CVE-2022-49979

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> net: fix refcount bug in sk_psock_get (2)<br /> <br /> Syzkaller reports refcount bug as follows:<br /> ------------[ cut here ]------------<br /> refcount_t: saturated; leaking memory.<br /> WARNING: CPU: 1 PID: 3605 at lib/refcount.c:19 refcount_warn_saturate+0xf4/0x1e0 lib/refcount.c:19<br /> Modules linked in:<br /> CPU: 1 PID: 3605 Comm: syz-executor208 Not tainted 5.18.0-syzkaller-03023-g7e062cda7d90 #0<br /> <br /> __refcount_add_not_zero include/linux/refcount.h:163 [inline]<br /> __refcount_inc_not_zero include/linux/refcount.h:227 [inline]<br /> refcount_inc_not_zero include/linux/refcount.h:245 [inline]<br /> sk_psock_get+0x3bc/0x410 include/linux/skmsg.h:439<br /> tls_data_ready+0x6d/0x1b0 net/tls/tls_sw.c:2091<br /> tcp_data_ready+0x106/0x520 net/ipv4/tcp_input.c:4983<br /> tcp_data_queue+0x25f2/0x4c90 net/ipv4/tcp_input.c:5057<br /> tcp_rcv_state_process+0x1774/0x4e80 net/ipv4/tcp_input.c:6659<br /> tcp_v4_do_rcv+0x339/0x980 net/ipv4/tcp_ipv4.c:1682<br /> sk_backlog_rcv include/net/sock.h:1061 [inline]<br /> __release_sock+0x134/0x3b0 net/core/sock.c:2849<br /> release_sock+0x54/0x1b0 net/core/sock.c:3404<br /> inet_shutdown+0x1e0/0x430 net/ipv4/af_inet.c:909<br /> __sys_shutdown_sock net/socket.c:2331 [inline]<br /> __sys_shutdown_sock net/socket.c:2325 [inline]<br /> __sys_shutdown+0xf1/0x1b0 net/socket.c:2343<br /> __do_sys_shutdown net/socket.c:2351 [inline]<br /> __se_sys_shutdown net/socket.c:2349 [inline]<br /> __x64_sys_shutdown+0x50/0x70 net/socket.c:2349<br /> do_syscall_x64 arch/x86/entry/common.c:50 [inline]<br /> do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80<br /> entry_SYSCALL_64_after_hwframe+0x46/0xb0<br /> <br /> <br /> During SMC fallback process in connect syscall, kernel will<br /> replaces TCP with SMC. In order to forward wakeup<br /> smc socket waitqueue after fallback, kernel will sets<br /> clcsk-&gt;sk_user_data to origin smc socket in<br /> smc_fback_replace_callbacks().<br /> <br /> Later, in shutdown syscall, kernel will calls<br /> sk_psock_get(), which treats the clcsk-&gt;sk_user_data<br /> as psock type, triggering the refcnt warning.<br /> <br /> So, the root cause is that smc and psock, both will use<br /> sk_user_data field. So they will mismatch this field<br /> easily.<br /> <br /> This patch solves it by using another bit(defined as<br /> SK_USER_DATA_PSOCK) in PTRMASK, to mark whether<br /> sk_user_data points to a psock object or not.<br /> This patch depends on a PTRMASK introduced in commit f1ff5ce2cd5e<br /> ("net, sk_msg: Clear sk_user_data pointer on clone if tagged").<br /> <br /> For there will possibly be more flags in the sk_user_data field,<br /> this patch also refactor sk_user_data flags code to be more generic<br /> to improve its maintainability.