Vulnerabilities

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> hpfs: fix a crash if hpfs_map_dnode_bitmap fails<br /> <br /> If hpfs_map_dnode_bitmap fails, the code would call hpfs_brelse4 on<br /> uninitialized quad buffer head, causing a crash.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> mm/vmalloc: do not trigger BUG() on BH disabled context<br /> <br /> __get_vm_area_node() currently triggers a BUG() if in_interrupt() returns<br /> true. However, in_interrupt() also reports true when BH are disabled.<br /> <br /> The bridge code can call rhashtable_lookup_insert_fast() with bottom<br /> halves disabled:<br /> <br /> __vlan_add()<br /> -&gt; br_fdb_add_local()<br /> spin_lock_bh(&amp;br-&gt;hash_lock); fdb_add_local()<br /> -&gt; fdb_create()<br /> -&gt; rhashtable_lookup_insert_fast()<br /> -&gt; kvmalloc()<br /> -&gt; vmalloc()<br /> -&gt; __get_vm_area_node()<br /> -&gt; BUG_ON(in_interrupt())<br /> spin_unlock_bh(&amp;br-&gt;hash_lock)<br /> <br /> this triggers the BUG() despite the caller not being in NMI or<br /> hard IRQ context.<br /> <br /> Replace the in_interrupt() check with in_nmi() || in_hardirq().

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> USB: serial: cypress_m8: fix memory corruption with small endpoint<br /> <br /> Make sure that the interrupt-out endpoint max packet size is at least<br /> eight bytes to avoid user-controlled slab corruption or NULL-pointer<br /> dereference should a malicious device report a smaller size.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> USB: serial: safe_serial: fix memory corruption with small endpoint<br /> <br /> Make sure that the bulk-out buffer size is at least eight bytes to avoid<br /> user-controlled slab corruption in "safe" mode should a malicious device<br /> report a smaller size.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> usb: typec: ucsi: validate connector number in ucsi_connector_change()<br /> <br /> The connector number in a UCSI CCI notification is a 7-bit field<br /> supplied by the PPM. ucsi_connector_change() uses it to index the<br /> ucsi-&gt;connector[] array without checking it against the number of<br /> connectors the PPM reported at init time, so a buggy or malicious PPM<br /> (EC firmware, or an I2C-attached UCSI controller on the ccg / stm32g0 /<br /> glink transports) can drive schedule_work() on memory past the end of<br /> the array.<br /> <br /> Reject connector numbers that are zero or exceed cap.num_connectors<br /> before dereferencing the array.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> usb: typec: tcpm/tcpci_maxim: validate header NDO against RX_BYTE_CNT<br /> <br /> A broken/malicious port can transmit a CRC-valid frame whose header<br /> advertises up to seven data objects but whose body carries fewer than<br /> that. Check for this, and rightfully reject the message, instead of<br /> reading from uninitialized stack memory.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> usb: typec: wcove: don&amp;#39;t write past struct pd_message in wcove_read_rx_buffer()<br /> <br /> wcove_read_rx_buffer() copies the PD RX FIFO into the caller&amp;#39;s<br /> struct pd_message with<br /> <br /> for (i = 0; i regmap, USBC_RX_DATA + i, msg + i);<br /> <br /> which has two problems:<br /> <br /> USBC_RXINFO_RXBYTES() is a 5-bit field (max 31) while struct pd_message<br /> is 30 bytes (__le16 header + __le32 payload[PD_MAX_PAYLOAD], packed).<br /> The byte count latched in RXINFO is the number of bytes the port partner<br /> put on the wire, so a malicious partner that transmits a 31-byte frame<br /> can drive the loop one byte past the destination if the WCOVE BMC<br /> receiver does not enforce the PD object-count limit in hardware. The<br /> existing FIXME flagged this as unverified.<br /> <br /> Independently, regmap_read() takes an unsigned int * and stores a full<br /> unsigned int at the destination. Passing the byte pointer msg + i means<br /> each iteration writes four bytes; the high three are zero (val_bits is<br /> 8) and are normally overwritten by the next iteration, but the final<br /> iteration&amp;#39;s high bytes are not. With RXBYTES == 30 the i == 29 iteration<br /> already writes three zero bytes past msg, which sits on the IRQ thread&amp;#39;s<br /> stack in wcove_typec_irq().<br /> <br /> Clamp the loop to sizeof(struct pd_message) and read each register into<br /> a local before storing only its low byte, so the copy can never exceed<br /> the destination regardless of what RXINFO reports.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> Bluetooth: ISO: serialize iso_sock_clear_timer with socket lock<br /> <br /> iso_sock_close() calls iso_sock_clear_timer() before acquiring<br /> lock_sock(sk).<br /> <br /> iso_sock_clear_timer() reads iso_pi(sk)-&gt;conn twice without the<br /> socket lock held:<br /> <br /> if (!iso_pi(sk)-&gt;conn)<br /> return;<br /> cancel_delayed_work(&amp;iso_pi(sk)-&gt;conn-&gt;timeout_work);<br /> <br /> Concurrently, iso_conn_del() executes under lock_sock(sk) and calls<br /> iso_chan_del(), which sets iso_pi(sk)-&gt;conn to NULL and may result in<br /> the final reference to the connection being dropped:<br /> <br /> CPU0 CPU1<br /> ---- ----<br /> iso_sock_clear_timer()<br /> if (conn != NULL) ... lock_sock(sk)<br /> iso_chan_del()<br /> iso_pi(sk)-&gt;conn = NULL<br /> cancel_delayed_work(conn) /* NULL deref or UAF */<br /> <br /> iso_pi(sk)-&gt;conn is not stable across the unlock window, causing a<br /> NULL pointer dereference or use-after-free.<br /> <br /> Serialize iso_sock_clear_timer() with the socket lock by moving it<br /> inside lock_sock()/release_sock(), matching the pattern used in<br /> iso_conn_del() and all other call sites.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> Bluetooth: ISO: fix UAF in iso_recv_frame<br /> <br /> iso_recv_frame reads conn-&gt;sk under iso_conn_lock but releases the lock<br /> before using sk, with no reference held. A concurrent iso_sock_kill()<br /> can free sk in that window, causing use-after-free on sk-&gt;sk_state and<br /> sock_queue_rcv_skb().<br /> <br /> Fix by replacing the bare pointer read with iso_sock_hold(conn), which<br /> calls sock_hold() while the spinlock is held, atomically elevating the<br /> refcount before the lock drops. Add a drop_put label so sock_put() is<br /> called on all exit paths where the hold succeeded.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> Bluetooth: HIDP: fix missing length checks in hidp_input_report()<br /> <br /> hidp_input_report() reads keyboard and mouse payload data from an skb<br /> without first verifying that skb-&gt;len contains enough data.<br /> <br /> hidp_recv_intr_frame() pulls the 1-byte HIDP header before dispatching<br /> to hidp_input_report(). If a paired device sends a truncated packet,<br /> the handler reads beyond the valid skb data, resulting in an<br /> out-of-bounds read of skb data. The OOB bytes may be interpreted as<br /> phantom key presses or spurious mouse movement.<br /> <br /> Replace the open-coded length tracking and pointer arithmetic with<br /> skb_pull_data() calls. skb_pull_data() returns NULL if the requested<br /> bytes are not present, eliminating the need for a manual size variable<br /> and the separate skb-&gt;len guard.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> Bluetooth: L2CAP: fix chan ref leak in l2cap_chan_timeout() on !conn<br /> <br /> __set_chan_timer() takes a l2cap_chan reference via l2cap_chan_hold()<br /> before scheduling the delayed work. The normal path in<br /> l2cap_chan_timeout() drops this reference with l2cap_chan_put() at the<br /> end, but the early return when chan-&gt;conn is NULL skips the put,<br /> leaking the reference.<br /> <br /> Add the missing l2cap_chan_put() before the early return.

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> auxdisplay: line-display: fix OOB read on zero-length message_store()<br /> <br /> linedisp_display() unconditionally reads msg[count - 1] before<br /> checking whether count is zero, so a write of zero bytes to the<br /> message sysfs attribute hits msg[-1]:<br /> <br /> write(fd, "", 0);<br /> <br /> -&gt; message_store(..., buf, count=0)<br /> -&gt; linedisp_display(linedisp, buf, count=0)<br /> -&gt; msg[count - 1] == &amp;#39;\n&amp;#39; ; OOB read<br /> <br /> The kernfs write buffer for that store is a 1-byte allocation<br /> (kernfs_fop_write_iter() does kmalloc(len + 1) with len == 0),<br /> so msg[-1] is a 1-byte read before the slab object. On a<br /> KASAN-enabled kernel this trips an out-of-bounds report and<br /> panics; on stock kernels it silently reads adjacent slab data<br /> and, if that byte happens to be &amp;#39;\n&amp;#39;, the following count--<br /> wraps ssize_t 0 to -1 and is then passed to kmemdup_nul().<br /> <br /> linedisp_display() is reached from the message_store() sysfs<br /> callback (drivers/auxdisplay/line-display.c message attribute,<br /> mode 0644) and from the in-tree initial-message setup with<br /> count == -1, so the OOB path is only userspace-triggerable via<br /> zero-byte writes; vfs_write() does not short-circuit on<br /> count == 0 and kernfs_fop_write_iter() dispatches the store<br /> callback regardless.<br /> <br /> Guard the trailing-newline trim with a count check. The<br /> existing if (!count) block then takes the clear-display path<br /> unchanged.<br /> <br /> Affects every auxdisplay driver that registers via<br /> linedisp_register() / linedisp_attach(): ht16k33, max6959,<br /> img-ascii-lcd, seg-led-gpio.