| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| The userspace verifier z_vrfy_rtio_sqe_copy_in_get_handles() in subsys/rtio/rtio_syscalls.c (subsys/rtio/rtio_handlers.c before v4.3.0) validated the RTIO object handle and the sqes input array, but not the handle out-parameter. On the first loop iteration it executed *handle = sqe, storing the kernel address of the newly acquired submission-queue entry through a pointer taken verbatim from user mode, with no K_SYSCALL_MEMORY_WRITE check in front of it.
Any user-mode thread that has been granted a struct rtio kernel object can invoke the syscall with an arbitrary address in handle. That is the ordinary way an unprivileged thread uses the RTIO API, for example via sensor_read_async_mempool() or the async ADC helpers, which call rtio_sqe_copy_in_get_handles() internally. The store happens in supervisor mode before any submission-entry validation, so it fires regardless of whether the SQE contents are subsequently rejected. Only builds with CONFIG_USERSPACE and CONFIG_RTIO are affected; without CONFIG_USERSPACE the verifier is not compiled and the caller is already privileged.
The write address is fully attacker-chosen and the written value is a pointer into the caller's own RTIO ring, whose contents the caller controls (the following *sqe = sqes[i] copies an attacker-supplied struct rtio_sqe into that slot). This yields a write-what-where primitive placing a pointer to attacker-controlled data at any kernel address, sufficient to corrupt kernel function pointers, thread structures, or memory-domain partition tables, and thus to escalate from user mode to kernel mode, defeating the isolation boundary CONFIG_USERSPACE is meant to enforce. At minimum it is a reliable kernel memory-corruption and crash primitive. The reporter reproduced the write on qemu_x86: a K_USER thread changed a supervisor global from NULL to a live kernel SQE pointer.
The fix adds K_SYSCALL_MEMORY_WRITE(handle, sizeof(*handle)) (guarded by the existing optional-NULL semantics) before the loop, so the destination must lie in the calling thread's writable memory domain or the thread is terminated by K_OOPS. The neighbouring verifier z_vrfy_rtio_cqe_get_mempool_buffer(), which checked its buff/buff_len out-parameters only for read although the implementation writes through them, was hardened separately by bea93400138 ("rtio: syscalls: validate output params as writable"); that residual was materially weaker, since a read check still confines the target to the caller's own memory domain. |
| The ADC API requires each driver to reject a sampling sequence whose destination buffer is too small: the buffer_size field of struct adc_sequence in include/zephyr/drivers/adc.h documents that "the driver must ensure that samples are not written beyond the limit and it must return an error if the buffer turns out to be not large enough". The NXP MCUX LPADC driver did not honour that contract. mcux_lpadc_start_read() in drivers/adc/adc_mcux_lpadc.c performed no buffer-size check at all before assigning data->buffer = sequence->buffer. Each completed conversion then stores one 16-bit sample per enabled channel per sampling round through an unbounded *data->buffer++: in mcux_lpadc_isr() for interrupt-driven builds, and in mcux_lpadc_dma_callback() for DMA-driven builds on releases that have the DMA path. A sequence selecting two channels with a two-byte buffer, for example, has its second sample written past the end of the buffer.
On a build with CONFIG_USERSPACE, adc_read() and adc_read_async() are system calls. The handler in drivers/adc/adc_handlers.c copies the sequence in from user memory, verifies only that [buffer, buffer + buffer_size) is writable by the calling thread, and rejects a user-supplied options->callback; it deliberately leaves the size arithmetic to the driver. A user-mode thread that has been granted access to an LPADC device object therefore fully controls channels, buffer, buffer_size and options->extra_samplings, and can request far more samples than its buffer can hold: up to channels * 65536 samples into a two-byte buffer, since the sample pointer is only rewound on a repeat sampling, never on the extra samplings of a sequence.
The resulting stores are performed by the driver in kernel mode (in the ADC interrupt handler or the DMA completion callback), where the MPU does not restrict the thread's memory domain, so the write walks linearly out of the user partition and into adjacent memory such as other partitions, kernel data or thread stacks. The impact is kernel-memory corruption of attacker-chosen length at an attacker-chosen offset, a plausible privilege-escalation and denial-of-service primitive from an unprivileged user-mode thread. Builds without CONFIG_USERSPACE are affected only as a caller-side robustness defect, since the application itself supplies the buffer.
The fix calls the new shared helper adc_sequence_validate_buffer() in drivers/adc/adc_common.c from mcux_lpadc_start_read(). The helper computes active_channels sizeof(uint16_t) (1 + extra_samplings) and returns -ENOMEM before any sampling is started. |
| The ADC API requires each driver to reject a sampling sequence whose destination buffer is too small: the buffer_size field of struct adc_sequence in include/zephyr/drivers/adc.h documents that "the driver must ensure that samples are not written beyond the limit and it must return an error if the buffer turns out to be not large enough". The ADI MAX32 driver did not honour that contract. start_read() in drivers/adc/adc_max32.c compared buffer_size, a byte count, against a sample count ((1 + extra_samplings) channels), ignoring sizeof(uint16_t), so it accepted a buffer half the required size. The samples are then stored through the uint16_t data->buffer by Wrap_MXC_ADC_GetData(), which writes two bytes per sample and advances the pointer by one uint16_t: in adc_max32_start_channel() for synchronous reads, and in adc_max32_isr() for asynchronous ones. A sequence selecting two channels with a two-byte buffer, for example, passes the check and has its second sample written past the end of the buffer.
On a build with CONFIG_USERSPACE, adc_read() and adc_read_async() are system calls. The handler in drivers/adc/adc_handlers.c copies the sequence in from user memory, verifies only that [buffer, buffer + buffer_size) is writable by the calling thread, and rejects a user-supplied options->callback; it deliberately leaves the size arithmetic to the driver. A user-mode thread that has been granted access to a MAX32 ADC device object therefore fully controls channels, buffer, buffer_size and options->extra_samplings, and can make the driver write twice as many bytes as its buffer holds. Because the check scales with extra_samplings, the overrun equals the length of the buffer itself, up to channels * 65536 bytes past its end, since the sample pointer is only rewound on a repeat sampling, never on the extra samplings of a sequence.
The resulting stores are performed by the driver in kernel mode (in the system call itself, the ADC context timer, or the ADC interrupt handler for asynchronous reads), where the MPU does not restrict the thread's memory domain, so the write walks linearly out of the user partition and into adjacent memory such as other partitions, kernel data or thread stacks. The impact is kernel-memory corruption of attacker-chosen length at an attacker-chosen offset, a plausible privilege-escalation and denial-of-service primitive from an unprivileged user-mode thread. Builds without CONFIG_USERSPACE are affected only as a caller-side robustness defect, since the application itself supplies the buffer.
The fix replaces that check in start_read() with a call to the new shared helper adc_sequence_validate_buffer() in drivers/adc/adc_common.c, passing sizeof(uint16_t) as the sample size. The helper computes active_channels sizeof(uint16_t) (1 + extra_samplings) and returns -ENOMEM before any sampling is started. |
| ieee802154_send() in subsys/net/l2/ieee802154/ieee802154.c copies the outgoing packet into a single fixed 125-byte transmit buffer (tx_frame_buf_pool, sized IEEE802154_MTU). In builds with CONFIG_NET_L2_IEEE802154_FRAGMENT enabled (the default whenever CONFIG_NET_6LO is set), the branch taken when 6LoWPAN fragmentation is not required performed an unchecked net_buf_add_mem(frame_buf, pkt_buf->data, pkt_buf->len). The only guard was __ASSERT_NO_MSG() inside net_buf_simple_add(), which is compiled out without CONFIG_ASSERT, so an oversized packet silently overran the frame buffer.
The defect is not reachable from the radio: for NET_AF_INET6 packets ieee802154_6lo_encode_pkt() compares the whole packet length against IEEE802154_MTU and takes the fragmentation path when it does not fit, so every buffer copied on the unfragmented branch is within bounds. It is reachable through NET_AF_PACKET sockets bound to an 802.15.4 interface: for NET_SOCK_RAW the 6LoWPAN block is skipped entirely and for NET_SOCK_DGRAM it returns early on the address-family test, leaving no length validation anywhere on the transmit path (net_context_sendto() and net_if_tx() apply none, and pkt_buffer_length() does not clamp the allocation for this L2).
An application — or, in a CONFIG_USERSPACE build, an unprivileged application thread using the zsock_socket()/zsock_sendto() syscalls — can therefore drive a supervisor-mode out-of-bounds write of chosen bytes past the 125-byte pool buffer. With the default CONFIG_NET_BUF_FIXED_DATA_SIZE of 128 bytes the overrun is bounded to roughly ll_hdr_len + 3 bytes; with CONFIG_NET_BUF_VARIABLE_DATA_SIZE a single storage buffer can be as large as CONFIG_NET_PKT_BUF_TX_DATA_POOL_SIZE, making the overrun far larger. The consequence is corruption of memory adjacent to the pool, with a crash or further compromise of kernel state as the practical impact.
The fix validates ll_hdr_len + net_pkt_get_len(pkt) + authtag_len against IEEE802154_MTU before any copy and adds a tailroom-checking copy_pkt_to_frame() helper that returns -EMSGSIZE instead of overrunning the buffer. The same change also linearizes the whole net_buf chain into one MAC frame, so packet storage boundaries no longer become frame boundaries on the wire. |
| The CoAP link-format helper match_path_uri() in subsys/net/lib/coap/coap_link_format.c compares a registered resource path against the URI carried in a Uri-Query href= option. That URI is not NUL terminated, but the inner character loop advanced its index k once per path character without ever testing it against the option length len. When a registered path segment is longer than the supplied URI and the URI is a prefix of it, the loop reads uri[len] and beyond, past the end of the option value.
The path is reached from coap_well_known_core_get_len() and coap_well_known_core_get() via match_queries_resource(), i.e. by any unauthenticated GET /.well-known/core?href=/<prefix> request to a device that serves /.well-known/core (for the CoAP server subsystem, CONFIG_COAP_SERVER_WELL_KNOWN_CORE, default y) and has at least one resource that declares struct coap_core_metadata attributes.
The over-read does not reach the receive buffer. The well-known-core builders parse the query into a stack-local struct coap_option, whose value is a fixed array (value[12], or CONFIG_COAP_EXTENDED_OPTIONS_LEN_VALUE bytes) that the option bytes are copied into, so uri points into that copy. Reading past len therefore reads the unused, uninitialized tail of the array and, when the option fills it, the bytes just past it in the same stack frame. (In the ZoAP library of v1.8.0 to v1.9.x the option value was instead a pointer into the received packet, and the over-read ran past the option inside the packet buffer.)
The impact is bounded. The number of bytes read past the end is limited by the length of the resource path segment, and each additional byte is only read if it happens to equal the next path character, so in practice the over-read is one byte. It also cannot influence the response: returning a match requires the final compared index to be len - 1 or len, both in bounds, so out-of-bounds bytes only ever steer the loop to the next candidate resource. The consequence is undefined behaviour, not information disclosure and not a matching error.
The fix adds a k >= len guard at the top of the inner loop, so every uri[k] dereference is within the option value while still allowing a trailing * wildcard to match a longer path. |
| The Bluetooth host pre-allocates a bt_conn object (a "reservation") from the fixed acl_conns[] pool for every connectable advertising set it starts, in le_adv_start_add_conn() in subsys/bluetooth/host/adv.c. When an incoming peripheral connection completes, find_pending_connect() in subsys/bluetooth/host/hci_core.c looked that reservation up using only the identity and the peer address, with a directed-then-undirected fallback, and had no knowledge of which advertising set actually terminated. With CONFIG_BT_EXT_ADV_MAX_ADV_SET > 1, where the LE Enhanced Connection Complete event is cached until the LE Advertising Set Terminated event identifies the set, this lookup can return the reservation belonging to a different, still-running advertising set.
The defect is reachable over the air by any adjacent device. If an application runs two connectable extended advertising sets on the same identity — one undirected and one directed at a peer P (the common "directed reconnect to a bonded central while still accepting new centrals" pattern) — a remote initiator whose address maps to P connecting on the undirected set causes the directed set's reservation to be consumed instead. The initiator address is chosen freely by the remote device and is not authenticated at link establishment, so no bonding, pairing or prior relationship is required to trigger the mis-attribution.
The undirected set's own reservation is then stranded: its BT_ADV_ENABLED flag has already been cleared, and le_adv_stop_free_conn() — the only path that returns a reservation to the pool — is reached only for sets that are still enabled, so bt_le_ext_adv_stop() returns early and the object is never released. Each occurrence permanently consumes one CONFIG_BT_MAX_CONN slot until bt_disable() or reboot. In application patterns that restart the directed advertising set after each disconnect, an attacker can repeat the sequence until the pool is exhausted, after which connectable advertising fails with -ENOMEM and the device can no longer accept connections. As a side effect the directed set keeps advertising without a reservation, so a genuine connection on it is dropped by bt_hci_disconnect(). The impact is availability only; no memory corruption or information disclosure results, and configurations with a single advertising set are unaffected. |
| The Espressif AES driver drivers/crypto/crypto_esp32_aes.c advertises a multi-session cipher API — it hands out up to CONFIG_ESP32_CRYPTO_AES_SESSIONS_MAX (default 4) independent sessions, each carrying its own key and direction. The AES peripheral, however, has a single chip-global set of key and mode registers, and aes_begin_session() programmed them exactly once, at session start, outside the aes_lock mutex that serialises the transforms. The per-operation handlers aes_ecb_op(), aes_cbc_op() and aes_ctr_op() then ran aes_hal_transform_block() against whatever key and direction the registers happened to hold.
As a result, once a second session is opened with a different key, every subsequent operation on the first session silently uses the second session's key and direction, and no error is reported. Because the key programming happened outside the lock, the correct usage pattern of "each thread does begin, then op, then free on its own session" also races: a second thread's cipher_begin_session() landing between the first thread's begin and its operation makes that operation run under the second thread's key. An attacker who can induce session creation on the device — for example a protocol peer whose connection causes a session key to be installed — can influence which key is resident when a victim operation runs.
The impact is a cryptographic failure rather than memory corruption: data the application believed it was protecting under one key is encrypted or decrypted under another. Where the resident key is one the attacker knows or supplied, the affected plaintext becomes recoverable by that attacker, and the operation's direction (encrypt versus decrypt) can likewise be inverted. Exposure requires an application that holds two or more concurrent sessions with different keys against this driver; no in-tree Zephyr subsystem does so today (the 802.15.4 security layer, which does open two sessions, uses CCM mode, which this driver rejects), so the risk falls on application code that uses the documented multi-session API.
The fix removes all hardware access from aes_begin_session() and reprograms the key and direction at the top of every operation, inside aes_lock, so the chip-global registers can no longer be observed stale by another session. |
| In the Bluetooth Classic Hands-Free Profile Audio Gateway role, hfp_ag_create() in subsys/bluetooth/host/classic/hfp_ag.c starts an SDP discovery toward the peer, which links the embedded ag->sdp_param._node into the SDP client's pending request list in subsys/bluetooth/host/classic/sdp.c. If the subsequent bt_rfcomm_dlc_connect() call fails, bt_hfp_ag_connect() cleaned up by memset()-ing the entire AG object while that node was still linked, and while ag->slc_work could still be queued on the system work queue. An in-flight SDP discovery cannot be cancelled — bt_sdp_discover_cancel() is declared in the public header but has no implementation — so the request stays in the list with its func, pool and uuid pointers zeroed.
The RFCOMM channel passed to bt_hfp_ag_connect() is taken from the peer's own SDP Hands-Free service record in the normal AG flow (bt_sdp_get_proto_param() performs no range validation, and samples/bluetooth/classic/handsfree_ag feeds the value straight through), so a remote BR/EDR device in radio range that advertises an RFCOMM channel outside the valid range 0x01..0x1e deterministically forces bt_rfcomm_dlc_connect() to fail and reach the wipe. Reading the peer's SDP record requires no pairing, and the peer can also trigger the failure by dropping the ACL or racing an RFCOMM DLC on the same DLCI at the right moment.
When the SDP client next advances its request list, sdp_client_discover() dereferences the zeroed parameters — net_buf_alloc(param->pool, …) with a NULL pool and param->uuid->type — and the result-notification paths call through the NULL params->func. Zeroing a linked node also truncates the request list behind a stale tail pointer, and re-arming the freed AG slot re-appends the same node and can produce a self-looping list that hangs the Bluetooth RX thread. The impact is a crash or hang of the Bluetooth stack; the written value is a constant zero at a fixed static address, so no attacker-controlled write is available.
Only builds with CONFIG_BT_CLASSIC and CONFIG_BT_HFP_AG that use the Audio Gateway role are affected. |
| bt_hfp_hf_connect() in subsys/bluetooth/host/classic/hfp_hf.c allocates the per-connection HFP Hands-Free object with hfp_hf_create(), which starts an SDP discovery by passing the object's embedded hf->sdp_param to bt_sdp_discover(). That call links the caller-owned parameter block into the SDP client's request list (session->reqs in subsys/bluetooth/host/classic/sdp.c). If the subsequent bt_rfcomm_dlc_connect() then failed, the affected code wiped the entire object with (void)memset(new_hf, 0, sizeof(*new_hf)) while that discovery was still outstanding, zeroing the sys_snode_t _node that is still linked in the SDP client's list along with the request's func, uuid and pool pointers, and zeroing the hf->slc_work work item that may already be queued on the system workqueue. The discovery could not be retracted instead: bt_sdp_discover_cancel() is declared in include/zephyr/bluetooth/classic/sdp.h but has no implementation.
The failing branch is reachable with peer-supplied input. The documented use of bt_hfp_hf_connect() is to pass the RFCOMM channel read out of the remote Audio Gateway's SDP record; neither bt_sdp_get_proto_param() nor bt_hfp_hf_connect() range-checks that value, so a remote device advertising an HFP AG record with an RFCOMM channel outside 1-30 makes bt_rfcomm_dlc_connect() return -EINVAL and drives the local device into the wipe. The same branch is reached without any application misuse when the remote peer controls connection timing, for example when the RFCOMM session is in BT_RFCOMM_STATE_DISCONNECTING or the ACL is dropped between object creation and the RFCOMM call.
Once the object has been wiped, the Bluetooth host continues to use it: sdp_client_alloc_buf() dereferences session->param->pool, which is now NULL, and the SDP receive paths call session->param->func() through a null function pointer, neither of which is NULL-checked. Because the wipe also clears hf->acl, the pool slot is considered free and a retry or an incoming HFP connection re-appends the same still-linked list node, turning session->reqs into a self-referential list that the host's list walkers never leave. The result is a fault or an unrecoverable hang in the Bluetooth host thread; the stray indirect call is through an all-zero pointer whose target the attacker does not control, so the practical impact is denial of service.
The fix keeps the object allocated and marks it with a new BT_HFP_HF_FLAG_RELEASING flag, letting bt_hfp_hf_discover_cb() release it once the discovery completes instead of destroying it under the SDP client. Only builds with CONFIG_BT_CLASSIC and CONFIG_BT_HFP_HF enabled are affected. |
| The Object Transfer Service server builds its Object Action Control Point (OACP) response indication in oacp_ind_send() (subsys/bluetooth/services/ots/ots_oacp.c). Before this fix the four-byte Checksum Value field was appended whenever the requested op code was BT_GATT_OTS_OACP_PROC_CHECKSUM_CALC, regardless of the result code. The response payload is a bare, uninitialized four-byte stack buffer declared with NET_BUF_SIMPLE_DEFINE(resp_param, sizeof(uint32_t)) in bt_gatt_ots_oacp_write(), and it is filled only on the success path of oacp_checksum_proc_validate(). On every failure path the buffer is still empty, so net_buf_simple_pull_le32() was invoked on a zero-length buffer.
The OACP characteristic and its client characteristic configuration descriptor are declared with BT_GATT_PERM_WRITE and BT_GATT_PERM_READ | BT_GATT_PERM_WRITE, so no pairing, encryption, or authentication is required. Any peer within radio range that can connect to a device running the OTS server can enable OACP indications and write a nine-byte Calculate Checksum command that fails validation — for example by issuing it with no object selected, which returns BT_GATT_OTS_OACP_RES_INV_OBJ. Because oacp_command_decode() accepts the Calculate Checksum op code unconditionally (it is not guarded by CONFIG_BT_OTS_OACP_CHECKSUM_SUPPORT), builds without checksum support are reachable too: they answer BT_GATT_OTS_OACP_RES_OPCODE_NOT_SUP and take the same bad path.
On builds with CONFIG_ASSERT enabled, the zero-length pull trips the assertion inside net_buf_simple_pull_mem() and panics the device, giving a remotely triggered denial of service. On builds with assertions disabled there is no check: the pull returns the untouched stack array and four bytes of uninitialized stack memory are copied into the indication and transmitted to the peer, repeatable on every request. The disclosure is bounded to four bytes per request and the write stays inside the seven-byte res indication buffer, so there is no memory corruption; the impact is uninitialized-memory disclosure plus an availability loss in assertion-enabled configurations. |
| The IPv6 stack derived the long-lived secret key used to compute RFC 7217 stable interface identifiers (gen_stable_iid() in subsys/net/ip/ipv6.c) and RFC 8981 temporary interface identifiers (gen_temporary_iid() in subsys/net/ip/ipv6_pe.c) with sys_rand_get(). That API is documented as not cryptographically secure, while both RFCs require this key to be an unpredictable random value of at least 128 bits — the unpredictability of every generated IID rests entirely on it.
Depending on the platform, sys_rand_get() resolves either to a hardware entropy device or, on boards whose entropy driver sets CONFIG_ENTROPY_NEEDS_PRNG, to the non-cryptographic xoshiro128++ generator in subsys/random/random_xoshiro128.c, which keeps a single global 128-bit state shared with every other sys_rand_get() consumer. Many of those consumers put their output directly on the wire (ephemeral source ports, DNS query IDs, CoAP message IDs and tokens, DHCPv6 transaction IDs, WebSocket masking keys), and the xoshiro state transition is invertible, so an observer of a device's traffic can recover the generator state and reconstruct the IID secret key. Additionally, with the entropy-device backend sys_rand_get() silently falls back to k_cycle_get_32() samples when the entropy driver fails, and returns no error, so the key could be derived from the cycle counter without the caller noticing.
With the key known, stable IIDs become directly computable for any prefix (the remaining HMAC inputs — prefix, interface index, interface name, DAD counter — are guessable), and temporary IIDs become linkable after a cheap search over the mixed-in k_uptime_get_32() value. This defeats the properties both features exist to provide: an attacker can track a device across networks and prefixes, and can predict its address inside a /64 without scanning, removing the usual barrier to directly targeting the host. There is no memory-safety, integrity, or availability impact.
The fix generates the key with sys_csrand_get() and aborts IID generation with -EIO when no cryptographic randomness is available, instead of silently deriving addresses from a predictable key. Both CONFIG_NET_IPV6_IID_STABLE and CONFIG_NET_IPV6_PE are experimental and disabled by default; builds using CONFIG_TEST_RANDOM_GENERATOR remain predictable by design, as sys_csrand_get() is deliberately routed back to the test generator there. |
| Zephyr's OpenThread integration defined CONFIG_OPENTHREAD_JOINER_PSKD with default "J01NME" in modules/openthread/Kconfig.thread (previously subsys/net/l2/openthread/Kconfig, then subsys/net/l2/openthread/Kconfig.thread). J01NME is the example Joining Device Credential used throughout the public OpenThread and Thread commissioning documentation, including Zephyr's own samples/net/sockets/echo_client/README.rst, so it is not a secret. A firmware image built with CONFIG_OPENTHREAD_JOINER_AUTOSTART=y that never overrode the PSKd therefore shipped with a publicly known joining credential compiled in, with no build-time or run-time warning.
The PSKd is the only value that authenticates Thread commissioning: joiner and commissioner run a DTLS-PSK handshake keyed on it, after which the commissioner supplies the full operational dataset. In ot_configure_instance() (modules/openthread/openthread.c), a device with no stored dataset calls otJoinerStart() with the configured PSKd, and ot_joiner_start_handler() calls otThreadSetEnabled() on success. An unauthenticated attacker within 802.15.4 radio range of an uncommissioned device (factory-fresh, after a factory reset, or after a settings erase) can run any stock Thread commissioner registered with J01NME, complete the handshake, and hand the device an attacker-controlled operational dataset.
The device then joins and enables Thread on the attacker's mesh. The attacker becomes its router, seeing, dropping, injecting and spoofing all of the device's IPv6 traffic and impersonating the peers the application expects, while the device is cut off from the network it was meant to join and stays mis-provisioned until it is reset. Only application-layer security with independently provisioned credentials survives. Exposure is limited to builds that enable joiner autostart and leave the PSKd unset; the option is not enabled by default and no in-tree sample uses the vulnerable combination.
The fix removes the default from CONFIG_OPENTHREAD_JOINER_PSKD and adds a BUILD_ASSERT() in modules/openthread/openthread.c, active only under CONFIG_OPENTHREAD_JOINER_AUTOSTART, that fails the build unless a 6-to-32-character PSKd is configured. Products should provision a per-device PSKd from factory data and call otJoinerStart() from the application rather than relying on a credential compiled identically into every unit. |
| The WireGuard VPN service in subsys/net/lib/wireguard/wg.c updated the peer endpoint from a message it could not authenticate. In handle_transport_data(), when the 32-bit receiver index of an incoming transport-data message matched no session, the code both scheduled a new handshake and copied the datagram's source address into peer->endpoint with memcpy(&peer->endpoint, peer_addr, sizeof(peer->endpoint)). Because the receiver index is unknown, there is no key with which such a message could be verified, so the address written into peer->endpoint is entirely attacker-supplied. Every other endpoint update in the implementation happens only after authentication — after the Noise handshake in handle_handshake_init(), or after AEAD decryption plus the anti-replay check in wg_process_data_message().
The WireGuard listener is registered with net_udp_register() using no remote address or port filter and defaults to UDP port 51820 (CONFIG_WIREGUARD_PORT), and wg_ctrl_recv() resolves the peer by interface rather than by any cryptographic identity. Any host able to send a UDP datagram to that port can therefore reach the defect off-path, with no need to observe legitimate traffic: a 16-byte datagram carrying the transport-data type, zeroed reserved bytes and an arbitrary receiver index is sufficient. The vulnerable branch runs whenever the peer has neither a valid handshake nor a current keypair — the state at boot and after a session expires.
A single spoofed datagram redirects peer->endpoint to an address of the attacker's choosing, and all subsequent handshake initiations, keepalives and encrypted data packets are sent there. The legitimate peer never receives the initiations and so never re-authenticates, and nothing resets the endpoint afterwards, leaving the tunnel persistently unusable on devices whose peer does not initiate handshakes itself. The attacker can also relay the redirected initiations to the real peer to place itself in the path of the tunnel, observing ciphertext and traffic metadata and selectively dropping packets. WireGuard's end-to-end cryptography still prevents decryption or forgery of tunnel content, so the impact is loss of availability plus limited confidentiality, not key compromise.
The fix removes the endpoint learning from this path entirely: the stale-session recovery is kept, but the handshake is sent only to an endpoint learnt from an authenticated message, falling back to the configured cfg_endpoint, and the message is dropped when neither is known. CONFIG_WIREGUARD is marked experimental and is not enabled by default. |
| The RTIO streaming completion callback adxl345_process_fifo_samples_cb() in drivers/sensor/adi/adxl345/adxl345_stream.c clamps the amount of FIFO data it will report (read_len = MIN(fifo_bytes, buf_avail)) to the space actually returned by rtio_sqe_rx_buf(), but then queues one 6-byte bus read per hardware FIFO entry at consecutive offsets from read_buf, bounded by the device-reported entry count rather than by the clamp. When the allocated buffer is smaller than a full batch, the driver writes up to 186 bytes past the end of the allocation; a companion defect left data->fifo_total_bytes un-reset at batch start, so an aborted or errored chain could leave a stale write offset that pushes the next batch's writes out of bounds as well.
A short buffer is an expected, first-class outcome of the RTIO allocator: rtio_block_pool_alloc() walks the requested size down one block at a time until a contiguous run is available, and the driver only demands sizeof(struct adxl345_fifo_data) + 6 bytes as its minimum. Any streaming configuration whose mempool block granularity is smaller than a full FIFO batch (204 bytes) — for example the sensor shell's RTIO_DEFINE_WITH_MEMPOOL(sensor_read_rtio, 8, 8, 32, 64, 4) used by sensor stream <dev> on fifo_wm incl, or the 16-byte-block streaming example in the documentation — takes this path whenever the pool is fragmented or the consuming thread falls behind.
The out-of-bounds writes are performed by the I2C/SPI RTIO backend (DMA on many SoCs) into the shared RTIO block pool, corrupting neighbouring live buffers or, for a pool smaller than one batch, memory beyond the pool itself. The written bytes are accelerometer samples and the destination follows the allocation, so there is no controllable corruption primitive: the realistic impact is silent corruption of other RTIO consumers' data or a crash. Reaching it requires a board with an interrupt-wired ADXL345 built with CONFIG_ADXL345_STREAM, plus buffer-pool starvation, which a co-resident local workload can induce but which also occurs spontaneously under load; there is no network, USB or radio path to this code. |
| The arm64 MMU backend invalidates a single virtual address from the TLB with invalidate_tlb_page() in arch/arm64/core/mmu.c, which issued TLBI VAE1{IS} with an operand of virt >> 12. That operand encodes the ASID in bits [63:48], so the invalidation only ever matched ASID 0 (the kernel page tables) and global entries. Memory domains, however, are allocated ASIDs starting at 1 in arch_mem_domain_init(), and every domain-private leaf created by private_map() is mapped with MT_NG, i.e. non-global. A domain's cached translation therefore survived the invalidation.
All callers of invalidate_tlb_page() are demand-paging paths. When the eviction algorithm reclaims a frame, page_frame_prepare_locked() in kernel/mmu.c calls arch_mem_page_out(), which marks the kernel leaf invalid, propagates the invalid descriptor into every domain via sync_domains(), and then flushes only ASID 0. do_page_fault() immediately reuses the very same page frame for the incoming mapping through arch_mem_page_in(). A user thread that had already touched the evicted page — a k_mem_partition added with arch_mem_domain_partition_add(), or its own stack mapped by map_thread_stack() — retains a valid EL0 read/write translation to that physical frame under its own ASID.
Unprivileged code running in a user thread can drive memory pressure to force eviction and then keep accessing the stale address, obtaining read and write access to whatever the kernel subsequently placed in that frame: supervisor data, another memory domain's partition, or another thread's stack. This defeats the isolation that CONFIG_USERSPACE is meant to enforce, allowing information disclosure and memory corruption up to full compromise of the system.
The fix switches the implementation to TLBI VAAE1{IS}, which matches every ASID and both global and non-global entries, so no domain-tagged translation survives a page out, dirty-bit update or access-flag clear. Exploitation requires an arm64 build with both CONFIG_USERSPACE and CONFIG_DEMAND_PAGING enabled, an active backing store, and at least one demand-paged page that is also privately mapped into a user memory domain. The same commit additionally stops globalize_table() from stripping EL0 permission and non-global attributes off domain-private leaves during paging synchronization, which previously caused user threads to lose access to their own partitions and stacks. |
| The DHCPv4 client in subsys/net/lib/dhcpv4/dhcpv4.c maintained its transaction identifier (iface->config.dhcpv4.xid) as a counter rather than as a per-transaction random value. It was seeded once with sys_rand32_get() in dhcpv4_start_internal(), incremented by one in dhcpv4_send_discover() and dhcpv4_send_decline(), and never changed by dhcpv4_send_request() — so every RENEWING and REBINDING exchange reused, unchanged, the identifier that had already appeared on the wire in the broadcast DHCPDISCOVER at boot, and where the identifier did move it moved predictably by one. RFC 2131 4.1 requires a single identifier per transaction, chosen so that clients are unlikely to collide.
This identifier is the client's only unpredictable reply-validation token. net_dhcpv4_input() accepts a BOOTP reply on op, a matching xid, and a matching chaddr (the interface MAC, which is public); the UDP callback is registered with a wildcard remote address and port, and no server-identifier or source-address check is applied. An attacker who learns one identifier — for example any host on the same segment that sees the broadcast DHCPDISCOVER, or a DHCP server or relay on a network the device previously joined — therefore retains a valid token for the device's whole uptime, and can predict the identifiers of later exchanges, without having to observe the unicast renew traffic or race the legitimate server.
A forged reply that passes those checks has effect regardless of client state, because dhcpv4_parse_options() applies its side effects before the state machine gate in dhcpv4_handle_reply(): the router option calls net_if_ipv4_set_gw() and the DNS option calls dns_resolve_reconfigure(). A single accepted datagram can thus repoint the device's default gateway and resolvers at attacker-controlled addresses, redirecting its traffic; during RENEWING or REBINDING a forged DHCPACK re-binds the lease with attacker-chosen parameters and a forged DHCPNAK removes the address and restarts configuration.
The fix draws a fresh sys_rand32_get() value at each point where a transaction begins — entering SELECTING, RENEWING and REBINDING, and at the INIT/INIT-REBOOT entry taken when an interface with a lease goes down — while REQUESTING and DECLINE keep the identifier of the exchange they belong to, as RFC 2131 requires. As a side effect the client also no longer discards a slow server's offer after retransmitting a discover, and the initial-discover backoff now draws its own randomness instead of deriving it from the identifier. |
| The DNS stub resolver in subsys/net/lib/dns/resolve.c binds one UDP socket per configured server when the resolver context is initialised (dns_resolve_init_locked() -> register_dispatcher() -> dns_dispatcher_register()), with the local port left at 0 so the stack picks an ephemeral port exactly once. dns_write() then sends every subsequent query from that same socket without re-binding, so the source port is constant for the lifetime of the resolver context while only the 16-bit transaction identifier (ctx->queries[i].id, drawn with sys_rand16_get()) varies per query. RFC 5452 section 9.2 requires both to vary; with a fixed port, an off-path attacker has to guess only the identifier.
The stable port is cheap to learn. Any single observed query reveals it permanently, and it can also be found by probing the ephemeral UDP range: subsys/net/ip/connection.c answers unmatched unicast UDP with an ICMP port-unreachable, so the resolver's socket is the one port in that range that stays silent. Acceptance of a reply is decided solely by get_slot_by_id() matching the transaction identifier and a CRC-16 of the queried name — the reply's source address is never checked (dispatcher_cb() does ARG_UNUSED(addr)), so forged answers need not even be spoofed to appear to come from the configured server; they only need to reach the resolver's port.
An attacker with UDP reachability to the device (same LAN, Wi-Fi or mesh segment, or a globally routable device address) can therefore flood answers for a predictable query name across the 16-bit identifier space during a query window and race the legitimate server. A win makes the application resolve an attacker-chosen address, and with CONFIG_DNS_RESOLVER_CACHE enabled the forged record is retained for its attacker-chosen TTL. The attack still requires winning that race, and applications that validate TLS certificates for the resolved endpoint are largely protected.
The fix draws a fresh local port before each query to a server that has nothing outstanding: dns_randomize_source_port() unregisters the dispatcher, closes and reopens the socket, and re-registers it bound to port 0 so the system selects a new port, raising the per-query guessing space from roughly 16 to roughly 31 bits. Multicast DNS and LLMNR are excluded because their ports are protocol-fixed. The behaviour is controlled by CONFIG_DNS_RESOLVER_RANDOMIZE_SOURCE_PORT (default y). |
| The Bluetooth HCI USB transport class of the new USB device stack, subsys/usb/device_next/class/bt_hci.c, copies every controller-to-host HCI packet into a USB transfer buffer before enqueuing it on an endpoint. bt_hci_tx_sync_in() allocated that buffer from bt_hci_ep_pool, a fixed pool whose slots are only USBD_MAX_BULK_MPS bytes (64 on full-speed-only builds), and then performed net_buf_add_mem(buf, bt_buf->data, bt_buf->len) with no length check. HCI event and ACL buffers are sized by CONFIG_BT_BUF_EVT_RX_SIZE and CONFIG_BT_BUF_ACL_RX_SIZE and can be far larger, so the copy runs past the end of the 64-byte slot into the adjacent slots and into whatever follows the statically allocated pool.
The device runs the local Bluetooth controller in raw HCI mode (CONFIG_BT_HCI_RAW), so the oversized packets originate over the air. With the HCI buffer sizes recommended for full-featured controllers, an unauthenticated attacker within Bluetooth range can supply the overflowing bytes: a connected peer that negotiates Data Length Extension can send ACL fragments up to CONFIG_BT_CTLR_DATA_LENGTH_MAX bytes, and an attacker that merely broadcasts extended advertising data reaches the same path through LE Extended Advertising Report events, which the controller fragments only to CONFIG_BT_BUF_EVT_RX_SIZE. In the stock configuration the overflow is already reached without any radio traffic: the mandatory Read Supported Commands command-complete event is 70 bytes and overruns the 64-byte slot by six bytes on every host initialization.
Because net_buf_add_mem() validates tailroom only through an assertion unless CONFIG_NET_BUF_HARDENING is enabled, builds with CONFIG_ASSERT=n — the usual production setting — silently corrupt memory adjacent to the USB buffer pool with attacker-chosen contents, which on a target without memory protection between subsystems can lead to arbitrary code execution. Builds with assertions enabled fail the assertion instead and panic, giving a remotely triggerable denial of service. The fix introduces a separate bt_hci_tx_pool sized to MAX(BT_BUF_ACL_RX_SIZE, BT_BUF_EVT_RX_SIZE) for the controller-to-host direction while leaving the host-to-device pool at the endpoint packet size, which is already bounds-checked during reassembly. |
| The MCTP-over-I2C+GPIO target binding in subsys/pmci/mctp/mctp_i2c_gpio_target.c emulates a small pseudo-register file, where the bus controller first writes a register number and then writes or reads payload bytes. The selected register (b->reg_addr) and the data-phase flag (b->rxtx) were never cleared when the controller issued an I2C STOP, so after a completed transfer the target still had a data register selected while the packet buffer that register refers to had already been consumed and set to NULL by mctp_i2c_gpio_target_stop().
A controller on the bus can re-enter the same branch of mctp_i2c_gpio_target_stop() with a NULL buffer using two ordinary transactions and no authentication: first a normal packet write to MCTP_I2C_GPIO_RX_MSG_ADDR followed by STOP (which delivers and frees b->rx_pkt), then a bare read transaction followed by STOP. The failing read_requested() call leaves the stale state untouched and the I2C driver still delivers the STOP callback, which then calls mctp_bus_rx() with a NULL packet. The symmetric case exists on the transmit side, where the stale MCTP_I2C_GPIO_TX_MSG_ADDR selection dereferences b->tx_pkt->end after b->tx_pkt has been cleared and additionally signals b->tx_complete a second time, causing the next outbound packet to be dropped.
The dereference happens in the I2C target interrupt handler, so on any target that traps accesses to address 0 the result is a fatal exception and a device reset or halt — a denial of service inflicted by any device able to act as controller on the shared management bus (for example compromised firmware on a peer MCTP endpoint or on the BMC). On parts where address 0 is readable, the impact degrades to a malformed packet buffer being handed to the MCTP stack and lost transmit packets. No attacker-controlled value is written and no confidentiality or integrity boundary is crossed.
The fix resets b->reg_addr to MCTP_I2C_GPIO_INVALID_ADDR and clears b->rxtx at every STOP, so a new register must be selected before any data phase can run again. Only builds that opt into the experimental CONFIG_MCTP subsystem with CONFIG_MCTP_I2C_GPIO_TARGET=y are affected. |
| The Nuvoton NPCX eSPI driver parses Target-Attached-Flash (TAF) requests in espi_taf_parse() (drivers/espi/espi_npcx.c), taking the 12-bit length field directly out of the request header in the FLASHRXBUF FIFO. For write and RPMC OP1 requests it copied DIV_ROUND_UP(len, 4) words into struct espi_taf_pckt.src, a fixed 16-word (64-byte) member of a static structure, without ever comparing the length against that buffer or against the negotiated eSPI receive payload size (rx-plsize, 64 bytes on all in-tree NPCX/NPCK SoCs). Both the loop counter and the word count were uint8_t, so the copy ran up to 255 iterations.
An eSPI controller that emits a flash request whose header claims a payload larger than the negotiated maximum therefore drives up to 239 32-bit stores past the end of the structure — roughly 956 bytes of out-of-bounds write into .bss — from the eSPI flash receive interrupt handler, while the source reads walk past the end of the 18-word FLASHRXBUF register array into neighbouring eSPI registers. The same unvalidated length was then passed on to flash_write() in espi_taf_npcx_flash_write() with the 64-byte src buffer as source, reading up to 4 KB of adjacent controller SRAM into SPI flash at a host-chosen address, where the host can read it back. A related fixed-size stack buffer in taf_npcx_completion_handler() was overflowed by four bytes on the maximum-size completion path on SoCs whose tx-plsize is 64.
The defect is reachable only from the eSPI host side, and only by a request that violates the negotiated payload size, so it requires host firmware or bus-level control rather than ordinary host software; a compliant maximum-length write fits the buffer exactly and is handled correctly. Because TAF exists specifically so that the embedded controller arbitrates host access to shared flash (espi_taf_check_read_protect() / espi_taf_check_write_protect()), corrupting controller memory from a host-supplied field crosses the trust boundary the feature asserts; consequences range from controller crash or hang to corruption of driver state that gates the flash protection regions.
The fix widens the counters, rejects any request whose word count exceeds both ARRAY_SIZE(FLASHRXBUF) - NPCX_ESPI_FLASH_HEADER_ADDRESS_LEN and NPCX_TAF_RX_PAYLOAD_WORDS (derived from the devicetree rx-plsize), marks such packets with a new invalid_flag so espi_taf_work() answers them with an unsuccessful completion instead of processing them, bounds the transmit payload against FLASHTXBUF, and adds a default case for unsupported cycle types. Only configurations with CONFIG_ESPI_TAF enabled on an NPCX/NPCK target are affected; no in-tree board enables it by default. |
