| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| The Xtensa MMU userspace port never initialized the per-thread page-table pointer thread->arch.ptables during thread setup, while arch_mem_domain_thread_add() in arch/xtensa/core/ptables.c uses a non-NULL value of that field as its "this thread is migrating from another memory domain" predicate. For a thread whose struct k_thread backing memory is not zeroed, the field holds stale data and the kernel takes the migration branch, passing that stale value to reset_region() as the base of a level-1 page table. The equivalent field on x86 is explicitly cleared in z_x86_userspace_prepare_thread() for exactly this reason; the Xtensa port adopted the predicate without the initialization.
The condition is reachable from an unprivileged user-mode thread on a build with CONFIG_XTENSA_MMU, CONFIG_USERSPACE and dynamically allocated kernel objects. k_object_alloc(K_OBJ_THREAD) is a system call and allocates the thread object from the caller's resource pool without zeroing it, and the k_thread_create() verifier accepts such an object while forcing the K_USER option, so the migration branch is taken whenever the recycled heap bytes at that offset are non-zero. Because a user thread can also allocate, fill and free other kernel objects from the same heap, the stale bytes are groomable rather than merely random.
Walking the bogus table in region_map_update() performs an unguarded index into l2_page_tables_counter[], may copy from and write through the attacker-influenceable level-1 base in dup_l2_table_if_needed(), and finally writes a page-table entry through a second pointer read out of that memory — once per page of the new thread's stack. The result is out-of-bounds reads and writes executed in supervisor mode, ranging from an immediate kernel panic (observed as Xtensa EXCAUSE 13 on Intel ADSP DSP firmware with user threads) to corruption of kernel memory and escape from the user-mode sandbox. |
| usbh_device_init() in subsys/usb/host/usbh_device.c rejects a device whose device descriptor advertises bNumConfigurations == 0, but it only logs the error and jumps to the cleanup label without setting the return code. At that point the local err is still 0 from the preceding successful descriptor read, so the function returns success while having skipped usbh_device_set_configuration() — the only place udev->cfg_desc is allocated. The device structure is zero-initialised by usbh_device_alloc(), so udev->cfg_desc remains NULL.
usbh_device_connect() treats the zero return as a successful enumeration and calls usbh_class_probe_device(), which starts its descriptor walk from udev->cfg_desc. usbh_desc_get_next_function() passes that NULL pointer to usbh_desc_is_valid(), whose only guard compares the caller-supplied size, so head->bLength is read through a NULL pointer.
The condition is triggered entirely by the attached peripheral: attaching a USB device (or any USB gadget emulating one) that reports bNumConfigurations = 0 reaches the defect directly from the UHC_EVT_DEV_CONNECTED_FS/_HS event handler in subsys/usb/host/usbh_core.c, with no authentication and no user interaction beyond the physical attachment.
On targets where page zero is unmapped or CONFIG_NULL_POINTER_EXCEPTION_DETECTION is enabled this raises a fatal exception in the USB host thread, denying service to the host. Where address zero is readable, the descriptor walker instead parses low memory as a descriptor chain and can dispatch a class probe() with attacker-influenced values, causing further out-of-bounds reads through the still-NULL udev->cfg_desc in usbh_desc_get_iface(). No attacker-controlled write was identified, so the impact is availability plus a bounded out-of-bounds read. Note that CONFIG_USB_HOST_STACK is marked experimental. |
| The HTTP client's debug helper print_header_field() in subsys/net/lib/http/http_client.c formatted an http_parser data-callback slice with a %s conversion: snprintk(output, len + 1, "%s", str). The parser hands its callbacks a pointer and a length into the caller's receive buffer, and those bytes are not NUL terminated. The size argument of snprintk() bounds only the destination, while the %s conversion measures the source first — lib/os/cbprintf_complete.c uses strlen() when no precision is given, and lib/os/cbprintf_nano.c uses strlen() unconditionally — so the read walks past the end of the header field and past the end of the receive buffer until it finds a zero byte.
The slice reaches the helper from on_url(), on_header_field() and on_header_value(), which the parser invokes on bytes that http_wait_data() read straight off the socket into the application-supplied recv_buf. A malicious or compromised HTTP server, or an attacker able to answer a plaintext request, therefore controls both the header field and the bytes that follow it in the buffer, and can guarantee that no zero byte appears before the buffer ends, forcing the over-read. The in-tree tests/net/lib/http_client suite reproduces the over-read under CONFIG_ASAN=y on an ordinary GET.
The impact is bounded. The whole helper body is compiled only when the HTTP client log module is explicitly set to Debug (CONFIG_NET_HTTP_LOG_LEVEL_DBG=y, which the global CONFIG_LOG_DEFAULT_LEVEL does not set), so default builds are unaffected. No out-of-bounds write occurs and none of the over-read bytes are emitted into the log, so there is no information disclosure; the consequence is an out-of-bounds read that can fault and panic the device where the buffer abuts an unmapped or MPU/MMU-protected region.
The fix copies exactly the len bytes the parser offered with memcpy() and terminates them, instead of relying on a %s conversion (a %.*s precision would not have sufficed, since cbprintf_nano.c measures the source in full before applying it). |
| When CONFIG_THREAD_STACK_MEM_MAPPED is enabled, each thread stack is mapped into the kernel virtual address space with a guard page on either side, and the mapping is released during abort cleanup. k_thread_abort_cleanup() in kernel/thread.c has two paths: a deferred one for self-abort, which stashes the mapping in thread_cleanup_stack_addr/thread_cleanup_stack_sz, and an immediate one for a thread aborted by some other thread. The immediate path called do_thread_cleanup(), which ignored its thread argument and only ever unmapped the stashed address — a value populated exclusively by the self-abort path. The mapped stack of a thread aborted by another thread was therefore never unmapped.
Every k_thread_abort() of a non-current thread leaks the thread's stack region plus its two guard pages from the kernel virtual address space. Threads that end by returning from their entry function take the working self-abort path, so the leak is specific to explicit cancellation — k_thread_abort(), pthread_cancel(), osThreadTerminate(), or the kernel thread kill shell command. Both k_thread_create() and k_thread_abort() are system calls, so with CONFIG_USERSPACE an unprivileged user-mode thread that has been granted a thread object and a user-capable stack object can drive the leak from user mode, consuming a kernel-side resource it cannot otherwise touch.
With CONFIG_KERNEL_VM_SIZE defaulting to 8 MB, a few hundred create/abort cycles exhaust the runtime mapping region. k_mem_map_phys_guard() then returns NULL, and map_thread_stack() guards that only with __ASSERT_NO_MSG(), which is compiled out of production builds; thread creation proceeds to write through a NULL-derived stack base and faults in kernel mode. The impact is a persistent denial of service — a kernel fatal error requiring a reboot — with no memory disclosure or corruption, and it is limited to configurations where memory-mapped thread stacks are enabled (x86 targets with X86_MMU and demand paging disabled, where the option is implied by default). |
| The Renesas RA SCI UART driver's receive-error ISR, uart_ra_sci_eri_isr() in drivers/serial/uart_renesas_ra_sci.c, acknowledged the error condition only when a Zephyr interrupt-driven callback had already been registered. With CONFIG_UART_INTERRUPT_DRIVEN=y and CONFIG_UART_ASYNC_API=n, if data->user_cb is NULL the ISR body is empty: the SCI error flags (SSR.ORER, SSR.FER, SSR.PER) are left set and the ICU event flag R_ICU->IELSR_b[eri_irq].IR is never cleared, so the interrupt re-pends immediately and the ISR is re-entered without end.
The error interrupt is masked at boot by uart_ra_sci_init(), but FSP's R_SCI_UART_Open() unconditionally re-enables it, and uart_ra_sci_configure() - the back end of the public uart_configure() API, enabled by default via CONFIG_UART_USE_RUNTIME_CONFIGURE - performs a Close()/Open() pair without re-masking it. Any runtime uart_configure() therefore leaves the error interrupt live regardless of whether a callback has been set, as does calling uart_irq_err_enable() on its own. An instance used with the polling API, or the window between uart_configure() and uart_irq_callback_user_data_set(), thus runs with the error interrupt unmasked and no callback registered.
The trigger is an ordinary receive line error - overrun, framing, parity, or a break condition - which is fully determined by the device attached to the serial link. A peer able to assert a break or transmit at a mismatched baud rate produces such an error at will, so a single frame is enough to enter the state.
The impact is availability only: the CPU is captured in the re-entering ISR and never returns to thread context, hanging the entire system until an external reset or hardware watchdog fires. There is no memory corruption or information disclosure. Builds using the asynchronous UART API are unaffected, because the fall-through to FSP's sci_uart_eri_isr() clears both the error flags and the pending IRQ. |
| The NXP ENET QoS Ethernet driver in drivers/ethernet/eth_nxp_enet_qos serialises transmission with a binary semaphore, data->tx.tx_sem, taken in eth_nxp_enet_qos_tx() and returned only by the transmit-completion interrupt handler tx_dma_done(). A frame that has been handed to the transmit DMA when the Ethernet link goes down is never written back, so that completion interrupt never arrives. The PHY callback eth_nxp_enet_qos_phy_cb() reacts to carrier loss only by calling net_eth_carrier_set(); it does not stop the DMA, take the descriptors back or release the outstanding frame, and no timeout exists to do so.
The trigger is a link-down/link-up transition that lands while a frame is under the transmit DMA. It can be produced by anyone who controls the link partner or the cable — for example by administratively downing and restoring the attached switch port, or by restarting auto-negotiation from a peer device — and can be made reliable by generating traffic that keeps the device transmitting and repeating the transition. No credentials or user interaction are required, and the trigger is at the link layer rather than over IP.
Because the semaphore is never given back, every subsequent call to eth_nxp_enet_qos_tx() fails with -EBUSY for the remaining life of the boot; restoring the link does not recover the interface, since the link-up path only rewrites MAC_CONFIGURATION. The affected interface continues to receive but can no longer transmit anything — no ARP replies, no TCP acknowledgements, no DHCP renewals — until the device is reset. The net_pkt and the net_buf fragments of the abandoned frame are likewise never unreferenced and never return to their pools. The security-relevant property is persistence: a momentary link flap leaves the interface permanently unable to send, so the denial of service outlasts the attacker's access to the link. There is no confidentiality or integrity impact.
The fix adds a claim/release pair around the outstanding packet so that completion and abandonment are mutually exclusive, aborts an outstanding frame from the PHY callback on carrier loss (stopping the DMA, flushing the MTL transmit queue and reinitialising the descriptor ring), and arms a transmit watchdog, CONFIG_ETH_NXP_ENET_QOS_TX_TIMEOUT_MS, so that any non-completing DMA — not only a carrier loss — reclaims the transmit slot and the packet. |
| The NXP ENET-QOS Ethernet driver's receive work handler eth_nxp_enet_qos_rx() in drivers/ethernet/eth_nxp_enet_qos hands each processed descriptor back to the DMA by writing desc->read.control = rx_desc_refresh_flags, which sets OWN_FLAG and BUF1_ADDR_VALID_FLAG. RDES0 of a receive descriptor is the DMA buffer address in read format, but the hardware write-back aliases that word with the VLAN tag fields (or, for a context descriptor, the timestamp low word), so it is destroyed on every completion. The success path rewrites desc->read.buf1_addr before re-arming; the four drop paths — errored or non-first descriptor, net_pkt_rx_alloc() failure, invalid descriptor length, and net_pkt_get_frag() failure — did not, re-arming the ring slot with the write-back residue as its DMA target address.
The drop paths are driven entirely by received traffic and need no privilege or special configuration. With the defaults for this driver (CONFIG_NET_BUF_DATA_SIZE of 128 and 16 receive descriptors) a single full-size frame consumes twelve ring slots and twelve K_NO_WAIT fragment allocations, so a short burst of frames sent to the device exhausts the receive pools and reaches the allocation-failure paths; a frame with a receive error, or any frame whose first descriptor was already dropped, reaches the non-first-descriptor path.
Once a slot has been poisoned, the DMA writes the next frame received in that slot to an address derived from the frame's VLAN tag fields rather than to a receive buffer. On the supported MCX SoCs that address falls in flash or reserved space, so the transfer raises a fatal bus error that the driver only logs — there is no reset or recovery path — and reception stops permanently until the device is rebooted; where the write is silently discarded instead, the untouched buffer is later handed to the network stack with a length, delivering stale buffer contents as if they were a received frame. A single triggering event is enough to make the condition persistent, so an attacker able to send traffic to the device can permanently disable its network interface. |
| The Mbed TLS crypto shim driver in drivers/crypto/crypto_mbedtls_shim.c drives PSA multipart cipher and AEAD operations. In mbedtls_cbc() and mbedtls_aead(), every error path after a successful psa_cipher_encrypt_setup() / psa_aead_decrypt_setup() returned directly without calling psa_cipher_abort() / psa_aead_abort(). Because the abort is the only thing that reaches mbedtls_cipher_free(), each failed operation permanently leaked the heap-allocated mbedtls_aes_context (roughly 288 bytes) that setup had allocated from Mbed TLS's fixed static heap (_mbedtls_heap[CONFIG_MBEDTLS_HEAP_SIZE], default 512 bytes). With CONFIG_PSA_CRYPTO_PROVIDER_TFM the leaked object is instead a finite secure-partition multipart-operation slot.
The failure is chosen by an unauthenticated attacker. With CONFIG_NET_L2_IEEE802154_SECURITY backed by this driver, ieee802154_decrypt_auth() in subsys/net/l2/ieee802154/ieee802154_security.c invokes cipher_ccm_op() on every received secured frame, with no authentication ahead of the MIC check. A device in radio range can transmit a stream of syntactically valid 802.15.4 frames carrying a bogus MIC; each one makes psa_aead_verify() fail and leaks a block. The CBC path is reachable the same way through the OSDP secure channel, where CONFIG_OSDP_SC_ENABLED selects this shim and osdp_encrypt() / osdp_decrypt() leave out_buf_max unset, so every operation fails inside psa_cipher_update().
Exhausting the Mbed TLS heap — which the default configuration reaches after a couple of failed operations, and any realistic configuration reaches in well under a second of frames — makes all subsequent mbedtls_calloc() calls fail. Every later cryptographic operation in the image then fails permanently, including link-layer security and any TLS/DTLS sharing the same heap, until the device is rebooted. There is no memory corruption or information disclosure: frames with an invalid MIC are still rejected, so the impact is limited to availability.
The fix routes all post-setup failures in mbedtls_cbc() and mbedtls_aead() through a common error: label that calls the matching PSA abort before returning. Setup failures need no change, since the PSA core aborts the operation itself when setup fails. |
| The entropy-device random number generator in subsys/random/random_entropy_device.c returned early with -ENODEV from rand_get() as soon as device_is_ready() failed, before reaching the non-cryptographic system-timer fallback further down the same function. Because z_impl_sys_rand_get() returns void and discards that error — and the public sys_rand_get() syscall is itself declared void, so it cannot report failure — the caller's output buffer was left untouched and its previous contents were handed back as random data. The thin wrappers make this a plain uninitialized read: sys_rand32_get() declares uint32_t ret;, calls sys_rand_get(&ret, sizeof(ret)) and returns ret.
The "not ready" state is a normal runtime condition rather than a fault. Entropy drivers initialise at PRE_KERNEL_1 with CONFIG_ENTROPY_INIT_PRIORITY (default 50), so any earlier initialiser — kernel objects run at priority 30 — observes an unready device; the sibling generator subsys/random/random_xoshiro128.c and z_early_rand_get() in kernel/init.c both already carried an explicit timer fallback for exactly this case. The state is also permanent for the boot when an entropy driver's init hook fails, which several in-tree drivers can do on real hardware (entropy_stm32 on an RNG clock configuration error, entropy_psa_crypto when psa_crypto_init() fails, entropy_nrf_cracen, entropy_esp32, entropy_virtio).
When this occurs, every non-cryptographic random value in the system is either a constant or stale memory. coap_next_token() in subsys/net/lib/coap/coap.c fills a static buffer, so every CoAP token becomes the same .bss constant; tcp_init_isn() in subsys/net/ip/tcp.c returns an uninitialized stack word as the TCP initial sequence number; DHCPv4/DHCPv6 transaction IDs, WebSocket masking keys and RTP SSRCs are similarly degraded, and arch/arm64/core/pac.c derives per-thread Pointer Authentication keys from the same source. Where the bytes are transmitted directly — SSH packet padding, the netmidi2 session nonce, a DTLS connection ID on builds without CONFIG_CSPRNG_ENABLED — stale caller stack or heap contents are disclosed on the wire.
The fix initialises ret to -ENODEV and only calls entropy_get_entropy() when the device is ready, so the not-ready case now falls through to the timer fallback and the buffer is always written. The cryptographically secure path is unchanged: sys_csrand_get() still propagates the error to the caller and never silently degrades. An attacker cannot induce the unready state, but can observe and exploit the resulting non-random values remotely on an affected device. |
| The RTIO submission-queue verifier rtio_vrfy_sqe() in subsys/rtio/rtio_syscalls.c (subsys/rtio/rtio_handlers.c before v4.3.0) accumulated the results of K_SYSCALL_MEMORY() into its valid_sqe flag with &=. That macro returns nonzero on failure (it is built on K_SYSCALL_VERIFY_MSG(), which yields !(expr)), so every buffer permission check contributed the opposite of its result. A submission whose tx.buf, rx.buf or txrx.tx_buf/txrx.rx_buf pointer was not accessible to the calling thread was accepted, while a legitimate, thread-owned buffer caused the calling thread to be killed with a kernel oops.
The verifier runs on behalf of the rtio_sqe_copy_in_get_handles() system call (also reached through the inline rtio_sqe_copy_in() wrapper), so it is driven directly by untrusted user-mode input on any image built with CONFIG_USERSPACE. A user thread that has been granted an RTIO context and an rtio_iodev object — the intended usage of RTIO userspace support, established via rtio_access_grant() — can submit an SQE whose buffer pointer and length name arbitrary kernel memory; the object check on sqe->iodev is unaffected and still passes, and the crafted SQE is forwarded to the bus driver unchanged.
Downstream consumers dereference the pointer with supervisor privilege. The generic I2C and I3C RTIO fallback handlers (drivers/i2c/i2c_rtio_default.c, drivers/i3c/i3c_rtio_default.c) copy sqe.rx.buf/sqe.tx.buf into struct i2c_msg and call i2c_transfer(), and the vendor SPI/I2C RTIO drivers behave equivalently. An RTIO_OP_RX submission therefore has peripheral data written into kernel memory of the attacker's choosing (arbitrary kernel write), an RTIO_OP_TX submission clocks kernel memory out over the bus (kernel memory disclosure), and a pointer to unmapped memory faults the kernel. This is a full escape from the user-mode isolation boundary that CONFIG_USERSPACE exists to enforce.
The defect was masked for years because the in-tree user-mode RTIO tests only exercise RTIO_OP_NOP and mempool-backed reads, neither of which reaches the affected checks. The fix negates each K_SYSCALL_MEMORY() result and additionally validates the TXRX transmit buffer read-only, matching the SPI, I2C and I3C handlers. |
| The nRF link-layer receive paths of the open-source Bluetooth LE Controller did not validate the on-air LENGTH field of a received PDU against the maximum receive length that was used to program the radio. isr_rx_pdu() in subsys/bluetooth/controller/ll_sw/nordic/lll/lll_conn.c enqueued any non-empty data PDU after only the sequence-number and (for encrypted links) MIC checks, and the corresponding isr_rx() handlers in lll_central_iso.c, lll_peripheral_iso.c, lll_sync_iso.c and lll_scan_aux.c did the same for ISO and extended-advertising PDUs. The nRF radio stores the received header verbatim and truncates only the payload at PCNF1.MAXLEN, so a frame declaring more octets than were configured leaves a node buffer whose len field describes more data than the buffer holds.
An attacker within radio range that can emit a hand-crafted link-layer PDU — for a connection, any peer that establishes an unencrypted LE connection to a connectable advertiser; for CONFIG_BT_CTLR_SYNC_ISO, any broadcaster the device is synchronised to — can set the length byte up to 255 while the controller has programmed a much smaller maximum (lll->dle.eff.max_rx_octets, which is 27 until a Data Length Update, or the negotiated cis_lll->rx.max_pdu). No pairing, authentication or user interaction is required. On encrypted links the CCM MIC check normally rejects such a frame first, so the exposure is on unencrypted traffic, which is why the added check is placed after decryption.
The unchecked length reaches hci_acl_encode() in subsys/bluetooth/controller/hci/hci.c, which performs net_buf_add(buf, pdu_data->len) followed by a memcpy() of that many bytes out of the node_rx pool element. Because CONFIG_BT_CTLR_DATA_LENGTH_MAX and CONFIG_BT_BUF_ACL_RX_SIZE both default to 27, this is an out-of-bounds read of up to roughly 228 bytes past the received PDU and a corresponding out-of-bounds write past the HCI ACL buffer into neighbouring buffers of the same pool; the tailroom test in net_buf_simple_add() is only enforced when CONFIG_NET_BUF_HARDENING (or CONFIG_ASSERT) is enabled. On the ISO paths the same unchecked length makes isoal_rx_append_to_sdu() read past the ISO node buffer and emit adjacent controller RAM to the host as ISO data. The practical result is remote memory corruption and controller crashes, plus disclosure of controller memory into the host data path.
The fix rejects any received PDU whose length exceeds the value that was used to configure the radio reception. Note that lll_sync.c (periodic advertising sync reception) was not changed and still enqueues received PDUs without an equivalent bound. |
| The user-mode syscall verifier z_vrfy_i2s_configure() in drivers/i2s/i2s_handlers.c copies the caller's struct i2s_config in from user memory and then, when cfg->frame_clk_freq is 0 (the API's "unconfigure the stream" request), jumps straight to z_impl_i2s_configure() without performing the K_SYSCALL_OBJ(config.mem_slab, K_OBJ_MEM_SLAB) ownership/type check or the block_size sanity check. The mem_slab field is a struct k_mem_slab *, so on that path a completely unvalidated, user-chosen pointer reaches driver code that is entitled to trust it as a real kernel object.
Any thread running under CONFIG_USERSPACE that has access to an I2S device object can trigger this with a single i2s_configure() call. Most in-tree drivers special-case frame_clk_freq == 0 and never store the caller's structure, but drivers/i2s/i2s_litex.c does not: i2s_litex_configure() only logs a warning for a mismatched frame clock, then memcpy()s the caller's config into stream->cfg and marks the stream I2S_STATE_READY. A subsequent i2s_trigger() with I2S_TRIGGER_START enables the FIFO interrupt, and the interrupt handlers i2s_litex_isr_rx() / i2s_litex_isr_tx() then call k_mem_slab_alloc() and k_mem_slab_free() on the attacker-supplied pointer in supervisor context.
The result is an arbitrary-pointer read/write primitive executed by the kernel on behalf of an unprivileged thread — k_mem_slab_alloc() takes a spinlock at slab->lock and reads and writes slab->free_list and the link word inside the returned "block" — which defeats the CONFIG_USERSPACE isolation boundary entirely. The same missing validation is a hazard for out-of-tree I2S drivers that retain the configuration on the unconfigure path. Builds without CONFIG_USERSPACE are unaffected, since the verifier is not used.
The fix zeroes config.mem_slab and config.block_size on the unconfigure path so the driver can never receive an unvalidated pointer, while still allowing a stream to be unconfigured by a caller that does not own a memory slab, as the API documents. |
| The user-mode verification handler for the i3c_transfer() system call, z_vrfy_i3c_transfer() in drivers/i3c/i3c_handlers.c, validated the caller-supplied struct i3c_device_desc in place instead of snapshotting it. It confirmed only that the descriptor was readable by the calling thread (K_SYSCALL_MEMORY_READ()), read target->bus once to check it with K_SYSCALL_OBJ(), and then passed the caller's live pointer down. The implementation z_impl_i3c_transfer() re-reads target->bus from that user memory and dispatches through it as DEVICE_API_GET(i3c, target->bus)->i3c_xfers(...), so the bus that was validated is not necessarily the bus that is used. DEVICE_API_GET() performs no runtime type check outside CONFIG_DEVICE_API_ASSERT builds.
On a build with CONFIG_USERSPACE and an I3C controller driver, a user thread that has been granted an I3C device object can exploit this with a second thread in the same memory domain: one thread calls i3c_transfer() while the other rewrites the bus field of the descriptor. The window between the check and the use spans the num_msgs validation, the message-array access check, a stack VLA allocation and a memcpy() of up to 32 struct i3c_msg, so the race is repeatable and can simply be retried until it lands.
When the substitution succeeds, the kernel dereferences an attacker-chosen pointer as a struct device, loads a function pointer from the attacker-controlled api table and calls it in supervisor mode with attacker-influenced arguments. This yields kernel-mode code execution from an unprivileged user thread — a complete escape of the userspace memory-protection sandbox; lesser outcomes are an arbitrary two-level kernel read or a kernel fault. Builds without CONFIG_USERSPACE, or without an I3C driver, are unaffected.
The fix copies the descriptor into a kernel-stack local with k_usermode_from_copy(), validates the bus in that copy, and dispatches on the copy, so the validated value is provably the value used. |
| The Bluetooth Classic AVDTP layer defers releasing a session's stream endpoints: bt_avdtp_l2cap_disconnected() in subsys/bluetooth/host/classic/avdtp.c ends with k_work_submit(&session->_release_work), placing the embedded struct k_work on the system work queue's pending list. When the same struct bt_avdtp object is immediately reused for a new signalling connection, the session memory is re-initialized — a2dp_get_connection() in subsys/bluetooth/host/classic/a2dp.c memset()s the whole struct bt_a2dp, and bt_avdtp_l2cap_accept() / bt_avdtp_connect() call k_work_init() on session->_release_work — while that work item is still linked into k_sys_work_q.pending. Zeroing a linked sys_snode_t truncates the list while its tail pointer keeps pointing past the truncation.
The reuse window is deterministic rather than a race. All host RX processing runs on the Bluetooth RX work queue at a cooperative priority (CONFIG_BT_RX_PRIO), while _release_work is submitted to the lower-priority system work queue, so the release handler cannot run until the RX thread blocks. Because l2cap_br_recv() processes every signalling command contained in a single L2CAP signalling PDU back-to-back, a peer can send one PDU holding an L2CAP Disconnection Request for the AVDTP signalling channel followed by a Connection Request for PSM 0x0019; the second command reaches the session-reuse path with the release work provably still queued. The peer must already hold an encrypted AVDTP signalling channel, i.e. be a paired/bonded BR/EDR device.
The result is corruption of the global system work queue's pending list: either the queue silently stops executing any work at all (subsequent submissions are appended to a chain the head can no longer reach, and the orphaned items keep their K_WORK_QUEUED flag forever) or, in the tail variant, the list acquires a self-referencing node that the queue thread spins on. The written values are pointers to other struct k_work objects and are not attacker-controlled, so the impact is a device-wide, reboot-required denial of service rather than a controllable write primitive. Only builds with CONFIG_BT_CLASSIC and CONFIG_BT_A2DP are affected.
The fix is a two-commit unit, and neither commit closes the path on its own. ec822ceed83b adds a pending-release check to bt_avdtp_connect() and bt_avdtp_l2cap_accept(), but the A2DP upper layer's a2dp_get_connection() memset()s the session — including the queued work item's list node and flags — before those checks run, so k_work_busy_get() would observe an already-cleared state. 2bd5ea88c69b closes that by making a2dp_get_connection() itself skip the reuse and return NULL while the release work is still pending, which both callers already handle by rejecting the connection attempt. Apply both commits together; either one alone leaves the device vulnerable. |
| Two RTIO syscall verifiers, in subsys/rtio/rtio_syscalls.c (subsys/rtio/rtio_handlers.c before v4.3.0), did not validate user-supplied pointers before the kernel wrote through them.
z_vrfy_rtio_sqe_cancel() performed no validation at all. It passed the caller's struct rtio_sqe pointer straight to z_impl_rtio_sqe_cancel(), which converts it with CONTAINER_OF to a struct rtio_iodev_sqe, ORs RTIO_SQE_CANCELED (BIT(3)) into its 16-bit flags field in supervisor mode, then loads its next pointer and repeats until that pointer is NULL. The syscall takes no RTIO context argument, so it needs no object grant: any user-mode thread on a CONFIG_USERSPACE build can set that bit in a 16-bit word at an address of its choosing in kernel memory, and the kernel continues the write along whatever pointer chain it finds at the corresponding offset. Setting chosen bits in kernel data structures, such as object permission or thread state fields, can be used to escalate privilege from user mode.
z_vrfy_rtio_sqe_copy_in_get_handles() validated its RTIO context and the input SQE array, but only NULL-checked the handle out-parameter before writing a kernel SQE pointer through it, with no K_SYSCALL_MEMORY_WRITE check. A user thread that holds a granted RTIO context can therefore make the kernel store a pointer value at an arbitrary address.
The fixes validate the handle out-parameter as writable, bound the cancel handle to real submission-pool storage, and require permission on the RTIO context that owns the matching pool. Without that last check a bounded handle still let any user thread cancel submissions belonging to other contexts, silently aborting their I/O. Systems built without CONFIG_USERSPACE or without CONFIG_RTIO are unaffected. |
| The MCUmgr SMP-over-console transport decodes a base64 frame, reads a 16-bit packet length from it, verifies a CRC and then unconditionally strips the trailing CRC with rx_ctxt->nb->len -= 2U; in mcumgr_serial_process_frag() (subsys/mgmt/mcumgr/transport/src/serial_util.c). mcumgr_serial_extract_len() accepted any declared length, including 0 and 1, and a packet declaring length 0 passes the checksum test for free because crc16_itu_t() over zero bytes returns the zero seed. Since net_buf::len is a uint16_t, the subtraction underflows and the buffer is handed to SMP claiming roughly 65 KB of payload while its data area is only CONFIG_MCUMGR_TRANSPORT_NETBUF_SIZE bytes (default 384).
The trigger is a single unauthenticated 7-byte line on the management console — the 0x06 0x09 packet marker followed by the base64 group AAA= and a newline — delivered to any transport built on this helper: CONFIG_MCUMGR_TRANSPORT_UART (smp_uart.c) or CONFIG_MCUMGR_TRANSPORT_SHELL (smp_shell.c), both of which select MCUMGR_TRANSPORT_SERIAL_HAS_SMP_OVER_CONSOLE. No prior session state, fragmentation or credentials are required to trigger the underflow, and the malformed frame is mishandled before any command handler or command-level access control runs. The attacker only needs write access to that console, which on many boards is a USB CDC-ACM port rather than a bare UART header.
With the inflated length, smp_process_request_packet() in subsys/mgmt/mcumgr/smp/src/smp.c loses its bound: cbor_nb_reader_init() gives the CBOR decoder a ~65 KB window into a 384-byte buffer, and each request header's nh_len is checked only against the inflated length. On its own the 7-byte frame re-parses whatever stale bytes the reused pool buffer still holds, typically a replay of the previously received request followed by a parse error, without leaving the buffer. Because the transport is unauthenticated, though, the attacker also controls the frames sent before the trigger, and can stage buffer contents so that a request succeeds with an nh_len larger than the buffer; net_buf_pull(), guarded only by __ASSERT_NO_MSG, then moves the parse cursor out of bounds and the loop reads further headers and CBOR from adjacent memory. The consequence is an out-of-bounds read that can fault the MCUmgr thread (denial of service); memory disclosure is also possible, since the default-enabled os echo handler (CONFIG_MCUMGR_GRP_OS_ECHO) decodes its string inside that window and copies it into its response. There is no integrity gain beyond what the unauthenticated transport already permits.
The fix rejects any declared packet length of two bytes or fewer in mcumgr_serial_extract_len(), so the CRC-strip subtraction can no longer underflow. The identical pattern remains in the test-only loopback transport subsys/mgmt/mcumgr/transport/src/smp_dummy.c (CONFIG_MCUMGR_TRANSPORT_DUMMY), which has no external input path and therefore carries no practical exposure. |
| The native BSD-socket layer recorded a pending asynchronous socket error by type-punning it into struct net_context's void user_data field (ctx->user_data = INT_TO_POINTER(-status) in zsock_accepted_cb(), zsock_received_cb(), zsock_connected_cb() and zsock_close_ctx() in subsys/net/lib/sockets/sockets_inet.c), reading it back with POINTER_TO_INT(). That same field is owned by the network stack for listening TCP contexts: net_tcp_accept() stores the parent context pointer there and the TCP core passes it back to the registered accept callback. A failed accept therefore left a small integer (an errno value) where the stack expected a struct net_context .
When the network interface carrying a listening TCP socket goes down, close_tcp_conn() in subsys/net/ip/tcp.c invokes the accept callback with -ENETDOWN and the context's user_data. In v4.3.0 the callback was not disarmed afterwards, so a second interface-down event forwarded the previously stored errno to zsock_accepted_cb(), which dereferenced it as the parent context and performed several stores through it (sock_set_error()'s read-modify-write of socket_data, k_fifo_cancel_wait(&parent->recv_q)) — the crash described in the fix's commit message. v4.3.1 and v4.4.x carry a later change clearing conn->accept_cb after the error callback (269cb8823d3 on the v4.3 branch, 913fae5169425550f2364655298fceb79b320066 on main), which closes that repeat path; on those releases the poisoned cookie remains reachable only by a narrower race, a handshake completing alongside the interface-down still passing the stale cookie to k_fifo_put(&parent->accept_q, ...), and by getsockopt(SO_ERROR), which reads the field back unconditionally.
On v4.3.0 an application that keeps a listening TCP socket open across repeated link-down events is sufficient to reach the defect; the triggering condition is a network-interface state change, not attacker-supplied packet data, so the practical attacker is one able to force the link down repeatedly (for example an adjacent attacker disrupting a wireless link) or one with local/physical access. Because both the faulting address and the stored data are fixed small constants derived from the errno value, the outcome is a wild-pointer access leading to a kernel fatal error — a denial of service (device crash or reset) rather than an attacker-directed memory corruption.
The fix stores the pending error in a dedicated net_context.sock_error field and converts every producer and consumer to sock_set_error()/sock_get_error(), leaving user_data untouched. As a side effect it also stops getsockopt(SO_ERROR) — which is evaluated unconditionally — from returning the kernel address held in user_data to a userspace application. |
| parse_write_op() in subsys/net/lib/lwm2m/lwm2m_message_handling.c handles inbound CoAP WRITE/CREATE requests that carry a Block1 option. For the first block of a transfer it called init_block_ctx() and then immediately stored the peer-selected block size with block_ctx->ctx.block_size = block_size before inspecting the return code. init_block_ctx() sets the caller's pointer to NULL and returns -ENOMEM when no entry of the static block1_contexts[] pool is free or timed out, so that store dereferences a NULL pointer.
The pool holds CONFIG_LWM2M_NUM_BLOCK1_CONTEXT entries (default 3) and an entry is only reclaimed once its transfer completes, fails, or ages past 30 seconds. A peer that reaches the client's LwM2M socket can therefore start three block-wise writes on three distinct object paths with the CoAP More bit set and leave them incomplete, then send the first block of a fourth write on a new path to reach the unguarded dereference. Reachability is gated only by the connected UDP socket's source-address filter unless CONFIG_LWM2M_DTLS_SUPPORT is enabled — which has no default — so in a NoSec deployment an on-path or address-spoofing attacker needs no credentials; the same sequence is also reachable from a bootstrap or lower-trust server, and can be hit accidentally by a legitimate server running four concurrent block transfers.
The write targets a fixed low address with a value between 0 and 7, so the consequence is a fatal memory fault (BusFault or corrupted low memory leading to a fault) rather than a usable memory-corruption primitive: the device crashes or resets. Confidentiality and integrity are not affected. The fix moves the store below the guard and validates the context pointer itself instead of the return code, so the context is only touched once it is known to be valid. |
| The HTTP/2 server in subsys/net/lib/http/http_server_http2.c clears the connection's stream binding at the start of every frame. handle_http_frame_header() executed client->current_stream = NULL unconditionally, but a CONTINUATION frame does not open a new stream and enter_http_frame_continuation_state() never re-bound one, so the rest of the header block was processed with client->current_stream unset. Two paths then write through the NULL pointer. A request for an unregistered resource, or a method the matched resource does not support, is answered by send_http2_404() or send_http2_405(), and both call send_headers_frame(), which unconditionally sets client->current_stream->headers_sent = true. A POST, PUT or PATCH to a registered dynamic resource reaches handle_http2_dynamic_resource(), which unlike its siblings (handle_http2_static_resource(), dynamic_get_del_opts_req_v2(), dynamic_post_put_req_v2()) had no client->current_stream == NULL guard, and stores through the pointer with client->current_stream->current_detail = dynamic_detail.
Any peer that can reach the server can trigger this without authentication or prior state. A single stray CONTINUATION frame with END_HEADERS set, sent right after the client preface and requesting any unregistered path, is enough, because nothing rejected a CONTINUATION that no HEADERS frame had opened. A legitimate split header block (a HEADERS frame with END_HEADERS clear whose payload ends on a header-field boundary, then a CONTINUATION with END_HEADERS set) reaches the same state. The in-tree test test_http2_post_headers_with_continuation() did not catch this because it splits the header block mid field, which takes the handle_incomplete_http_header() shortcut that keeps the stream bound.
The result is a write to a fixed low address (a small offset into struct http2_stream_ctx), which faults on mainstream targets (flash/vector-table region on Cortex-M, or a trap under MPU/MMU null-pointer detection), and the default fatal-error handler halts the system. Impact is an unauthenticated, remotely triggerable denial of service against any device running the HTTP/2 server: the 404 path needs no particular resource to be registered, and HTTP/2 is built into CONFIG_HTTP_SERVER with cleartext h2c accepted, so TLS does not prevent it. Neither the target address nor the written value is attacker-controlled, so there is no confidentiality or integrity impact. Applications built without CONFIG_HTTP_SERVER are not affected. |
| 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. |
