CVE-2024-26923

En el kernel de Linux, se resolvió la siguiente vulnerabilidad: af_unix: corrige la ejecución del recolector de basura contra connect() El recolector de basura no tiene en cuenta el riesgo de que el embrión quede en cola durante la recolección de basura. Si dicho embrión tiene un par que porta SCM_RIGHTS, dos pases consecutivos de scan_children() pueden ver un conjunto diferente de niños. Lo que lleva a un recuento en vuelo elevado incorrectamente y luego a un puntero colgante dentro de gc_inflight_list. los sockets son AF_UNIX/SOCK_STREAM S es un socket no conectado L es un socket de escucha en vuelo vinculado a addr, no en fdtable El fd de V se pasará a través de sendmsg(), se aumenta el recuento en vuelo connect(S, addr) sendmsg(S, [ V]); cerrar(V) __unix_gc() ---------- ------------------------- -- --------- NS = unix_create1() skb1 = sock_wmalloc(NS) L = unix_find_other(addr) unix_state_lock(L) unix_peer(S) = NS // V count=1 en vuelo=0 NS = unix_peer(S ) skb2 = sock_alloc() skb_queue_tail(NS, skb2[V]) // V se convirtió en vuelo // V recuento=2 en vuelo=1 close(V) // V recuento=1 en vuelo=1 // Condición candidata de GC cumplida para u en gc_inflight_list: if (total_refs == inflight_refs) agregue u a gc_candidates // gc_candidates={L, V} para u en gc_candidates: scan_children(u, dec_inflight) // el embrión (skb1) aún no era // accesible desde L , por lo que V's // en vuelo permanece sin cambios __skb_queue_tail(L, skb1) unix_state_unlock(L) para u en gc_candidates: if (u.inflight) scan_children(u, inc_inflight_move_tail) // V count=1 inflight=2 (!) Si hay un socket de escucha candidato a GC, bloquear/desbloquear su estado. Esto hace que GC espere hasta el final de cualquier conexión () en curso a ese socket. Después de girar la cerradura, un embrión posiblemente cargado de SCM ya está en cola. Y si viene otro embrión, no es posible que porte SCM_RIGHTS. En este punto, unix_inflight() no puede ocurrir porque unix_gc_lock ya está en uso. El gráfico a bordo no se ve afectado.

A flaw was found in the Linux kernel, where the management of inter-process communication uses AF_UNIX sockets. The issue arises from a race condition where a partially initialized socket with specific permissions carrying SCM_RIGHTS is improperly handled during garbage collection. This situation leads to an incorrect count of active sockets, potentially causing resources to remain unaccounted for and never released.

In the Linux kernel, the following vulnerability has been resolved: af_unix: Fix garbage collector racing against connect() Garbage collector does not take into account the risk of embryo getting enqueued during the garbage collection. If such embryo has a peer that carries SCM_RIGHTS, two consecutive passes of scan_children() may see a different set of children. Leading to an incorrectly elevated inflight count, and then a dangling pointer within the gc_inflight_list. sockets are AF_UNIX/SOCK_STREAM S is an unconnected socket L is a listening in-flight socket bound to addr, not in fdtable V's fd will be passed via sendmsg(), gets inflight count bumped connect(S, addr) sendmsg(S, [V]); close(V) __unix_gc() ---------------- ------------------------- ----------- NS = unix_create1() skb1 = sock_wmalloc(NS) L = unix_find_other(addr) unix_state_lock(L) unix_peer(S) = NS // V count=1 inflight=0 NS = unix_peer(S) skb2 = sock_alloc() skb_queue_tail(NS, skb2[V]) // V became in-flight // V count=2 inflight=1 close(V) // V count=1 inflight=1 // GC candidate condition met for u in gc_inflight_list: if (total_refs == inflight_refs) add u to gc_candidates // gc_candidates={L, V} for u in gc_candidates: scan_children(u, dec_inflight) // embryo (skb1) was not // reachable from L yet, so V's // inflight remains unchanged __skb_queue_tail(L, skb1) unix_state_unlock(L) for u in gc_candidates: if (u.inflight) scan_children(u, inc_inflight_move_tail) // V count=1 inflight=2 (!) If there is a GC-candidate listening socket, lock/unlock its state. This makes GC wait until the end of any ongoing connect() to that socket. After flipping the lock, a possibly SCM-laden embryo is already enqueued. And if there is another embryo coming, it can not possibly carry SCM_RIGHTS. At this point, unix_inflight() can not happen because unix_gc_lock is already taken. Inflight graph remains unaffected.

This update for the Linux Kernel 5.14.21-150400_24_111 fixes several issues. The following security issues were fixed. Fixed int overflow for stack access size. Sch_cake: do not call cake_destroy from cake_init. Fixed potential UAF in is_valid_oplock_break. Smb: client: fix use-after-free bug in cifs_debug_data_proc_show. Fixed potential UAF in smb2_is_network_name_deleted. Fixed potential UAF in cifs_stats_proc_show. Fixed potential UAF in smb2_is_valid_lease_break. Intermittent nfs mount failures. Fixed potential UAF in cifs_signal_cifsd_for_reconnect. Gpiolib: cdev: Fix use after free in lineinfo_changed_notify. Net: do not leave a dangling sk pointer, when socket creation fails hfsplus: fix uninit-value in copy_name. Fs/9p: only translate RWX permissions for plain 9P2000. Hsr: Prevent use after free in prp_create_tagged_frame. Fixed a general protection fault in i915_perf_open_ioctl. Set gtt bound flag in amdgpu_ttm_gart_bind. Fixed use-after-free bugs caused by sco_sock_timeout. Drm/client: Fully protect modes with dev->mode_config.mutex. Fixed false-positive lockdep splat for spin_lock in __unix_gc. Fixed double free of the ha->vp_map pointer. Fixed underflow in parse_server_interfaces. Fixed Integer Overflow or Wraparound vulnerability in x86 and ARM md, raid, raid5 modules. Fixed use-after-free in ip6_route_mpath_notify. Fixed memory corruption in wifi/iwlwifi. Fixed an out-of-bound bug in ipvlan caused by unset skb->mac_header. Fixed SDMA off-by-one error in _pad_sdma_tx_descs. Fixed a race condition in nfc_llcp_sock_get and nfc_llcp_sock_get_sn. Fixed a race condition in the GSM 0710 tty multiplexor via the GSMIOC_SETCONF ioctl that could lead to local privilege escalation.