CVE-2026-98369
Gravedad CVSS v3.1:
ALTA
Tipo:
No Disponible / Otro tipo
Fecha de publicación:
06/10/2026
Última modificación:
07/10/2026
Descripción
*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br />
<br />
xfrm: add missing rcu_read_lock(), skb_dst_force() and dev_hold() for xfrm_trans_reinject()<br />
<br />
syzbot reported a suspicious RCU usage warning in ip6_pkt_drop():<br />
<br />
WARNING: suspicious RCU usage in ip6_pkt_drop<br />
include/net/addrconf.h:389 suspicious rcu_dereference_check() usage!<br />
<br />
Call Trace:<br />
__in6_dev_get_safely include/net/addrconf.h:389 [inline]<br />
ip6_pkt_drop+0x596/0x610 net/ipv6/route.c:4620<br />
ip6_pkt_discard+0x1c/0x30 net/ipv6/route.c:4651<br />
xfrm_trans_reinject+0x324/0x630 net/xfrm/xfrm_input.c:806<br />
process_one_work kernel/workqueue.c:3322 [inline]<br />
process_scheduled_works+0xa8e/0x14e0 kernel/workqueue.c:3405<br />
worker_thread+0xa47/0xfb0 kernel/workqueue.c:3486<br />
<br />
When commit 4f4920669d21 ("xfrm: Reinject transport-mode packets through<br />
workqueue") converted xfrm_trans_reinject from a tasklet to a workqueue,<br />
the reinjection loop ceased running in softirq context. Workqueue workers<br />
run in process context where local_bh_disable() does not enter an RCU<br />
read-side critical section under CONFIG_PREEMPT_RCU.<br />
<br />
Because finish callbacks (such as ip6_rcv_finish) expect to run under an<br />
RCU read lock (performing route lookups, l3mdev lookups, and accessing<br />
RCU-protected data structures), invoking them in workqueue context without<br />
rcu_read_lock() triggers RCU lockdep warnings.<br />
<br />
Furthermore, packets queued to the workqueue via xfrm_trans_queue_net()<br />
may carry non-refcounted (noref) dst entries (e.g. from ip_route_input_noref).<br />
Additionally, on netdevice unregistration, dst_dev_put() replaces dst->dev<br />
with blackhole_netdev, so dst entries do not keep skb->dev alive while<br />
queued in the workqueue.<br />
<br />
Fix these issues by:<br />
1. Calling skb_dst_force(skb) in xfrm_trans_queue_net() while still in the<br />
caller&#39;s RCU section to ensure dst is reference-counted before queuing.<br />
2. Holding a reference on skb->dev via dev_hold()/dev_put() across workqueue<br />
deferral so skb->dev remains valid during finish() callback processing.<br />
3. Acquiring rcu_read_lock() around the finish callback invocation loop in<br />
xfrm_trans_reinject().
Impacto
Puntuación base 3.x
7.80
Gravedad 3.x
ALTA
Referencias a soluciones, herramientas e información
- https://git.kernel.org/stable/c/0cda8273265d30cac6423834fd7d4acb75f04fdb
- https://git.kernel.org/stable/c/41e47f1664be86c91326f0afe0504a1162d00907
- https://git.kernel.org/stable/c/6601d91a85761f33351c71e04ec0bbd294ca07ce
- https://git.kernel.org/stable/c/664fc0941df7c1918b2cd4de6ee00469ba77d8e4
- https://git.kernel.org/stable/c/68a317b4aec8ca1868a39d69e40f9e29baa4f40a
- https://git.kernel.org/stable/c/6eb3b071be8e260543c604550c54dac66e6b174b
- https://git.kernel.org/stable/c/d2f5082f9e84653fa1a9e8aebaaff23e688f5e19


