Threats affecting pkg:github/zephyr

Zephyr's IPv6 network stack can be prevented from receiving or processing future incoming packets by sending a small number of maliciously fragmented IPv6 packets. When such a packet is handled by the fragment-header processing path, the associated RX network packet buffer (allocated from a memory slab) is not released back to the pool. Repeating the malicious packet exhausts all RX buffer slots, after which the device can no longer obtain RX buffers and stops receiving traffic, resulting in a denial of service.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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.

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). The observed effects are an AddressSanitizer abort and, without ASan, heap corruption / fatal error.

A remote, unauthenticated attacker can trigger memory corruption in Zephyr's HTTP server WebSocket upgrade path by sending a crafted Sec-WebSocket-Key header. The HTTP/1 header parser copies the header into a fixed-size buffer using a bounded copy that does not guarantee NUL termination when the input length reaches the buffer size. During upgrade handling the buffer is copied to a local stack buffer and passed to strlen(); if no NUL exists in-bounds, strlen() reads beyond the stack buffer and subsequent concatenation with the WebSocket magic string can write out of bounds. This leads to out-of-bounds read and write on stack memory, resulting in crash (denial of service) and potentially code execution. The path is reachable when CONFIG_HTTP_SERVER_WEBSOCKET is enabled.

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. struct net_sockaddr is an opaque type, so 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 out-of-bounds write can lead to a crash and denial of service, and potentially to arbitrary code execution.

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 (reported_len) and computes reported_len - 3 without checking that reported_len >= 3. When reported_len is less than 3, the subtraction is performed in signed int arithmetic and yields a negative value that bypasses the length guard and is then implicitly converted to a very large size_t when passed to net_buf_simple_pull_mem(). In builds without assertions, this wraps the buffer length and advances the data pointer far out of bounds, so 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.

The SocketCAN implementation validates the length of a user-provided buffer containing a socketcan_frame object using only a NET_ASSERT statement in zcan_sendto_ctx() before dereferencing it in socketcan_to_can_frame(). In production builds where assertions are disabled, a userspace application that controls the length passed to a sendto syscall can supply an incomplete or truncated frame, causing socketcan_to_can_frame() to dereference fields beyond the end of the buffer. This results in an out-of-bounds read that can cause denial-of-service crashes or, because the parsed frame contents are transmitted on the network, leak adjacent memory.

A bitwise shift vulnerability in Zephyr's PTP subsystem allows a remote attacker to cause undefined behavior and potential system crashes. An attacker sends a crafted PTP_MSG_MANAGEMENT message to set an unvalidated negative log_announce_interval value in the port's data set. When a subsequent PTP_MSG_ANNOUNCE message is processed, port_timer_set_timeout_random computes a timeout as NSEC_PER_SEC >> -log_seconds; if the attacker-supplied value is sufficiently negative (e.g., -127), the shift amount exceeds the 64-bit integer width, triggering undefined behavior in C. This can cause a system crash via a compiler-generated illegal instruction trap on some architectures, or produce an erroneous zero timeout leading to resource starvation loops or other logical errors.

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.