_adin2111.c) reassembles received Ethernet frames in OPEN Alliance (OA) SPI mode by copying device-supplied 64-byte data chunks into a fixed static buffer ctx->buf of size CONFIG_ETH_ADIN2111_BUFFER_SIZE (default 1524 bytes). In eth_adin2111_oa_data_read(), each valid chunk was memcpy’d into ctx->buf[ctx->scur] and the write cursor scur advanced, with no check that scur + len stayed within the buffer. The number of chunks (up to 255, from the BUFSTS RCA field) and the per-chunk length are taken entirely from the frame data received off the wire; the cursor is only reset on a start-of-frame chunk. An attacker on the single-pair Ethernet segment can therefore send a frame whose reassembled size exceeds the configured buffer, causing the driver’s RX offload thread to write attacker-controlled frame bytes past the end of the static buffer into adjacent driver/kernel memory (up to roughly 14.8 KB in the worst case). This is a remotely/adjacently reachable out-of-bounds write (CWE-787) that can corrupt memory and cause denial of service or potentially code execution. The defect was introduced when OA SPI support was added (commit 0ca8b0756b1) and shipped in releases v3.7.0 through v4.4.0. The fix adds a bounds check that drops the oversized frame and resets the cursor before the copy.

This has been fixed in main for v4.5.0

CVE 2026-12629

PL011 UART error interrupts never cleared, enabling an external-peer interrupt-storm denial of service

The ARM PL011 UART driver in drivers/serial/uart_pl011.c fails to acknowledge receive error interrupts. On the PL011, the framing, parity, break, and overrun error interrupts (PL011_IMSC_ERROR_MASK) are cleared only by writing the interrupt-clear register UARTICR; reading the data register clears the RX interrupt and the per-byte RSR status but not the error interrupt status in MIS. The interrupt service routine pl011_isr() acknowledged only the CTS modem-status interrupt and never wrote icr for the error bits, so an asserted error interrupt remains pending after the ISR returns.

When an application enables error-interrupt reporting via the public uart_irq_err_enable() API, an attacker who controls the serial peer can deterministically assert these error bits by injecting line errors on the RX line — a baud/stop-bit mismatch or mid-character break (framing/break error), a flipped parity bit (parity error), or FIFO flooding (overrun error). Because the error interrupt is never cleared, the interrupt line stays asserted and the CPU re-enters pl011_isr() immediately and indefinitely, producing an interrupt-storm livelock from which the core makes no forward progress.

The impact is an availability-only denial of service (permanent hang), reachable from an external or removable UART peer. Exploitation is gated by configuration: the error interrupt is off by default and no in-tree subsystem enables it, so only applications that explicitly call uart_irq_err_enable() on a PL011-based, interrupt-driven port are affected. The fix makes pl011_isr() acknowledge the pending error bits via uart->icr, breaking the loop, and additionally clears the latched RSR status in pl011_err_check().

This has been fixed in main for v4.5.0

CVE 2026-12630

6LoWPAN IPHC uncompression out-of-bounds read on reserved destination addressing mode

Zephyr’s 6LoWPAN IP Header Compression (IPHC) uncompression code contains an out-of-bounds read in get_ihpc_inlined_size() (subsys/net/ip/6lo.c). The destination inline size is looked up in da_inline_size_table, which has 13 entries, using an index built from the M, DAC and DAM bits of the received IPHC dispatch word (iphc & NET_6LO_IPHC_DA_MASK, a 4-bit value of 0-15). The reserved combinations 13, 14 and 15 are not bounds-checked and read past the end of the table.

The iphc word is taken directly from the received frame, and get_ihpc_inlined_size() is reached on every inbound 6LoWPAN frame via net_6lo_uncompress() from the 802.15.4 receive path (subsys/net/l2/ieee802154/ieee802154_6lo.c and ieee802154_6lo_fragment.c). An unauthenticated attacker on the radio/adjacent link can therefore craft a frame whose destination addressing-mode nibble selects an out-of-range index, with no privileges or user interaction.

The out-of-bounds value becomes the computed inline_size, which then drives header reconstruction before the buffer-length check: it is used to dereference *(pkt->buffer->data + sizeof(iphc) + inline_size) and to compute a size_t diff that can underflow, leading to a further out-of-bounds read of the packet buffer and malformed uncompression. The practical impact is a radio-triggerable out-of-bounds read / denial-of-service on the receiver; the leaked byte is not returned to the attacker. The fix rejects any destination index beyond the table, aborting processing of the malformed frame.

This has been fixed in main for v4.5.0

CVE 2026-12631

Broken access-control denial in k_thread_join/k_thread_abort syscall validation in Zephyr kernel

The Zephyr kernel validates the k_thread_join() and k_thread_abort() system calls (declared __syscall in include/zephyr/kernel.h) through thread_obj_validate() in kernel/thread.c. Its default switch branch is the access-denied path, taken when k_object_validate() returns -EPERM (the calling user thread was never granted access to the target thread object) or -EBADF (the supplied pointer is not a registered kernel object of the right type). That branch invoked K_OOPS(K_SYSCALL_VERIFY_MSG(ret, "access denied")), but K_SYSCALL_VERIFY_MSG treats a true expression as success; the non-zero error code ret therefore read as “verified OK”, the kernel oops was never raised, and control fell through to CODE_UNREACHABLE.

Because k_thread_join() and k_thread_abort() are system calls, an unprivileged user-mode thread (under CONFIG_USERSPACE) can reach this denial path directly by calling either syscall on a thread object it does not own. Instead of the offending thread being cleanly terminated, execution reaches __builtin_unreachable() while running in supervisor mode inside the syscall handler.

On Clang builds CODE_UNREACHABLE emits an illegal-instruction trap, so a user thread can deterministically crash the kernel — a locally triggerable denial of service that escapes the userspace sandbox. On GCC builds the path is undefined behavior: the compiler may drop the return-value handling for thread_obj_validate(), so it can return an undefined bool; if that is false, the caller proceeds into the real k_thread_join()/k_thread_abort() implementation for a thread the user was never authorized to access, an access-control bypass.

The fix changes the verification expression to ret == 0, so a denied (non-zero) result now correctly raises K_OOPS and terminates the offending caller.

This has been fixed in main for v4.5.0

CVE 2026-12632

Out-of-bounds read in Zephyr PTP message parsing from unvalidated message type

Zephyr’s Precision Time Protocol receive handler ptp_msg_post_recv() in subsys/net/lib/ptp/msg.c takes the 4-bit message type straight off the wire via ptp_msg_type() (msg->header.type_major_sdo_id & 0xF, range 0-15) and uses it to index the msg_size[] table. That table only defines entries up to PTP_MSG_MANAGEMENT (0xD), giving it ARRAY_SIZE == 14. Before the fix there was no upper-bound check, so the undefined types 0xE and 0xF indexed one or two int slots past the end of the array — an out-of-bounds read of adjacent read-only data.

The out-of-bounds value is then reused as a length: it gates msg_size[type] > cnt, and when it is small or negative it makes cnt - msg_size[type] a large positive budget passed to msg_tlv_post_recv(), whose TLV loop then walks the message suffix past the received bytes, performing further out-of-bounds reads and in-place byte-swap writes on memory beyond the message slab.

The defect is reached directly from the network: ptp_port_event_gen() in subsys/net/lib/ptp/port.c reads a PTP frame with ptp_transport_recv() and calls ptp_msg_post_recv() with the attacker-chosen type. PTP uses UDP multicast or raw Ethernet (0x88F7) and is unauthenticated, so any host on the same link can trigger the indexing on a CONFIG_PTP-enabled node with no preconditions.

The reliably reproducible impact is a denial of service (fault/crash); a limited memory-corruption path exists but depends on the build-specific value adjacent to msg_size[], which the attacker cannot tune. The fix rejects type >= ARRAY_SIZE(msg_size) with -EBADMSG before any indexing.

This has been fixed in main for v4.5.0

CVE 2026-12633

Out-of-bounds write in IPv6 6LoWPAN Context Option handling via unauthenticated Router Advertisement

The IPv6 neighbor-discovery code in subsys/net/ip/ipv6_nbr.c processes the 6LoWPAN Context Option (6CO, RFC 6775) carried inside ICMPv6 Router Advertisements. In handle_ra_6co() the 8-bit context_len field is taken directly from the packet and was never bounded to the RFC maximum of 128. The function computes context->context_len / 8 and then performs memset(context->prefix + context_len, 0, sizeof(context->prefix) - context_len), where context->prefix is a fixed 16-byte array.

With context_len between 136 and 255 (and the option length field set to 3, which the pre-fix validation accepts), context_len / 8 evaluates to 17..31, so the memset length 16 - context_len/8 underflows the unsigned size_t argument to roughly SIZE_MAX. This produces an unbounded out-of-bounds memset that zeroes kernel memory well past the 6lo context structure.

The defect is reachable from unauthenticated, link-local input: any host on the same link can send a crafted Router Advertisement with a 6CO option. The RA handler validates only the option length field before calling handle_ra_6co(), so a single packet triggers the wild write. The code is compiled when CONFIG_NET_6LO_CONTEXT is enabled.

The impact is a reliable remote (adjacent) denial of service via memory corruption, with collateral integrity loss as the memset zeroes contiguous memory before the system faults. Router Advertisements are link-scoped and not forwarded, so the attacker must be on the same link (AV:A). The fix rejects any context_len greater than 128 before the length computation.

This has been fixed in main for v4.5.0

CVE 2026-12634

Out-of-bounds stack write in the settings NVS backend from over-reported nvs_read length

The NVS backend of the Zephyr settings subsystem (subsys/settings/src/settings_nvs.c) reads stored setting-name entries into fixed 74-byte stack buffers and NUL-terminates them with buf[rc] = '\0', where rc is the return value of nvs_read(). Per its contract, nvs_read() returns the full stored entry length (wlk_ate.len), which can exceed the supplied buffer length — only MIN(len, stored_len) bytes are actually copied, but the return value may be much larger, bounded only by the NVS sector size. Three sites (settings_nvs_cache_match(), settings_nvs_load(), and settings_nvs_save()) used this value directly as the NUL index without clamping, so an oversized stored name entry causes a single \0 byte to be written past the end of the stack buffer at an attacker-influenced offset (CWE-787).

The oversized entry cannot arise through the normal settings API, where names are bounded by SETTINGS_MAX_NAME_LEN. It requires an actor able to write the flash that backs the settings partition — a co-resident or untrusted component sharing the flash device, a malicious settings image/restore, or offline/physical flash access (a shared-flash threat model). The malformed entry is parsed when settings_load() runs at boot or subsystem init, or during settings_save().

The out-of-bounds write is a single NUL byte at an offset equal to the crafted entry length (up to the NVS sector size), so the practical impact is a crash or denial of service and limited stack corruption rather than reliable code execution. There is no confidentiality impact, and the path is not reachable from the network through the ordinary settings interface. The fix skips any entry whose nvs_read() length is greater than or equal to the buffer size before performing the NUL store.

This has been fixed in main for v4.5.0

CVE 2026-12999

Infineon Airoc Wi-Fi driver leaks TX buffers on send failure, leading to permanent pool exhaustion

The Infineon Airoc Wi-Fi driver’s transmit callback airoc_mgmt_send() in drivers/wifi/infineon/airoc_wifi.c allocates a net_buf from the fixed airoc_pool for every outbound packet. When whd_network_send_ethernet_data() returns a synchronous failure, the underlying WHD library does not take ownership of the buffer, but the pre-fix driver returned -EIO without releasing it. Each failed transmit therefore permanently leaks one buffer from the pool.

airoc_pool is small and fixed (AIROC_WIFI_TX_PACKET_POOL_COUNT + AIROC_WIFI_RX_PACKET_POOL_COUNT, default 20 buffers) and is shared by WHD’s whd_host_buffer_get callback for both transmit and receive. Once enough send failures have leaked the pool dry, airoc_wifi_host_buffer_get() returns WHD_BUFFER_ALLOC_FAIL for all subsequent allocations, so both transmit and the WHD-driven receive path fail and Wi-Fi connectivity is lost until the device is rebooted.

The leak occurs only on the transmit error path. A Wi-Fi-adjacent attacker can influence the conditions that cause synchronous send failures (for example by deauthenticating/disassociating the station while the local stack continues to attempt transmits), and ordinary transient failures over the device’s lifetime accumulate toward the same state. Reliable on-demand triggering is of high complexity and the impact is availability-only, but the resulting denial of service is permanent and non-recoverable without a reboot.

The fix releases the buffer with airoc_wifi_buffer_release() on the failure branch, returning it to the pool. The commit also removes a redundant k_sem_give() in airoc_mgmt_disconnect(); because data->sema_common is a binary semaphore (limit 1) the duplicate give merely saturated at 1 and had no security impact.

This has been fixed in main for v4.5.0

CVE 2026-13212

Zephyr virtio driver calls an arbitrary function pointer from an out-of-range used-ring descriptor id

The Zephyr virtio driver does not validate the descriptor-chain head id that the virtio device writes into the used ring. In virtio_isr() (drivers/virtio/virtio_common.c), the device-written vq->used->ring[idx].id is used directly as an index into vq->recv_cbs[] and vq->desc[], which are both allocated with exactly vq->num entries. recv_cbs[] holds {cb, opaque} callback entries, and the indexed callback pointer is then invoked as cbe.cb(cbe.opaque, used_len).

Because the id is consumed as a 16-bit value with no bound check, a malicious or compromised virtio backend (an untrusted hypervisor, or an untrusted hardware/peer-processor virtio device on a PCI or MMIO transport) can supply an id far beyond vq->num. This causes an out-of-bounds read of a {function pointer, argument} pair from heap memory beyond recv_cbs[], after which the driver calls that attacker-shaped pointer in the guest’s interrupt context. No guest privileges or user interaction are required; the backend triggers it by writing the shared used ring and raising the queue interrupt.

The result is an arbitrary / attacker-influenced function-pointer call in the Zephyr guest, i.e. a control-flow-hijack primitive that can lead to code execution or, at minimum, a reliable crash. The fix rejects any used-ring id >= vq->num before indexing recv_cbs[]/desc[] or invoking the callback. This affects builds using CONFIG_VIRTIO with the PCI or MMIO transport.

This has been fixed in main for v4.5.0

CVE 2026-13213

Bluetooth HAS: NULL-pointer dereference DoS when a bonded peer reconnects before bt_has_register

The Hearing Access Service (HAS) GATT server in subsys/bluetooth/audio/has.c installs a connection-callback set unconditionally via BT_CONN_CB_DEFINE, so security_changed() runs for every connection that establishes security even before the application has called bt_has_register(). The service attribute pointers hearing_aid_features_attr, preset_control_point_attr, and active_preset_index_attr remain NULL until bt_has_register() resolves them and sets has.registered.

With CONFIG_BT_SETTINGS, settings_set_cb() restores each bonded client’s persisted context at boot and unconditionally sets context->flags to BONDED_CLIENT_INIT_FLAGS (non-zero). When a previously bonded peer reconnects and re-establishes security during the startup window before bt_has_register() has been called, security_changed() sees the non-zero flags and schedules notify_work_handler, which calls bt_gatt_is_subscribed() with a still-NULL attribute pointer. That triggers an assertion (__ASSERT(attr, ...) in bt_gatt_is_subscribed()), or a NULL dereference of attr->uuid when assertions are compiled out.

The result is a remotely triggerable (Bluetooth, adjacent) crash of the HAS peripheral. Exploitation requires the peer to have previously bonded with the device and to reconnect within the boot-time race window before the application registers the service; a peer that reconnects persistently can prolong the outage. Impact is denial of service only, with no memory corruption or information disclosure.

The fix adds an early if (!has.registered) { return; } guard in security_changed(), so no notification work is scheduled until the GATT service is registered and its attribute pointers are valid.

This has been fixed in main for v4.5.0

CVE 2026-13214

Stack buffer overflow in OCPP GetConfiguration key parsing

The OCPP 1.6 client in subsys/net/lib/ocpp/ocpp_j.c contains a stack buffer overflow in parse_getconfig_msg(). When handling a GetConfiguration request from the central system, the handler copied the attacker-controlled JSON "key" string into the caller’s fixed 50-byte stack buffer (skey[CISTR50], declared in subsys/net/lib/ocpp/ocpp.c) using an unbounded strcpy(). The parsed key value points directly into the receive buffer, so its length is bounded only by the message size (CONFIG_OCPP_RECV_BUFFER_SIZE, default 2048).

The GetConfiguration message is delivered over the WebSocket connection that the charge point opens to its configured central system. The reader thread ocpp_wsreader() reads the message into ui->recv_buf and dispatches it to parse_getconfig_msg() via the PDU function table. An attacker who controls the central system endpoint, or a man-in-the-middle on an unencrypted connection, can send a GetConfiguration request whose "key" field exceeds 50 bytes and overflow the reader thread’s stack with attacker-chosen bytes.

The consequence is a remotely triggerable stack smash on the OCPP reader thread: at minimum a denial of service, and plausibly remote code execution depending on build-time hardening such as stack canaries and MPU configuration. The fix replaces the strcpy() with a bounded strncpy(key, payload.key[0], CISTR50 - 1) followed by explicit NUL termination, matching the bounded copies already used by the sibling handlers.

This has been fixed in main for v4.5.0

CVE 2026-13215

Zephyr ext2 mount: unvalidated superblock block size causes out-of-bounds write from a crafted filesystem image

The Zephyr ext2 filesystem driver fails to validate the s_log_block_size field of the on-disk superblock when mounting a filesystem. ext2_verify_disk_superblock() in subsys/fs/ext2/ext2_impl.c checks the magic number, revision, inode size and group counts, but never bounds s_log_block_size. On a successful verify, subsys/fs/ext2/ext2_ops.c computes fs->block_size = 1024 << superblock.s_log_block_size from this attacker-controlled uint32_t, so a crafted value either overflows the shift (undefined behaviour) or yields a block size far larger than CONFIG_EXT2_MAX_BLOCK_SIZE.

That block size is then passed to k_mem_slab_init() by ext2_init_blocks_slab() to carve CONFIG_EXT2_MAX_BLOCK_COUNT blocks out of the fixed static buffer __ext2_block_memory_buffer, whose size is CONFIG_EXT2_MAX_BLOCK_COUNT * CONFIG_EXT2_MAX_BLOCK_SIZE. k_mem_slab_init() does not verify that the requested blocks fit the buffer, and the ext2 wrapper discards its return value, so the slab is laid out past the end of the static buffer. The mount immediately reads block-group, bitmap and inode blocks of fs->block_size bytes each into these slab blocks, producing an out-of-bounds write into adjacent static memory on the first block read.

The entire path is gated only by data read from the mounted image, making this reachable by any attacker who can present a crafted ext2 image to a device that mounts it (for example a removable SD card or storage medium). Because the ext2 driver runs in kernel mode, supplying image bytes yields a supervisor-mode memory-corruption primitive, with impact ranging from denial of service to potential code execution.

The fix rejects s_log_block_size values that overflow the shift (greater than 11) or that produce a block size exceeding CONFIG_EXT2_MAX_BLOCK_SIZE, so the block slab can no longer be initialized larger than its backing buffer.

This has been fixed in main for v4.5.0

CVE 2026-13216

Out-of-bounds stack write in Zephyr virtio PCI driver from unvalidated device-supplied capability length

The virtio PCI driver (drivers/virtio/virtio_pci.c) parses a device’s PCI capability list during driver initialization. In virtio_pci_read_cap() the device-supplied capability length byte cap_len (read from PCI config space via pcie_conf_read()) was only checked with assert(tmp.cap_len == cap_struct_size). That assert resolves to __ASSERT_NO_MSG(), gated by CONFIG_ASSERT, which defaults off in production builds, so the value reached the copy logic completely unvalidated.

The length then drives a loop that copies extra capability dwords into a fixed-size stack buffer supplied by the caller. A cap_len below the 24-byte base struct virtio_pci_cap underflows the unsigned extra_data_words count to a near-SIZE_MAX value, producing an effectively unbounded stack write; a cap_len above the caller’s buffer (up to 255) writes up to roughly 228 bytes of device-controlled data past the buffer. Both are out-of-bounds writes of attacker-controlled content executed in kernel mode during boot-time device probe.

The input originates from the virtio device. In the common deployment where Zephyr runs as a guest under a hypervisor, the device backend is the host, which already fully outranks the guest, so the bug yields no privilege escalation. The exploitable case is a virtio device that is untrusted relative to the Zephyr kernel — an untrusted or physical/passthrough virtio PCIe device on a bare-metal system, or a confidential-computing posture where the guest must defend against the host — where a malicious device can corrupt the kernel stack and potentially achieve code execution or a crash.

The fix replaces the compiled-out assert with a runtime range check rejecting cap_len outside [sizeof(struct virtio_pci_cap), cap_struct_size] before any arithmetic or copy.

This has been fixed in main for v4.5.0

CVE 2026-13217

NULL-pointer dereference in Zephyr OCPP CALLRESULT parsing via unchecked strtok_r/atoi

The OCPP 1.6 client in subsys/net/lib/ocpp/ocpp.c reconstructs a session handle and PDU id from the uid field of a CALLRESULT message. In ocpp_process_server_msg() the code calls atoi(strtok_r(uid, "-", &tmp)) without checking the strtok_r return value. When the server-supplied uid is empty or contains no - delimiter, strtok_r() returns NULL and atoi(NULL) dereferences a NULL pointer, which is undefined behaviour.

The uid originates from network data: parse_rpc_msg() in subsys/net/lib/ocpp/ocpp_j.c JSON-parses a frame received from the OCPP central system over TCP/WebSocket and copies the server-controlled string into the local buffer. A malicious or compromised central system, or a man-in-the-middle on a non-TLS ws:// connection, can return a malformed uid to reach the defect. No authentication beyond the existing server connection (or MITM position) is required, and the reconstructed pointer is membership-validated by ocpp_session_is_valid(), so the impact is limited to the NULL dereference rather than arbitrary pointer use.

On Zephyr targets that trap access to address 0 (MMU/MPU platforms or CONFIG_NULL_POINTER_EXCEPTION_DETECTION), the dereference faults inside the OCPP reader thread and invokes the fatal handler, producing a remote denial of service of the charge point; on bare targets where address 0 is readable the call returns 0 and is benign, so the impact is availability-only and platform-conditional.

The applied fix guards only the first atoi(); the second strtok_r(NULL, "-", &tmp) followed by pdu = atoi(buf) in the same function remains unguarded and the identical NULL dereference is still reachable from the same network input when the uid has a first token but no second --delimited token. A complete fix should validate the second token as well.

This has been fixed in main for v4.5.0

CVE 2026-13343

Uninitialised stack memory disclosure in the MIDI 2.0 UMP Stream responder

The UMP Stream responder library in lib/midi2/ump_stream_responder.c builds reply packets in a 16-byte struct midi_ump (uint32_t data[4]). The builders make_endpoint_info() and make_function_block_info() populate only the first two words (res.data[0] and res.data[1]) and, before this fix, declared their result as an uninitialised local (struct midi_ump res;). The remaining two words (res.data[2], res.data[3]) retain stale stack contents.

Endpoint Info and Function Block Info notifications are UMP Stream messages (UMP_MT_UMP_STREAM), which are 4 words long, so the full 16-byte packet — including the two uninitialised words — is transmitted verbatim by cfg->send(). The responder is driven by attacker-supplied UMP Stream Endpoint-Discovery / Function-Block-Discovery requests via ump_stream_respond(). In the in-tree Network MIDI 2.0 server (subsys/net/lib/midi2/netmidi2.c) these requests arrive as UDP datagrams and, with the default no-authentication endpoint, a remote peer can establish a session and trigger the responses; the same library also serves USB MIDI 2.0 hosts.

Each discovery request causes the device to disclose 8 bytes of its own uninitialised stack memory to the peer, and the request is freely repeatable. This is a confidentiality-only information leak (root cause is use of an uninitialised variable, CWE-457/CWE-908); the leaked words could include residual data or pointer values. There is no memory-corruption, integrity, or availability impact.

The fix zero-initialises both result structs (struct midi_ump res = {0};), so the trailing words are cleared before transmission. These are the only two responder builders that left trailing words unset (send_string() already zeroes its buffer), so the leak is fully closed.

This has been fixed in main for v4.5.0

CVE 2026-13351

net: Maliciously fragmented IPv6 packets can prevent receiving/processing future incoming packets

The Zephyr network stack can be prevented from receiving or processing future incoming packets by sending a few maliciously fragmented IPv6 packets. When net_ipv6_handle_fragment_hdr() triggers an ICMPv6 error response for a malformed fragment, it returns NET_OK without unreferencing the packet, leaking the RX network packet buffer. Each call to k_mem_slab_alloc() lacks a counterpart k_mem_slab_free(), so replaying such a packet a few times exhausts the RX buffer slab, after which the driver repeatedly fails to obtain RX buffers, resulting in a denial of service.

This has been fixed in main for v4.4.0

CVE 2026-13478

Out-of-bounds read in Zephyr ext2 block-bitmap validation from a crafted s_blocks_count

The Zephyr ext2 filesystem driver validates the on-disk block bitmap in ext2_init_fs() (subsys/fs/ext2/ext2_impl.c) by passing fs_blocks = s_blocks_count - s_first_data_block to ext2_bitmap_count_set(). That helper (subsys/fs/ext2/ext2_bitmap.c) treats its argument as a number of bits and reads one bitmap byte per eight bits, but the bitmap buffer (BGROUP_BLOCK_BITMAP) is a single fetched block of only fs->block_size bytes (capacity fs->block_size * 8 bits). s_blocks_count and s_first_data_block are taken verbatim from the superblock and were never bounded against this single-group capacity; ext2_verify_disk_superblock() checks the magic, revision, and block-size shift but not the block count.

A crafted ext2 image with an oversized s_blocks_count (up to ~4 billion, against a maximum 4096-byte block / 32768-bit bitmap) makes ext2_bitmap_count_set() scan roughly 512 MB of memory past the bitmap block — a large out-of-bounds read of the static block slab and adjacent memory.

The defect is reached during mount: ext2_init_fs() is invoked from ext2_mount() (subsys/fs/ext2/ext2_ops.c), the registered .mount operation. Any path that mounts an attacker-supplied ext2 image (removable media, a disk/flash partition, or a downloaded image) triggers it. The kernel-privileged parser operates on attacker-controlled data, so the bug is exploitable wherever untrusted ext2 media can be mounted.

Impact is an out-of-bounds read only: the resulting bit count is compared internally and the mount is rejected, so no attacker-controlled bytes are returned (not a useful information leak). The ~512 MB over-read will almost certainly cross an unmapped or MPU-protected boundary and fault, crashing the system — a denial of service triggered by mounting a single malformed image. The fix rejects any image whose fs_blocks exceeds fs->block_size * 8 before the scan.

This has been fixed in main for v4.5.0

CVE 2026-13479

Out-of-bounds read in LoRaWAN clock-sync AppTimeAns downlink handler

The LoRaWAN application-layer clock-synchronization service parses downlinks in clock_sync_package_callback() (subsys/lorawan/services/clock_sync.c). Its command loop only guarantees that the one-byte command id is in bounds; for the CLOCK_SYNC_CMD_APP_TIME (AppTimeAns) command the handler then reads a 4-byte time correction via sys_get_le32() plus a 1-byte token without checking that 5 bytes remain in the receive buffer (len - rx_pos). A short or crafted AppTimeAns therefore reads up to 5 bytes past the end of the decrypted payload.

The payload (rx_buf/len) is the decrypted application frame delivered to the registered downlink callback (mcps_indication->Buffer/BufferSize). Reaching the handler requires a frame on the clock-sync port that passes LoRaWAN’s MAC integrity check and FRMPayload decryption, so the practical attacker is a malicious or compromised network/application server (the designated sender of AppTimeAns) or a party holding the session keys, rather than an arbitrary radio listener.

The over-read is bounded: the backing store is a fixed 255-byte static buffer, so the few stray bytes do not fault, and the read values (time_correction, token) are used only internally and never transmitted, so there is no disclosure to the attacker and no crash. The sole effect is that a stale token matching ctx.req_token can apply a garbage time_correction to the device’s own clock offset (ctx.time_offset), a minor integrity impact confined to the victim’s time estimate. The fix adds an explicit length check that drops a too-short AppTimeAns. Note the sibling one-byte reads in the periodicity and force-resync handlers remain unguarded with the same negligible impact.

This has been fixed in main for v4.5.0

CVE 2026-13480

Out-of-bounds read in LoRaWAN fragmented data block transport (FUOTA) downlink handler

The LoRaWAN TS004 Fragmented Data Block Transport handler frag_transport_package_callback() in subsys/lorawan/services/frag_transport.c parses downlink command bytes without validating that enough payload bytes remain before each access. The loop’s only bound is rx_pos < len; after consuming the one-byte command id the handler cast rx_buf + rx_pos to a 10-byte struct frag_transport_setup_req, and for a DATA_FRAGMENT command passed &rx_buf[rx_pos] to the fragment decoder, which reads exactly ctx.frag_size bytes — with no remaining-length check in either case.

The fragment size is attacker-chosen in a preceding FRAG_SESSION_SETUP command (ctx.frag_size = req->frag_size, capped at CONFIG_LORAWAN_FRAG_TRANSPORT_MAX_FRAG_SIZE, default 232). rx_buf aliases the 255-byte static MacCtx.RxPayload buffer in the loramac-node MAC layer, while len is the actual decrypted payload length. By padding a downlink with mismatched-index DATA_FRAGMENT filler commands (each advancing rx_pos by three bytes without producing an answer) and appending one matching-index fragment near the end of the payload, an attacker can make the decoder read up to roughly frag_size bytes past the end of RxPayload, copying adjacent static memory into the decoder buffers and the FUOTA flash image.

The handler runs only on downlinks that have already passed the LoRaWAN frame MIC and FRMPayload decryption, so the defect is reachable only by a party holding the device’s session keys (the FUOTA server or an attacker who has compromised those keys). The out-of-bounds bytes are never returned to the sender — the only uplink emitted is a status answer carrying fragment counts — so there is no direct disclosure channel, and on typical flat-memory LoRaWAN MCUs the over-read stays within mapped memory, making a crash unlikely. The impact is therefore a bounded out-of-bounds read with limited confidentiality consequence and no write or control-flow primitive. The fix adds remaining-length guards before each access.

This has been fixed in main for v4.5.0

CVE 2026-13481

Out-of-bounds read in PTP management TLV TIME parsing in Zephyr net PTP

The IEEE 1588 PTP management-message parser in subsys/net/lib/ptp/tlv.c mishandles the PTP_MGMT_TIME management id. In tlv_mgmt_post_recv(), the PTP_MGMT_TIME case casts mgmt_tlv->data to a 10-byte struct ptp_timestamp and reads it (then byte-swaps and writes it back) without first checking that the TLV data field is at least sizeof(struct ptp_timestamp). Every sibling management id in the same switch validates its length first; PTP_MGMT_TIME was the only case lacking that check.

The length passed in is the management data size (tlv->length - 2), and the upstream guard in ptp_tlv_post_recv() only requires tlv->length > 2, while msg_tlv_post_recv() validates only that the TLV fits within the received byte count, not a per-id minimum. A peer on the local PTP segment can therefore send a PTP_MSG_MANAGEMENT message carrying a short PTP_MGMT_TIME TLV (data as small as 2 bytes), causing the parser to read and write 8 bytes beyond the validated data. The message type and TLV contents are taken straight off the wire, so the path is reachable by any adjacent attacker when CONFIG_PTP is enabled.

The over-read and write-back stay within the struct ptp_msg allocation (mgmt_tlv->data lives in the leading mtu[NET_ETH_MTU] union member, so data + 10 lands at most a few bytes past mtu[], inside the same object), so this is an out-of-bounds read of adjacent in-object memory plus a bounded in-place corruption of the message’s parsed timestamp, not past-allocation memory corruption. Impact is limited to minor information exposure of adjacent bytes and corruption of the device’s parsed management TIME value; there is no crash on the access and no reachable reference-count corruption.

The fix adds if (length < sizeof(struct ptp_timestamp)) { return -EBADMSG; } before the cast, matching the other management-id cases and fully closing the receive-path defect.

This has been fixed in main for v4.5.0

CVE 2026-13734

Zephyr WireGuard mutates peer state before anti-replay check, enabling capture-replay endpoint hijack

Zephyr’s WireGuard VPN data-plane receive handler wg_process_data_message() in subsys/net/lib/wireguard/wg_crypto.c validated the anti-replay counter too late. After AEAD decryption of a MESSAGE_TRANSPORT_DATA packet succeeded, the code committed several peer-state changes — update_peer_addr() (endpoint roaming update), the keypair->last_rx/peer->last_rx liveness timers, and keypair_update() (promote next→current and destroy the previous keypair) — and only afterward called wg_check_replay(). On a replayed packet the replay check returned -EINVAL, but none of the preceding mutations were rolled back.

The AEAD tag authenticates content but not freshness, so a replayed-but-authentic transport packet decrypts correctly. An attacker who captures one valid ciphertext off the wire (an on-path or shared-medium observer) can re-inject it from an arbitrary spoofed source address. Reaching the handler requires no credentials: it is driven directly from inbound UDP datagrams via the dispatch in subsys/net/lib/wireguard/wg.c.

Because the state mutations committed before the replay check, the replay repoints the peer endpoint to the attacker-chosen source address (roaming hijack), redirecting the victim’s subsequent outbound tunnel traffic until the legitimate peer’s next packet re-corrects it; it also prematurely destroys the previous keypair and refreshes the RX liveness timer. The tunnel payload stays encrypted under the session keypair, so this is an integrity/availability impact (traffic redirection and session disruption), not payload disclosure. The fix moves wg_check_replay() to immediately after a successful decrypt, before any peer-state mutation, matching the WireGuard specification and the Linux reference implementation.

This has been fixed in main for v4.5.0

CVE 2026-13735

WireGuard keepalive transport-data messages accepted without Poly1305 authentication

Zephyr’s WireGuard implementation in subsys/net/lib/wireguard/wg_crypto.c mishandled keepalive packets. In wg_process_data_message(), any type-4 transport-data message whose payload was exactly 16 bytes (an empty plaintext plus a bare Poly1305 tag, i.e. a keepalive) was accepted and returned immediately, before wg_decrypt_packet() was ever called. The Poly1305 authentication tag was therefore never verified; the only preceding gates were a cleartext receiver-index lookup (get_peer_keypair_for_index() on the attacker-supplied data_hdr->receiver) and a non-cryptographic keypair validity/expiry check.

The path is reachable entirely from the network: inbound UDP on the WireGuard port is dispatched by wg_input() to handle_transport_data() and then wg_process_data_message(). The 32-bit receiver index is transmitted in cleartext in WireGuard handshake and data messages, so an on-path observer learns it directly and an off-path attacker can brute-force it against the UDP port. Given an active receiving-valid session for that index, an attacker could send a 16-byte garbage payload and have it accepted without possessing the session key.

On acceptance the unauthenticated message caused the management layer to observe a spoofed NET_EVENT_VPN_CONNECTED signal (setting peer->first_valid and notifying any net_mgmt listener) and incremented the keepalive-RX statistic. The impact is limited to integrity of this status signal: no plaintext is decrypted or injected, no key is disclosed, and the early-return path did not update the peer endpoint or liveness timers, so there is no traffic-injection, session-takeover, or availability consequence.

The fix removes the pre-decrypt early return so a 16-byte payload flows through wg_decrypt_packet(), which verifies the Poly1305 tag over the empty plaintext, followed by the existing anti-replay check; only an authenticated, non-replayed message is then recognised as a keepalive. Forged keepalives now fail the tag check and are counted as decrypt failures.

This has been fixed in main for v4.5.0

CVE 2026-14366

SiWx91x WiFi driver double-unref / use-after-free of caller-owned TX net_pkt

The Silicon Labs SiWx917 WiFi driver’s transmit callback siwx91x_send() in drivers/wifi/siwx91x/siwx91x_wifi.c frees a network packet it does not own. In the Zephyr TX path the net_pkt is owned by the L2/networking stack; the driver only borrows it to copy the frame bytes into a local net_buf. Before the fix, after transmitting, siwx91x_send() additionally called net_pkt_unref(pkt) on the caller-owned packet, dropping its last reference and returning it to the shared packet pool prematurely. This code path is compiled in by default (CONFIG_WIFI_SILABS_SIWX91X_NET_STACK_NATIVE).

The caller, ethernet_send() in subsys/net/l2/ethernet/ethernet.c, keeps using the packet after the driver returns: it reads net_pkt_get_len(pkt), updates TX statistics, and then performs its own net_pkt_unref(pkt). Because the driver already released the packet, these are use-after-free reads followed by a second unref (a double free). When concurrent network activity recycles the freed slab slot between the two unrefs, the trailing unref decrements a different, live packet’s reference count and frees it, corrupting the net_pkt pool shared by both the receive and transmit paths.

The defect is exercised by ordinary transmission over the native-stack SiWx917 WiFi interface, and an adjacent attacker on the same WiFi network can induce transmissions (for example ARP or ICMP echo replies, or TCP handshakes) to drive the path. The primary observable impact is loss of availability (transmit hangs and crashes from pool corruption), with race-dependent memory corruption of the kernel networking buffer pool. The fix removes the erroneous net_pkt_unref(pkt) from siwx91x_send(); the driver’s receive-path unref, which correctly frees a packet the driver itself allocated, is unaffected.

This has been fixed in main for v4.5.0

CVE 2026-14367

I3C IBI work-node free-list data race between ISR and workqueue thread

The I3C IBI subsystem in drivers/i3c/i3c_ibi_workq.c hands out statically-allocated work nodes through a free-list i3c_ibi_work_nodes_free implemented as a plain sys_slist_t, which provides no synchronization. The allocation helpers (i3c_ibi_work_enqueue, i3c_ibi_work_enqueue_target_irq, i3c_ibi_work_enqueue_hotjoin, i3c_ibi_work_enqueue_controller_request, i3c_ibi_work_enqueue_cb) called sys_slist_get() directly from ISR context, while the workqueue handler i3c_ibi_work_handler() returned nodes with sys_slist_append() from the workqueue thread, with no lock on either side.

Because sys_slist_get() and sys_slist_append() are neither atomic nor interrupt-safe, an IBI interrupt that fires while the workqueue thread is mid-append (or a truly parallel access under CONFIG_SMP) races on the shared list. This corrupts the list linkage: a node may be handed to two consumers, a node may be lost, or the head/tail pointers may be left inconsistent so sys_slist_get() returns a stale or garbage pointer. In the double-hand-out case the subsequent memcpy(ibi_node, ibi_work, sizeof(*ibi_node)) overwrites a node still in flight; a garbage pointer turns the same memcpy into an out-of-bounds write.

The race is driven by I3C bus traffic — IBIs, hot-joins, and controller-role requests originate from target devices on the bus, and I3C supports hot-joining devices. An attacker controlling an I3C peripheral on the board’s chip-to-chip bus can generate high-frequency interrupts timed to collide with the free operation. Exploitation requires physical access to the bus and winning a narrow timing window; the most realistic impact is a crash or hang (denial of service), with memory corruption possible but hard to control.

The fix wraps all free-list sys_slist_get()/sys_slist_append() operations in the new ibi_work_alloc()/ibi_work_free() helpers, each guarded by a k_spinlock (ibi_work_lock), closing the race across ISR and thread contexts.

This has been fixed in main for v4.5.0

CVE 2026-14368

Off-by-one out-of-bounds NUL write in Zephyr LwM2M JSON string parser

The LwM2M JSON content formatter’s get_string() in subsys/net/lib/lwm2m/lwm2m_rw_json.c copies a parsed JSON string into a caller-supplied buffer and NUL-terminates it. The length guard used if (string_length > buflen), which accepts a string whose length is exactly buflen. After memcpy() fills the whole buffer, buf[string_length] = '\0' then writes one byte past the end of the buffer (CWE-787).

The string value and its length are taken directly from the incoming CoAP payload during a LwM2M WRITE: do_write_op_json() parses the payload obtained from coap_packet_get_payload(), and get_string() is invoked from lwm2m_write_handler() (engine_get_string() in subsys/net/lib/lwm2m/lwm2m_message_handling.c) for a LWM2M_RES_TYPE_STRING resource. The destination buf/buflen is either the resource instance’s fixed data buffer (res_inst->data_ptr/max_data_len) or the engine validation buffer (msg->ctx->validate_buf). A LwM2M server (the client’s DTLS peer) can therefore write a string resource with a value whose length equals the target buffer size and force a one-byte overflow.

The overflow is a single out-of-bounds write of the constant byte 0x00 immediately past the resource or validation buffer, corrupting the adjacent byte in memory. It is not an information leak and the written value is fixed, so it is not a direct code-execution primitive, but it can corrupt adjacent state (an adjacent resource value, a length/flag field, or a struct field) and cause data corruption or a crash. Triggering the write is deterministic; the resulting impact depends on memory layout.

The fix changes the guard to string_length >= buflen, rejecting the exact-length case and aligning the JSON formatter with the other content formatters (lwm2m_rw_plain_text.c, lwm2m_rw_oma_tlv.c, lwm2m_rw_senml_json.c, lwm2m_rw_cbor.c, lwm2m_rw_senml_cbor.c), which already used the correct boundary check.

This has been fixed in main for v4.5.0

CVE 2026-14696

Ethernet bridge RX packet leak enables denial of service via RX buffer-pool exhaustion

When Ethernet bridging is enabled (CONFIG_NET_ETHERNET_BRIDGE), eth_bridge_input_process() in subsys/net/l2/ethernet/bridge/bridge_input.c decides how each frame received on a bridge member interface is handled. For frames that must also be delivered to the local stack, the code called eth_bridge_handle_locally() and returned NET_OK. That helper does not consume the packet — it only calls bridge_iface_recv() (via virtual_recv()), which returns NET_CONTINUE without taking ownership of pkt.

The NET_OK verdict then propagates through ethernet_recv() up to processing_data() in subsys/net/ip/net_core.c, where NET_OK is interpreted as “the packet was consumed, do not free it.” Because no consumer actually took ownership, the RX net_pkt is never returned to the pool and is leaked. The concretely reproducible leak occurs for frames whose EtherType has no registered L3 handler when CONFIG_NET_ETHERNET_FORWARD_UNRECOGNISED_ETHERTYPE is set (default y when CONFIG_NET_SOCKETS_PACKET is enabled): the fall-through L3 dispatch does not overwrite the NET_OK verdict, so ethernet_recv() returns NET_OK and the buffer is never released.

Any device on a bridged L2 segment can emit broadcast/multicast frames carrying an arbitrary EtherType with no authentication. Each such frame permanently consumes one buffer from the finite RX pool (CONFIG_NET_PKT_RX_COUNT), so a brief broadcast flood exhausts the pool and the device can no longer receive traffic until it is rebooted — a persistent denial of service. There is no confidentiality or integrity impact.

The fix makes eth_bridge_handle_locally() propagate the real net_verdict and return NET_CONTINUE for locally-kept frames, writing the bridge interface back through a new dst_iface out-parameter so the packet follows the normal receive path and is unreferenced exactly once.

This has been fixed in main for v4.5.0

CVE 2026-14697

IPv6 Neighbor Solicitation packet leak causes TX pool exhaustion denial of service

net_ipv6_send_ns() in subsys/net/ip/ipv6_nbr.c allocates a transmit net_pkt for a Neighbor Solicitation. When it is called with a data packet pending on an unresolved neighbor and that neighbor’s pending_queue is already non-empty (an NS is already outstanding), the function appends the data packet and returns early without ever sending the NS via net_send_data() or releasing it with net_pkt_unref(). The freshly allocated NS net_pkt and its attached TX buffers are held only by a local variable and are leaked permanently, never returning to CONFIG_NET_PKT_TX_COUNT / CONFIG_NET_BUF_TX_COUNT.

The leaking branch sits on the normal IPv6 transmit path: net_ipv6_prepare_for_send() (called from net_if.c) invokes net_ipv6_send_ns() for any outbound or forwarded IPv6 packet whose next hop is not yet in the neighbor cache. An on-link (adjacent) attacker can drive it deterministically by sending a burst of request packets (for example ICMPv6 echo requests or UDP datagrams) that all spoof a single non-existent on-link source address: the node generates a reply to each, the first reply queues an NS, and every subsequent reply during the roughly three-second INCOMPLETE resolution window takes the leaking branch and loses one TX packet. Router-configured nodes forwarding attacker traffic toward a non-existent on-link host leak identically.

Because the leaked packets are never reclaimed and CONFIG_NET_PKT_TX_COUNT defaults to only 4 (14 for Ethernet), a brief low-rate burst exhausts the TX pool. Once exhausted the node can no longer allocate any transmit packet and cannot send TCP/UDP, ARP/ND, or any reply at all, producing a complete and persistent network denial of service that does not self-heal until reboot. The fix releases the unsent NS packet with net_pkt_unref(pkt) before the early return.

This has been fixed in main for v4.5.0

CVE 2026-14986

Under embargo until 2026-09-04

CVE 2026-15460

Missing channel-state validation in Zephyr Bluetooth Classic L2CAP receive path

The Bluetooth Classic (BR/EDR) L2CAP receive handler bt_l2cap_br_recv() in subsys/bluetooth/host/classic/l2cap_br.c dispatched inbound data PDUs based only on the destination channel ID, without checking that the target channel had reached the BT_L2CAP_CONNECTED state. A dynamic channel is assigned its RX CID and added to the connection’s channel list while still in BT_L2CAP_CONNECTING (and later BT_L2CAP_CONFIG) — before configuration completes and, for PSMs that require security, before the peer is authenticated (l2cap_br_conn_req()).

Because the channel is already findable by bt_l2cap_br_lookup_rx_cid() during this window, a remote peer within radio range can send a data PDU addressed to that CID and have it processed on a not-yet-established channel. The dispatch keys off channel fields (BR_CHAN(chan)->rx.mode, rx.mps) that are only initialized during configuration by l2cap_br_conf(); since channel objects are pooled and bt_l2cap_br_chan_del() does not reset rx.mode or the reassembly buffer _sdu, a reused channel can carry stale state into the CONNECTING window and route the frame into the retransmission/flow-control path (bt_l2cap_br_ret_fc_recv()) with stale parameters and a possibly stale _sdu pointer.

The impact is delivery of attacker data to upper-layer protocol handlers on a half-open (and possibly unauthenticated) channel, plus operation on stale or partially initialized channel state on reused channel objects — leading to channel/link teardown (denial of service) and, in the stale-_sdu case, a dangling-pointer condition. The fix adds an explicit BR_CHAN(chan)->state < BT_L2CAP_CONNECTED guard that drops any data received before the channel is fully connected.

This has been fixed in main for v4.5.0

CVE 2026-15461

Type confusion in Zephyr HL78xx GNSS NMEA driver causes wild-pointer write from GNSS input

The Sierra Wireless HL78xx modem GNSS driver (drivers/modem/hl78xx/, later drivers/modem/vendor_standalone/hl78xx/) embeds a generic struct gnss_nmea0183_match_data match_data inside struct hl78xx_gnss_data. The generic NMEA0183 match helper (drivers/gnss/gnss_nmea0183_match.c) requires that context to be the first member because its callbacks cast user_data directly to struct gnss_nmea0183_match_data *. In the affected releases match_data was the second member (after const struct device *dev), so it sat at a non-zero offset while gnss_nmea0183_match_init() initialized it at the correct address. The registered NMEA handlers instead pass the whole device data object (data->devices.gnss->data, offset 0), producing an offset-shifted type confusion between where state is initialized and where the parse callbacks read and write it.

When NMEA sentences from the GNSS receiver are parsed, the GGA/RMC callbacks write parsed fix data into the wrong location within the struct, and the GSV callback (gnss_nmea0183_match_gsv_callback, active under CONFIG_GNSS_SATELLITES) reads its satellites pointer and bound from the wrong offsets — non-pointer bytes of struct hl78xx_gnss_data — and then writes parsed struct gnss_satellite entries through that bogus pointer. This is a write through an uninitialized/wild pointer with a garbage bound.

The NMEA handlers are registered by default (CONFIG_HL78XX_GNSS_SOURCE_NMEA is the default GNSS source) on devices using the HL78xx GNSS. The driver runs in kernel context and the NMEA data originates from the GNSS radio front-end, so a party able to influence the GNSS signal (for example GNSS/GPS spoofing at radio proximity) can drive the kernel-side parser into the faulty write. The most likely impact is a crash (denial of service) because the bogus pointer resolves to a fixed near-NULL value, with adjacent-memory corruption possible on MMU-less targets. Confidentiality is not affected. Exploitation requires the satellites feature to be enabled and active, so attack complexity is high.

This has been fixed in main for v4.5.0

CVE 2026-6682

Integer overflow in FatFs FAT32 volume mount (mount_volume) yields an attacker-controlled file size and out-of-bounds access in Zephyr

Zephyr bundles ChaN’s FatFs (via the zephyrproject-rtos/fatfs west module) as the FAT/exFAT filesystem backing subsys/fs/fat_fs.c. In mount_volume() (modules/fs/fatfs/ff.c), the FAT area size is computed as fasize = ld_32(BPB_FATSz32); ... fasize *= fs->n_fats; — a 32-bit multiply with no overflow check.

A crafted FAT32 volume that sets BPB_FATSz32 = 0x80000001 with two FATs makes the product wrap (to 0x00000002), so the computed data region overlaps the FAT region. Because the pre-multiply value is kept in fs->fsize, the later plausibility check does not catch the wrap.

An attacker who can get the device to mount such a volume (a malicious SD card or USB medium) can place forged directory entries in the overlapped region, causing f_stat()/directory reads to return attacker-controlled file sizes; application code that reads a file using that size as a length then overflows its buffers, giving heap- or stack-based memory corruption during ordinary file operations.

This is a core FAT32 code path with no compile-time gate (FAT12/16/32 is always built), so a default Zephyr FatFs configuration is affected. Upstream FatFs is unmaintained for security purposes (the maintainer did not respond to runZero or JPCERT/CC), so Zephyr carries the fix in its vendored copy. Tracked upstream as CVE-2026-6682.

This has been fixed in main for v4.5.0

CVE 2026-6683

Divide-by-zero in FatFs exFAT sync (sync_fs) crashes Zephyr on a crafted exFAT volume

Zephyr’s bundled FatFs (zephyrproject-rtos/fatfs) supports exFAT when CONFIG_FS_FATFS_EXFAT is enabled. In sync_fs() (modules/fs/fatfs/ff.c) the free-cluster bookkeeping divides by (fs->n_fatent - 2).

The cluster count is read from the exFAT boot region as ncl = ld_32(BPB_NumClusEx) and is validated only against an upper bound (> MAX_EXFAT), never a lower bound, so a crafted exFAT volume with BPB_NumClusEx = 0 yields n_fatent = 2 and a divisor of zero.

Mounting such a volume and performing any write/sync triggers a divide-by-zero (SIGFPE / CPU fault), a denial of service.

The defect is present only when exFAT is compiled in, which is not the default Zephyr configuration; devices that enable exFAT and mount untrusted removable media are exposed. Upstream FatFs is unmaintained for security, so Zephyr carries the fix in its vendored copy. Tracked upstream as CVE-2026-6683.

This has been fixed in main for v4.5.0

CVE 2026-6685

Integer underflow in FatFs dirty-sector cache (f_read/f_write) causes wrong-sector I/O on crafted fragmented media in Zephyr

In FatFs’s read/write path (f_read/f_write in modules/fs/fatfs/ff.c), the dirty-sector cache-refill decision compares fp->sect - sect against the run length cc using unsigned arithmetic. On a fragmented FAT layout where a later cluster maps to a lower absolute sector than the currently cached one, the subtraction underflows (wraps to a large unsigned value), so the guard that decides whether the cached window overlaps the requested range is evaluated incorrectly.

The result is that FatFs flushes or reuses the wrong cached sector, reading or writing file data to/from an incorrect on-disk location — cross-file data corruption and potential disclosure of unrelated file contents, and a path to out-of-bounds behaviour during ordinary file operations.

A crafted volume with a deliberately fragmented cluster chain (attacker-controlled removable media) triggers the condition. This is on the default read/write path (no exFAT or LFN gating). Upstream FatFs is unmaintained for security, so Zephyr carries the fix in its vendored copy. Tracked upstream as CVE-2026-6685.

This has been fixed in main for v4.5.0

CVE 2026-6686

FatFs f_lseek past end-of-file exposes uninitialized/stale cluster contents (deleted file data) in Zephyr

In FatFs (f_lseek in modules/fs/fatfs/ff.c), seeking a file opened for write to an offset beyond its current end extends the cluster chain via create_chain() but does not zero the newly allocated clusters.

FatFs marks the file as larger without initializing the backing sectors, so a subsequent read of the grown region returns whatever was previously on the medium in those clusters — typically the residual contents of deleted files. On a device where a lower-privileged or later actor can read a file that was extended this way, previously deleted or unrelated file data is disclosed (CWE-908, use of uninitialized resource). No memory-safety corruption occurs; the impact is confidentiality of on-media data.

The bug is on the default write path (no exFAT/LFN gating). Upstream FatFs is unmaintained for security, so Zephyr carries the fix in its vendored copy. Tracked upstream as CVE-2026-6686.

This has been fixed in main for v4.5.0

CVE 2026-6687

Stack buffer overflow in FatFs exFAT volume-label read (f_getlabel) via an unvalidated on-disk length in Zephyr

In FatFs’s f_getlabel() (modules/fs/fatfs/ff.c), the exFAT volume-label copy loop is bounded by the raw on-disk byte dj.dir[XDIR_NumLabel] (0–255) rather than the spec maximum of 11 characters.

A crafted exFAT volume that sets XDIR_NumLabel to a large value (e.g. 128) makes the loop read label characters past the 32-byte directory entry and write up to that many UTF-decoded characters into the caller-supplied label[] buffer, overflowing a typical fixed-size label array (e.g. char label[12]/label[24]) on the stack — memory corruption with potential control-flow impact.

In Zephyr this is reachable only in downstream applications: f_getlabel is compiled solely with CONFIG_FS_FATFS_EXTRA_NATIVE_API=y, exFAT must be enabled, and no in-tree Zephyr code calls it (application code supplies the buffer). It is nonetheless a genuine defect in the vendored library, and because upstream FatFs is unmaintained for security Zephyr carries the fix (clamp the label length) in its vendored copy so opted-in applications are protected. Tracked upstream as CVE-2026-6687.

This has been fixed in main for v4.5.0

CVE 2026-15890

Under embargo until 2026-09-21

CVE 2026-15891

NULL pointer dereference in Zephyr MQTT-SN client when removing a non-responsive gateway

The MQTT-SN client keepalive handler process_ping() in subsys/net/lib/mqtt_sn/mqtt_sn.c removes the gateway record after PINGREQ retries are exhausted. It invoked SYS_SLIST_PEEK_HEAD_CONTAINER(&client->gateways, gw, next) but discarded the result. That macro is a pure expression that does not assign to gw, so gw retained its NULL initializer regardless of the list contents.

The code then dereferences the NULL gw (gw->gw_id) and passes it to mqtt_sn_gw_destroy(), reaching k_mem_slab_free(&gateways, NULL). With CONFIG_MEM_SLAB_POINTER_VALIDATE enabled this triggers k_panic(); in the default configuration it performs a write through the NULL pointer (*(char **)mem = slab->free_list;) and corrupts the slab free list. The outcome is a crash/kernel panic or, on targets where address 0 is writable, silent memory-allocator corruption.

The vulnerable branch runs whenever the connected MQTT-SN gateway fails to answer keepalive PINGREQs for the configured number of retries. This condition is controlled by the remote peer: a malicious or compromised gateway, or an on-path/adjacent attacker that advertises itself as a gateway and then stops responding (or blackholes the real gateway’s PINGRESPs), forces the client into the defect. MQTT-SN runs over UDP and no authentication is required.

The impact is a remotely triggerable denial of service (availability) of the affected MQTT-SN client; there is no attacker-controlled data written. The sibling remover process_advertise() uses SYS_SLIST_FOR_EACH_CONTAINER_SAFE and is not affected. The fix assigns the macro’s return value to gw.

This has been fixed in main for v4.5.0

CVE 2026-15892

Heap memory leak in mcumgr settings-management handlers on access-hook rejection leads to denial of service

The mcumgr SMP settings-management group handlers settings_mgmt_read(), settings_mgmt_write(), and settings_mgmt_delete() in subsys/mgmt/mcumgr/grp/settings_mgmt/src/settings_mgmt.c allocate a key_name buffer (and, for read, a data buffer) via k_malloc() when CONFIG_MCUMGR_GRP_SETTINGS_BUFFER_TYPE_HEAP is enabled, relying on the end: label to k_free() them. When CONFIG_MCUMGR_GRP_SETTINGS_ACCESS_HOOK is also enabled and the application access hook rejects a request by returning status MGMT_CB_ERROR_RC, the handler executed return ret_rc; directly, bypassing end: and leaking the heap allocation on every rejected request.

The settings handlers are reachable over the unauthenticated SMP transport (Bluetooth LE, UART, or UDP, depending on product configuration). The access hook is the mechanism applications use to deny unauthorized settings access, and MGMT_CB_ERROR_RC is a common rejection style, so an attacker who can send settings read/write/delete commands that the hook rejects triggers a heap leak on each attempt.

Because the leaked memory is never reclaimed until reboot, a sustained stream of rejected requests monotonically exhausts the kernel heap until k_malloc() fails, denying mcumgr service and impacting any other heap consumer on the device — a denial of service. The impact is availability-only; there is no memory corruption or information disclosure. Only configurations that select the heap buffer type, enable the access hook, and register a hook that returns MGMT_CB_ERROR_RC are affected (the default stack buffer type cannot leak).

This has been fixed in main for v4.5.0

CVE 2026-15893

Zephyr IPv6 Neighbor Discovery zero reachable time from crafted Router Advertisement causes assertion/DoS

net_if_ipv6_calc_reachable_time() in subsys/net/ip/net_if.c derives a randomized ND reachable time from ipv6->base_reachable_time as min_reachable + sys_rand32_get() % (max_reachable - min_reachable), where min_reachable = base/2 and max_reachable = 3*base/2 using integer division. When base_reachable_time is 1, both min_reachable and the modulus collapse so the function returns 0, and net_if_ipv6_set_reachable_time() stores that 0 into ipv6->reachable_time.

The base_reachable_time is attacker-controlled: handle_ra_input() in subsys/net/ip/ipv6_nbr.c accepts the Reachable Time field of an incoming Router Advertisement whenever it is nonzero and <= MAX_REACHABLE_TIME, so a single unauthenticated, link-local RA carrying a Reachable Time of 1 drives the computed reachable time to 0. Router Advertisements are unauthenticated by default and require only adjacency to the target link.

When a neighbor is subsequently confirmed reachable, net_ipv6_nbr_set_reachable_timer() reads the value and executes NET_ASSERT(time, "Zero reachable timeout!"). On builds with CONFIG_ASSERT enabled this triggers a fatal kernel assertion — a remote denial of service; on builds without assertions the reachable timer is armed with K_MSEC(0) and fires immediately, forcing reachable neighbors into perpetual re-solicitation (STALE), degrading Neighbor Discovery. The impact is limited to availability; there is no memory-safety, confidentiality, or integrity consequence.

This has been fixed in main for v4.5.0

CVE 2026-15894

Under embargo until 2026-10-07

CVE 2026-15923

Infinite loop denial of service in Zephyr SDIO byte-I/O from a card-supplied zero max_blk_size

The Zephyr SDIO subsystem function sdio_io_rw_extended_helper() in subsys/sd/sdio.c finishes transfers with a byte-I/O loop that uses size = MIN(remaining, func->cis.max_blk_size) as the per-iteration step. The value func->cis.max_blk_size is decoded directly from the SDIO card’s CIS FUNCE tuple in sdio_decode_cis() and is not validated. When a card reports a maximum block size of zero, size is always 0, remaining never decreases, and the loop spins forever.

The loop is reached from the public SDIO client API used by drivers, including sdio_read_fifo(), sdio_write_fifo(), and the incrementing register read/write helpers, each of which enters the loop while holding the per-card mutex func->card->lock. A card advertising max_blk_size == 0 therefore hangs the calling thread permanently on its first non-block-aligned transfer and never releases the mutex, denying service to the SDIO peripheral (and any subsystem such as Wi-Fi that depends on it) until the device is reset.

The malicious value must come from the SDIO card itself, so the defect is exploitable where a removable SDIO/combo card slot lets an attacker insert a crafted or malfunctioning card (a physical attack vector); on boards with a soldered SDIO peripheral it is not attacker-influenceable. There is no memory-safety, confidentiality, or integrity impact — only a permanent availability loss. The fix returns -EIO when func->cis.max_blk_size is zero, before the loop is entered.

This has been fixed in main for v4.5.0

CVE 2026-15924

Use-after-free / double-free from unsynchronized concurrent access to the TLS client session cache in Zephyr sockets

Zephyr’s TLS socket layer in subsys/net/lib/sockets/sockets_tls.c keeps a single process-global array, client_cache, of cached client sessions that is shared by every TLS socket context. The functions that mutate and read it — tls_session_save(), tls_session_get(), tls_session_cache_reset(), and the settings restore handler — allocate, free, and dereference each entry’s heap buffer (entry->session). Before the fix these accesses were serialized only by the per-socket context mutex ctx->lock (assigned per socket in ctx_set_lock()), which provides no mutual exclusion between different sockets touching the shared cache.

Because CONFIG_NET_SOCKETS_TLS_MAX_CLIENT_SESSION_COUNT defaults to 1, any two concurrent client sockets contend for the same slot. A thread in tls_session_get() reading entry->session inside mbedtls_ssl_session_load() can run concurrently with another thread in tls_session_save() that selects the same entry for reuse and executes mbedtls_free(entry->session) before reallocating — a use-after-free read, and a double-free when two saves evict the same entry. Both corrupt the mbedTLS heap. The cache is reached on ordinary client paths: at connect time via tls_session_store()/tls_session_restore(), and (on main) whenever a TLS 1.3 session ticket arrives during recv()/poll() via tls_session_store_current().

Exploitation requires an application that opts into per-socket client session caching (the TLS_SESSION_CACHE socket option, off by default) and runs concurrent TLS client connections on multiple threads; the timing that opens the window is influenced by the remote peer(s), so a malicious or compromised server can raise session-ticket frequency to widen it. The reliably-demonstrable impact is memory corruption leading to a crash or heap corruption (denial of service). The fix adds a dedicated session_cache_lock mutex taken across every accessor of client_cache, serializing all reads and frees and closing the race.

This has been fixed in main for v4.5.0

CVE 2026-16147

it82xx2 USB device controller submits incomplete OUT transfer buffers, causing use-after-free and event-list corruption

The ITE IT82xx2 USB device-controller driver (drivers/usb/udc/udc_it82xx2.c) mishandles multi-packet OUT transfers on non-control endpoints. In work_handler_out() the active transfer buffer is obtained with udc_buf_peek() (which does not dequeue it); when a full max-packet-size packet arrives but the buffer still has tailroom (the transfer is not yet complete), the pre-fix code both re-arms the endpoint to keep filling that same buf via work_handler_xfer_continue() and simultaneously hands the same, still-being-filled buffer to the upper stack with udc_submit_ep_event().

Because udc_submit_ep_event() transfers ownership of the buffer to the USB device stack (usbd_event_carrier() appends &buf->node to uds_ctx->ep_events, after which the class handler processes and net_buf_unref()s it), the driver continues to DMA subsequent host-controlled OUT packets into a buffer the upper stack may already have freed and recycled — a use-after-free write. In addition, since the buffer was never dequeued, the completing packet runs udc_buf_get() on the same object and submits it a second time, appending &buf->node to the event slist twice (singly-linked-list corruption) and causing a double net_buf_unref().

The IT82xx2 is a USB peripheral controller, so the untrusted USB host controls OUT-transfer packetization and can force this path against any non-control OUT endpoint whose queued buffer exceeds one packet — an ordinary bulk/interrupt pattern. The driver and USB device stack run in kernel context above the external host, giving the host a device-side kernel heap-corruption primitive: a reliable denial of service and, because the written bytes are attacker-controlled, plausible corruption of adjacent net_buf pool memory. The vector is physical (USB attach). The fix defers submission until the buffer is completely filled and lets xfer_work_handler() drive continuation, so each OUT buffer is submitted to the upper stack exactly once.

This has been fixed in main for v4.5.0

CVE 2026-16148

Kernel panic in the it82xx2 USB device controller driver via re-initialization of a busy delayable work item

The ITE it82xx2 USB device-controller driver initialized its bus-suspend detection work with k_work_init_delayable(&priv->suspended_work, suspended_handler) inside it82xx2_enable() (the driver’s .enable op) in drivers/usb/udc/udc_it82xx2.c. This work item is scheduled essentially continuously while the USB bus is active: the interrupt handler reschedules it on every SOF frame and suspended_handler() reschedules itself, so its timeout node is normally linked in the kernel timeout list / a workqueue pending queue.

k_work_init_delayable() (kernel/work.c) unconditionally overwrites the entire k_work_delayable structure, including its timeout and queue linkage, with no busy check. Because it82xx2_disable() does not cancel the work, a normal disable-then-enable cycle re-runs api->enable() (udc_enable() only rejects a redundant enable, not a re-enable after disable) and re-initializes the still-pending work in place, corrupting the kernel timeout/workqueue linked lists and causing a kernel panic.

An external USB host — for example a host performing USB DFU detach (dfu-util --detach) or forcing repeated attach/reset/re-enumeration — drives the udc_disable()/udc_enable() transitions and controls suspend/resume timing, so it can arrange for the suspend work to be pending across a re-enable. This yields an unauthenticated denial of service (kernel panic) reachable across the USB boundary from a removable, physically-connected host, with no confidentiality or integrity impact demonstrated.

The fix moves the k_work_init_delayable() call into the one-time preinit function so the work is initialized exactly once, eliminating the re-initialization of an in-use item.

This has been fixed in main for v4.5.0

CVE 2026-16511

Under embargo until 2026-11-02

CVE 2026-16512

Under embargo until 2026-09-18

CVE 2026-16513

Under embargo until 2026-09-25

CVE 2026-16514

Under embargo until 2026-09-18

CVE 2026-16515

Under embargo until 2026-09-18

CVE 2026-17050

Under embargo until 2026-09-19

CVE 2026-17051

Under embargo until 2026-09-20

CVE 2026-17052

Under embargo until 2026-09-20

CVE 2026-17053

Under embargo until 2026-09-20

CVE 2026-17054

Under embargo until 2026-09-21

CVE 2026-2411

Bluetooth GATT notify/indicate enforces the wrong attribute’s permissions, bypassing encryption/authentication requirements on characteristic values

Zephyr’s Bluetooth host declares a GATT characteristic as two consecutive attributes: a Characteristic Declaration whose permission is hard-coded to BT_GATT_PERM_READ, and a Characteristic Value attribute that carries the application-specified security permissions (e.g. BT_GATT_PERM_READ_ENCRYPT / READ_AUTHEN / READ_LESC). The public notify and indicate APIs explicitly accept either attribute, and passing the declaration is the documented, common idiom. Before sending each notification or indication, the host re-checks link security with bt_gatt_check_perm() against params->attr in gatt_notify(), gatt_indicate(), and gatt_notify_multiple_verify_params() (subsys/bluetooth/host/gatt.c).

When the application passed the Characteristic Declaration attribute, the host correctly redirected the value handle but left params->attr pointing at the declaration, so the security check evaluated the declaration’s permissions (no security required) instead of the value’s. As a result the encryption/authentication/LESC requirement configured on the characteristic value was skipped. The Notify-Multiple path additionally used a mask that omitted the LE Secure Connections requirement.

A remote peer triggers the disclosure by connecting (optionally without pairing or encryption) and writing the Client Characteristic Configuration descriptor to enable notifications or indications, causing the server to emit the protected value over a link that has not reached the required security level. The impact is information disclosure / access-control bypass for characteristic values the application intended to expose only over a secured link; exposure depends on the application declaring encrypt/authen-required notify/indicate characteristics and on the CCC being writable at a lower security tier. There is no memory-safety or availability impact.

The fix adds bt_gatt_attr_resolve_value(), which maps a declaration attribute to the following value attribute before the permission check, and switches the Notify-Multiple path to the full BT_GATT_PERM_READ_ENCRYPT_MASK so the LESC requirement is also enforced.

This has been fixed in main for v4.5.0

CVE 2026-18413

Under embargo until 2026-09-26

CVE 2026-18414

Under embargo until 2026-09-26

CVE 2026-18415

Under embargo until 2026-09-26

CVE 2026-18416

Under embargo until 2026-09-26

CVE 2026-18417

Under embargo until 2026-09-27

CVE 2026-18418

Under embargo until 2026-10-11

CVE 2026-18746

Under embargo until 2026-09-28

CVE 2026-18747

Under embargo until 2026-09-28

CVE 2026-18748

Under embargo until 2026-10-20

CVE 2026-19184

Under embargo until 2026-10-04

CVE 2026-19185

Under embargo until 2026-10-04

CVE 2026-19186

Under embargo until 2026-10-07

CVE 2026-19569

Under embargo until 2026-10-09

CVE 2026-19570

Under embargo until 2026-10-09

CVE 2026-19571

Under embargo until 2026-10-09

CVE 2026-19574

Under embargo until 2026-10-09

CVE 2026-19575

Under embargo until 2026-10-09

CVE 2026-19576

Under embargo until 2026-10-10

CVE 2026-19577

Under embargo until 2026-10-10

CVE 2026-19595

Under embargo until 2026-10-18

CVE 2026-19669

Under embargo until 2026-10-10

CVE 2026-19673

Under embargo until 2026-10-14

CVE 2026-19676

Under embargo until 2026-10-17

CVE 2026-19735

Under embargo until 2026-10-11

CVE 2026-19736

Under embargo until 2026-10-11

CVE 2026-19737

Under embargo until 2026-10-11

CVE 2026-19738

Under embargo until 2026-10-11

CVE 2026-19739

Under embargo until 2026-10-11

CVE 2026-19740

Under embargo until 2026-10-11

CVE 2026-19741

Under embargo until 2026-10-22

CVE 2026-19742

Under embargo until 2026-10-25

CVE 2026-19809

Under embargo until 2026-10-19

CVE 2026-19935

Under embargo until 2026-10-11

CVE 2026-19936

Under embargo until 2026-10-12

CVE 2026-19937

Under embargo until 2026-10-12

CVE 2026-19938

Under embargo until 2026-10-12

CVE 2026-19939

Under embargo until 2026-10-13

CVE 2026-19940

Under embargo until 2026-10-13

CVE 2026-19947

Under embargo until 2026-10-21

CVE 2026-75083

Under embargo until 2026-10-14

CVE 2026-75084

Under embargo until 2026-10-14

CVE 2026-75085

Under embargo until 2026-10-14

CVE 2026-76787

Under embargo until 2026-10-16

CVE 2026-76788

Under embargo until 2026-10-16

CVE 2026-77684

Under embargo until 2026-10-18

CVE 2026-77685

Under embargo until 2026-10-18

CVE 2026-78116

Under embargo until 2026-10-20

CVE 2026-78117

Under embargo until 2026-10-20

CVE 2026-78118

Under embargo until 2026-10-20

CVE 2026-78119

Under embargo until 2026-10-20

CVE 2026-79976

Under embargo until 2026-10-21

CVE 2026-79977

Under embargo until 2026-10-22

CVE 2026-79978

Under embargo until 2026-10-23

CVE 2026-79979

Under embargo until 2026-10-23

CVE 2026-79980

Under embargo until 2026-10-23

CVE 2026-79981

Under embargo until 2026-10-23

CVE 2026-79982

Under embargo until 2026-10-23

CVE 2026-81038

Under embargo until 2026-10-24

CVE 2026-81039

Under embargo until 2026-10-24

CVE 2026-82388

Under embargo until 2026-10-26

CVE 2026-82389

Under embargo until 2026-10-26

CVE 2026-82390

Under embargo until 2026-10-26

CVE 2026-82391

Under embargo until 2026-10-26

CVE 2026-82961

Under embargo until 2026-10-27

CVE 2026-82962

Under embargo until 2026-10-27

CVE 2026-85032

Under embargo until 2026-10-30

CVE 2026-85033

Under embargo until 2026-10-31

CVE 2026-85034

Under embargo until 2026-10-31

CVE 2026-85035

Under embargo until 2026-10-31

CVE 2026-85036

Under embargo until 2026-10-31

CVE 2026-86086

Under embargo until 2026-12-01

CVE 2026-86092

Under embargo until 2026-12-03

CVE 2026-87038

Under embargo until 2026-11-02

CVE 2026-87039

Under embargo until 2026-11-02

CVE 2026-87040

Under embargo until 2026-11-02

CVE 2026-87041

Under embargo until 2026-11-02

CVE 2026-87042

Under embargo until 2026-11-03

CVE 2026-87043

Under embargo until 2026-11-03

CVE 2026-87044

Under embargo until 2026-11-03

CVE 2026-87045

Under embargo until 2026-11-06

CVE 2026-90585

Under embargo until 2026-11-07

CVE 2026-90586

Under embargo until 2026-11-08

CVE 2026-90587

Under embargo until 2026-11-08

CVE 2026-90588

Under embargo until 2026-11-08

CVE 2026-90589

Under embargo until 2026-11-08

CVE 2026-90590

Under embargo until 2026-11-09

CVE 2026-90591

Under embargo until 2026-11-09

CVE 2026-90832

Under embargo until 2026-11-10

CVE 2026-90833

Under embargo until 2026-11-10

CVE 2026-90834

Under embargo until 2026-11-10

CVE 2026-91007

Under embargo until 2026-11-12