[PATCH net v2 2/3] bpf: don't redirect decrypted skbs from egress to ingress

From: Todd Freed

Date: Tue Nov 17 2026 - 14:28:12 EST


On transmit, skb->decrypted means that the device must encrypt the
packet, e.g. a packet of a kTLS or PSP device offload socket.

On receive, skb->decrypted means that the device decrypted the packet,
and kTLS device offload trusts it as such (tls_device_decrypted()).

bpf_redirect() and bpf_clone_redirect() with BPF_F_INGRESS send the
packet to a device's receive path with the bit still set, where it claims
to have been decrypted by the device, although it was never encrypted.

Drop decrypted skbs which bpf redirects from egress to ingress, as mirred
does since the previous patch.

__bpf_redirect() tells egress from ingress by tc_at_ingress, which is
only kept up to date in tc. It is also reached from LWT xmit programs
(ip route ... encap bpf xmit), which run in ip_finish_output2(), before
__dev_queue_xmit() clears the bit. A forwarded packet can still have it
set there by the tc ingress hook of the device it came in on, and would
pass the check: whether the packet is dropped would depend on unrelated
tc configuration of its input device.

Clear the bit before running an LWT xmit program, as __dev_queue_xmit()
would right after. This also gives the program's other uses of the bit,
in the redirect helpers and in its __sk_buff->tstamp access, the egress
view, which matches where it runs.

Sashiko found the LWT case when reviewing this patch as patch 7 of v1
of the net-next series "psp: introduce socket-less associations", which
it was split out of.

Fixes: 414776621d10 ("net/tls: prevent skb_orphan() from leaking TLS plain text with offload")
Reported-by: Sashiko <sashiko-bot@xxxxxxxxxx>
Closes: https://lore.kernel.org/netdev/179471270091.2199510.4320994626979041884@xxxxxxxxxx
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Todd Freed <toddfreed@xxxxxxxx>
---
net/core/filter.c | 10 ++++++++++
net/core/lwt_bpf.c | 8 ++++++++
2 files changed, 18 insertions(+)

diff --git a/net/core/filter.c b/net/core/filter.c
index 946b07a17fce..defbef3a9f75 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -2226,6 +2226,16 @@ static int __bpf_redirect_common(struct sk_buff *skb, struct net_device *dev,
static int __bpf_redirect(struct sk_buff *skb, struct net_device *dev,
u32 flags)
{
+ /*
+ * On transmit, decrypted asks the device to encrypt the packet. Sent
+ * to ingress, the packet would read as decrypted by the device.
+ */
+ if (unlikely((flags & BPF_F_INGRESS) && !skb_at_tc_ingress(skb) &&
+ skb_is_decrypted(skb))) {
+ kfree_skb(skb);
+ return -EINVAL;
+ }
+
if (dev_is_mac_header_xmit(dev))
return __bpf_redirect_common(skb, dev, flags);
else
diff --git a/net/core/lwt_bpf.c b/net/core/lwt_bpf.c
index 72c86ff2a847..4861fe2b4959 100644
--- a/net/core/lwt_bpf.c
+++ b/net/core/lwt_bpf.c
@@ -13,6 +13,7 @@
#include <net/gre.h>
#include <net/ip.h>
#include <net/ip6_route.h>
+#include <net/tcx.h>

struct bpf_lwt_prog {
struct bpf_prog *prog;
@@ -327,6 +328,13 @@ static int bpf_xmit(struct sk_buff *skb)
__be16 proto = skb->protocol;
int ret;

+ /* The packet is on egress, but a forwarded one can still have
+ * tc_at_ingress set by the tc ingress hook of the device it
+ * came in on, until __dev_queue_xmit() clears it. Clear it
+ * now, so that the redirect helpers see the direction of this
+ * hook.
+ */
+ tcx_set_ingress(skb, false);
ret = run_lwt_bpf(skb, &bpf->xmit, dst, CAN_REDIRECT);
switch (ret) {
case BPF_OK:
--
2.53.1-Meta