CVE-2022-50396
Publication date:
18/09/2025
In the Linux kernel, the following vulnerability has been resolved:<br />
<br />
net: sched: fix memory leak in tcindex_set_parms<br />
<br />
Syzkaller reports a memory leak as follows:<br />
====================================<br />
BUG: memory leak<br />
unreferenced object 0xffff88810c287f00 (size 256):<br />
comm "syz-executor105", pid 3600, jiffies 4294943292 (age 12.990s)<br />
hex dump (first 32 bytes):<br />
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................<br />
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................<br />
backtrace:<br />
[] kmalloc_trace+0x20/0x90 mm/slab_common.c:1046<br />
[] kmalloc include/linux/slab.h:576 [inline]<br />
[] kmalloc_array include/linux/slab.h:627 [inline]<br />
[] kcalloc include/linux/slab.h:659 [inline]<br />
[] tcf_exts_init include/net/pkt_cls.h:250 [inline]<br />
[] tcindex_set_parms+0xa7/0xbe0 net/sched/cls_tcindex.c:342<br />
[] tcindex_change+0xdf/0x120 net/sched/cls_tcindex.c:553<br />
[] tc_new_tfilter+0x4f2/0x1100 net/sched/cls_api.c:2147<br />
[] rtnetlink_rcv_msg+0x4dc/0x5d0 net/core/rtnetlink.c:6082<br />
[] netlink_rcv_skb+0x87/0x1d0 net/netlink/af_netlink.c:2540<br />
[] netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]<br />
[] netlink_unicast+0x397/0x4c0 net/netlink/af_netlink.c:1345<br />
[] netlink_sendmsg+0x396/0x710 net/netlink/af_netlink.c:1921<br />
[] sock_sendmsg_nosec net/socket.c:714 [inline]<br />
[] sock_sendmsg+0x56/0x80 net/socket.c:734<br />
[] ____sys_sendmsg+0x178/0x410 net/socket.c:2482<br />
[] ___sys_sendmsg+0xa8/0x110 net/socket.c:2536<br />
[] __sys_sendmmsg+0x105/0x330 net/socket.c:2622<br />
[] __do_sys_sendmmsg net/socket.c:2651 [inline]<br />
[] __se_sys_sendmmsg net/socket.c:2648 [inline]<br />
[] __x64_sys_sendmmsg+0x24/0x30 net/socket.c:2648<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+0x63/0xcd<br />
====================================<br />
<br />
Kernel uses tcindex_change() to change an existing<br />
filter properties.<br />
<br />
Yet the problem is that, during the process of changing,<br />
if `old_r` is retrieved from `p->perfect`, then<br />
kernel uses tcindex_alloc_perfect_hash() to newly<br />
allocate filter results, uses tcindex_filter_result_init()<br />
to clear the old filter result, without destroying<br />
its tcf_exts structure, which triggers the above memory leak.<br />
<br />
To be more specific, there are only two source for the `old_r`,<br />
according to the tcindex_lookup(). `old_r` is retrieved from<br />
`p->perfect`, or `old_r` is retrieved from `p->h`.<br />
<br />
* If `old_r` is retrieved from `p->perfect`, kernel uses<br />
tcindex_alloc_perfect_hash() to newly allocate the<br />
filter results. Then `r` is assigned with `cp->perfect + handle`,<br />
which is newly allocated. So condition `old_r && old_r != r` is<br />
true in this situation, and kernel uses tcindex_filter_result_init()<br />
to clear the old filter result, without destroying<br />
its tcf_exts structure<br />
<br />
* If `old_r` is retrieved from `p->h`, then `p->perfect` is NULL<br />
according to the tcindex_lookup(). Considering that `cp->h`<br />
is directly copied from `p->h` and `p->perfect` is NULL,<br />
`r` is assigned with `tcindex_lookup(cp, handle)`, whose value<br />
should be the same as `old_r`, so condition `old_r && old_r != r`<br />
is false in this situation, kernel ignores using<br />
tcindex_filter_result_init() to clear the old filter result.<br />
<br />
So only when `old_r` is retrieved from `p->perfect` does kernel use<br />
tcindex_filter_result_init() to clear the old filter result, which<br />
triggers the above memory leak.<br />
<br />
Considering that there already exists a tc_filter_wq workqueue<br />
to destroy the old tcindex_d<br />
---truncated---
Severity CVSS v4.0: Pending analysis
Last modification:
14/01/2026
