CVE-2026-72327

The CVE-2026-72327 issue affects the Linux kernel DRM v3d indirect CSD setup. v3d_get_cpu_indirect_csd_params() looks up the indirect buffer object from a userspace-supplied handle and may return NULL; this NULL is stored in info->indirect and later dereferenced by the indirect CSD job, turnin...

CVE-2026-72326

CVE-2026-72326 affects the Linux kernel’s net/sched cake scheduler. CAKE accepts signed overhead values stored in an s16, but the adjusted packet length calculation uses unsigned arithmetic, allowing a negative effective length to wrap to a large value. This can cause rate accounting to depend on...

CVE-2026-72325

Technical details of CVE-2026-72325 are not public in the provided documents. Available descriptions summarize the issue but do not expose affected products, versions, or fixes. Monitor for updates from vendors/security advisories.

CVE-2026-72324

Summary: CVE-2026-72324 relates to the Linux kernel GPIO/mvebu domain handling. The issue arises because irq_alloc_domain_generic_chips() allocates generic chip data that must be freed via irq_domain_remove_generic_chips(), but the existing devres action mvebu_gpio_remove_irq_domain() only called...

CVE-2026-72322

CVE-2026-72322 concerns a race in the Linux kernel IPv6 multicast code (MLD delayed work) that can cause a Use-After-Free during device teardown and MLD query processing. The root cause is a use of in6_dev_hold() to take a reference on inet6_dev while it may already be freed (RCU-protected memory...

CVE-2026-72323

CVE-2026-72323 — Linux kernel igmp UAF vulnerability : A race between device teardown (inetdev_destroy) and IGMP query processing (igmp_rcv) can cause a Use-After-Free in the IGMP timer callback when the timer is re-armed. The in_device memory is freed via RCU after its refcount reaches 0, but ig...

CVE-2026-72321

CVE-2026-72321 affects the Linux kernel’s IGMP timer handling (ipv4: igmp). The root cause is a memory leak when deleting or stopping timers in igmp_mod_timer() and igmp_stop_timer(): refcount_dec(&im->refcnt) can drop the group entry im to zero on an RC reader path, leaving the im object leak...

CVE-2026-72319

The CVE-2026-72319 entry describes a Linux kernel vulnerability in IPVS where inner IP headers in ICMP errors must be kept within skb headroom after removing outer headers (pskb_pull). The root cause involves ensuring the inner headers remain present for functions such as ipv4_update_pmtu(), icmp...

CVE-2026-72320

CVE-2026-72320 concerns the Linux kernel netfilter implementation for nft_lookup. The issue arises in how inverted lookups (NFT_LOOKUP_F_INV) interact with the catchall element in interval sets, potentially causing an incorrect verdict. The documented fix folds the catchall lookup into ext before...

CVE-2026-72318

CVE-2026-72318 affects the Linux kernel CIFS/SMB path handling. The flaw arises in parse_dfs_referrals(), where string offsets in DFS referrals are not validated against the received buffer, allowing crafted referrals to cause pointers to read outside the buffer. This can lead to out-of-bounds re...

CVE-2026-72317

In the Linux kernel SUNRPC TLS connect path, a use-after-free occurs because upper rpc_clnt is not pinned across the delayed TLS connect_worker. The TLS handshake worker xs_tcp_tls_setup_socket() reads saved fields from a freed rpc_clnt when a handshake fails, leading to a refcount underflow. The...

CVE-2026-72316

In the Linux kernel vulnerability CVE-2026-72316, the metadata_open() path used by device-mapper metadata handling returns NULL on kzalloc_obj() failure, but era_ctr() only checks IS_ERR(md). Since IS_ERR(NULL) is false, a NULL is treated as valid and later dereferenced (era->md). The fix is t...

CVE-2026-72314

CVE-2026-72314 in the Linux kernel fixes a bug in regulator_lock_two() where the code compared the return value to -EDEADLOCK instead of -EDEADLK after ww_mutex_lock() returns -EDEADLK on contention. This mismatch could cause the lock to not back off and retry correctly, resulting in only one of ...

CVE-2026-72315

The CVE-2026-72315 entry documents a Linux kernel issue in the CIFS/SMB client where a delayed release of a dentry reference count during unmount after DIO could trigger a busy-dentry warning. A fix was implemented in cifs_kill_sb() by flushing deferredclose_wq, and later the problem in the DIO p...

CVE-2026-72313

CVE-2026-72313 concerns the Linux kernel DRM fb-helper: the fix ensures fbdev output is synchronized only to the vblank of an active CRTC. The change iterates the CRTC list and selects the first active CRTC for vblank synchronization, instead of assuming a fixed CRTC. This addresses a warning and...

CVE-2026-72311

CVE-2026-72311 affects the Linux kernel DRM/XE path. The issue occurs in xe_vm_madvise_ioctl(), which allocates madvise_range.vmas in get_vmas(). When get_vmas() succeeds and later validation fails, control incorrectly jumps to madv_fini, skipping kfree(madvise_range.vmas) and leaking the VMA arr...

CVE-2026-72312

CVE-2026-72312 affects the Linux kernel component octeontx2-af and fixes an issue where VF bringup could erroneously modify the PF’s promiscuous/allmulti MCAM rules. The root cause was that mbox handling of nix_set_rx_mode (and related APIs) acted only on the PF, not the VF, potentially clearing ...

CVE-2026-72310

CVE-2026-72310 : In the Linux kernel SMB client, the passthrough ioctl bounds check for smb2_ioctl_query_info() uses 32-bit fields for OutputOffset and input_buffer_length. The code adds these values directly, allowing 32-bit wrap-around, which can let a malicious server craft a large OutputOffse...

CVE-2026-72308

CVE-2026-72308 affects the Linux kernel’s mlxsw port code, where mlxsw_sp_port_lag_join() can leak a refcount if mlxsw_sp_port_lag_index_get() fails. The issue arises when an error path bypasses the standard cleanup that releases the lag reference with mlxsw_sp_lag_put(), leading to a single miss...

CVE-2026-72309

The CVE-2026-72309 vulnerability affects the Linux kernel’s tracing/remotes path. It fixes a leak in trace_remote_alloc_buffer() error path: if page allocation fails, desc->nr_cpus is not yet incremented for the current CPU, which can cause a half-allocated rb_desc to remain unreleased in trac...