CVE-2023-53354

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> skbuff: skb_segment, Call zero copy functions before using skbuff frags<br /> <br /> Commit bf5c25d60861 ("skbuff: in skb_segment, call zerocopy functions<br /> once per nskb") added the call to zero copy functions in skb_segment().<br /> The change introduced a bug in skb_segment() because skb_orphan_frags()<br /> may possibly change the number of fragments or allocate new fragments<br /> altogether leaving nrfrags and frag to point to the old values. This can<br /> cause a panic with stacktrace like the one below.<br /> <br /> [ 193.894380] BUG: kernel NULL pointer dereference, address: 00000000000000bc<br /> [ 193.895273] CPU: 13 PID: 18164 Comm: vh-net-17428 Kdump: loaded Tainted: G O 5.15.123+ #26<br /> [ 193.903919] RIP: 0010:skb_segment+0xb0e/0x12f0<br /> [ 194.021892] Call Trace:<br /> [ 194.027422] <br /> [ 194.072861] tcp_gso_segment+0x107/0x540<br /> [ 194.082031] inet_gso_segment+0x15c/0x3d0<br /> [ 194.090783] skb_mac_gso_segment+0x9f/0x110<br /> [ 194.095016] __skb_gso_segment+0xc1/0x190<br /> [ 194.103131] netem_enqueue+0x290/0xb10 [sch_netem]<br /> [ 194.107071] dev_qdisc_enqueue+0x16/0x70<br /> [ 194.110884] __dev_queue_xmit+0x63b/0xb30<br /> [ 194.121670] bond_start_xmit+0x159/0x380 [bonding]<br /> [ 194.128506] dev_hard_start_xmit+0xc3/0x1e0<br /> [ 194.131787] __dev_queue_xmit+0x8a0/0xb30<br /> [ 194.138225] macvlan_start_xmit+0x4f/0x100 [macvlan]<br /> [ 194.141477] dev_hard_start_xmit+0xc3/0x1e0<br /> [ 194.144622] sch_direct_xmit+0xe3/0x280<br /> [ 194.147748] __dev_queue_xmit+0x54a/0xb30<br /> [ 194.154131] tap_get_user+0x2a8/0x9c0 [tap]<br /> [ 194.157358] tap_sendmsg+0x52/0x8e0 [tap]<br /> [ 194.167049] handle_tx_zerocopy+0x14e/0x4c0 [vhost_net]<br /> [ 194.173631] handle_tx+0xcd/0xe0 [vhost_net]<br /> [ 194.176959] vhost_worker+0x76/0xb0 [vhost]<br /> [ 194.183667] kthread+0x118/0x140<br /> [ 194.190358] ret_from_fork+0x1f/0x30<br /> [ 194.193670] <br /> <br /> In this case calling skb_orphan_frags() updated nr_frags leaving nrfrags<br /> local variable in skb_segment() stale. This resulted in the code hitting<br /> i &gt;= nrfrags prematurely and trying to move to next frag_skb using<br /> list_skb pointer, which was NULL, and caused kernel panic. Move the call<br /> to zero copy functions before using frags and nr_frags.