x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 323 insertions, 0 deletions

diff --git a/net/ipv6/xfrm6_input.c b/net/ipv6/xfrm6_input.c
new file mode 100644
index 000000000..07edef258
--- /dev/null
+++ b/net/ipv6/xfrm6_input.c
@@ -0,0 +1,323 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * xfrm6_input.c: based on net/ipv4/xfrm4_input.c
+ *
+ * Authors:
+ *	Mitsuru KANDA @USAGI
+ *	Kazunori MIYAZAWA @USAGI
+ *	Kunihiro Ishiguro <kunihiro@ipinfusion.com>
+ *	YOSHIFUJI Hideaki @USAGI
+ *		IPv6 support
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter_ipv6.h>
+#include <net/ipv6.h>
+#include <net/xfrm.h>
+#include <net/protocol.h>
+#include <net/gro.h>
+
+int xfrm6_rcv_spi(struct sk_buff *skb, int nexthdr, __be32 spi,
+		  struct ip6_tnl *t)
+{
+	XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
+	XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
+	XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
+	return xfrm_input(skb, nexthdr, spi, 0);
+}
+EXPORT_SYMBOL(xfrm6_rcv_spi);
+
+static int xfrm6_transport_finish2(struct net *net, struct sock *sk,
+				   struct sk_buff *skb)
+{
+	if (xfrm_trans_queue(skb, ip6_rcv_finish)) {
+		kfree_skb(skb);
+		return NET_RX_DROP;
+	}
+
+	return 0;
+}
+
+int xfrm6_transport_finish(struct sk_buff *skb, int async)
+{
+	struct xfrm_offload *xo = xfrm_offload(skb);
+	struct net_device *dev = skb->dev;
+	int nhlen = -skb_network_offset(skb);
+
+	skb_network_header(skb)[IP6CB(skb)->nhoff] =
+		XFRM_MODE_SKB_CB(skb)->protocol;
+
+#ifndef CONFIG_NETFILTER
+	if (!async)
+		return 1;
+#endif
+
+	__skb_push(skb, nhlen);
+	ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
+	skb_postpush_rcsum(skb, skb_network_header(skb), nhlen);
+
+	if (xo && (xo->flags & XFRM_GRO)) {
+		/* The full l2 header needs to be preserved so that re-injecting the packet at l2
+		 * works correctly in the presence of vlan tags.
+		 */
+		skb_mac_header_rebuild_full(skb, xo->orig_mac_len);
+		skb_reset_network_header(skb);
+		skb_reset_transport_header(skb);
+		return 0;
+	}
+
+	NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING,
+		dev_net(dev), NULL, skb, dev, NULL,
+		xfrm6_transport_finish2);
+	return 0;
+}
+
+static int __xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb, bool pull)
+{
+	struct udp_sock *up = udp_sk(sk);
+	struct udphdr *uh;
+	struct ipv6hdr *ip6h;
+	int len;
+	int ip6hlen = sizeof(struct ipv6hdr);
+	__u8 *udpdata;
+	__be32 *udpdata32;
+	u16 encap_type;
+
+	encap_type = READ_ONCE(up->encap_type);
+	/* if this is not encapsulated socket, then just return now */
+	if (!encap_type)
+		return 1;
+
+	/* If this is a paged skb, make sure we pull up
+	 * whatever data we need to look at. */
+	len = skb->len - sizeof(struct udphdr);
+	if (!pskb_may_pull(skb, sizeof(struct udphdr) + min(len, 8)))
+		return 1;
+
+	/* Now we can get the pointers */
+	uh = udp_hdr(skb);
+	udpdata = (__u8 *)uh + sizeof(struct udphdr);
+	udpdata32 = (__be32 *)udpdata;
+
+	switch (encap_type) {
+	default:
+	case UDP_ENCAP_ESPINUDP:
+		/* Check if this is a keepalive packet.  If so, eat it. */
+		if (len == 1 && udpdata[0] == 0xff) {
+			return -EINVAL;
+		} else if (len > sizeof(struct ip_esp_hdr) && udpdata32[0] != 0) {
+			/* ESP Packet without Non-ESP header */
+			len = sizeof(struct udphdr);
+		} else
+			/* Must be an IKE packet.. pass it through */
+			return 1;
+		break;
+	}
+
+	/* At this point we are sure that this is an ESPinUDP packet,
+	 * so we need to remove 'len' bytes from the packet (the UDP
+	 * header and optional ESP marker bytes) and then modify the
+	 * protocol to ESP, and then call into the transform receiver.
+	 */
+	if (skb_unclone(skb, GFP_ATOMIC))
+		return -EINVAL;
+
+	/* Now we can update and verify the packet length... */
+	ip6h = ipv6_hdr(skb);
+	ip6h->payload_len = htons(ntohs(ip6h->payload_len) - len);
+	if (skb->len < ip6hlen + len) {
+		/* packet is too small!?! */
+		return -EINVAL;
+	}
+
+	/* pull the data buffer up to the ESP header and set the
+	 * transport header to point to ESP.  Keep UDP on the stack
+	 * for later.
+	 */
+	if (pull) {
+		__skb_pull(skb, len);
+		skb_reset_transport_header(skb);
+	} else {
+		skb_set_transport_header(skb, len);
+	}
+
+	/* process ESP */
+	return 0;
+}
+
+/* If it's a keepalive packet, then just eat it.
+ * If it's an encapsulated packet, then pass it to the
+ * IPsec xfrm input.
+ * Returns 0 if skb passed to xfrm or was dropped.
+ * Returns >0 if skb should be passed to UDP.
+ * Returns <0 if skb should be resubmitted (-ret is protocol)
+ */
+int xfrm6_udp_encap_rcv(struct sock *sk, struct sk_buff *skb)
+{
+	int ret;
+
+	if (skb->protocol == htons(ETH_P_IP))
+		return xfrm4_udp_encap_rcv(sk, skb);
+
+	ret = __xfrm6_udp_encap_rcv(sk, skb, true);
+	if (!ret)
+		return xfrm6_rcv_encap(skb, IPPROTO_ESP, 0,
+				       udp_sk(sk)->encap_type);
+
+	if (ret < 0) {
+		kfree_skb(skb);
+		return 0;
+	}
+
+	return ret;
+}
+
+struct sk_buff *xfrm6_gro_udp_encap_rcv(struct sock *sk, struct list_head *head,
+					struct sk_buff *skb)
+{
+	int offset = skb_gro_offset(skb);
+	const struct net_offload *ops;
+	struct sk_buff *pp = NULL;
+	int len, dlen;
+	__u8 *udpdata;
+	__be32 *udpdata32;
+
+	if (skb->protocol == htons(ETH_P_IP))
+		return xfrm4_gro_udp_encap_rcv(sk, head, skb);
+
+	len = skb->len - offset;
+	dlen = offset + min(len, 8);
+	udpdata = skb_gro_header(skb, dlen, offset);
+	udpdata32 = (__be32 *)udpdata;
+	if (unlikely(!udpdata))
+		return NULL;
+
+	rcu_read_lock();
+	ops = rcu_dereference(inet6_offloads[IPPROTO_ESP]);
+	if (!ops || !ops->callbacks.gro_receive)
+		goto out;
+
+	/* check if it is a keepalive or IKE packet */
+	if (len <= sizeof(struct ip_esp_hdr) || udpdata32[0] == 0)
+		goto out;
+
+	/* set the transport header to ESP */
+	skb_set_transport_header(skb, offset);
+
+	NAPI_GRO_CB(skb)->proto = IPPROTO_UDP;
+
+	pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
+	rcu_read_unlock();
+
+	return pp;
+
+out:
+	rcu_read_unlock();
+	NAPI_GRO_CB(skb)->same_flow = 0;
+	NAPI_GRO_CB(skb)->flush = 1;
+
+	return NULL;
+}
+
+int xfrm6_rcv_tnl(struct sk_buff *skb, struct ip6_tnl *t)
+{
+	return xfrm6_rcv_spi(skb, skb_network_header(skb)[IP6CB(skb)->nhoff],
+			     0, t);
+}
+EXPORT_SYMBOL(xfrm6_rcv_tnl);
+
+int xfrm6_rcv(struct sk_buff *skb)
+{
+	return xfrm6_rcv_tnl(skb, NULL);
+}
+EXPORT_SYMBOL(xfrm6_rcv);
+int xfrm6_input_addr(struct sk_buff *skb, xfrm_address_t *daddr,
+		     xfrm_address_t *saddr, u8 proto)
+{
+	struct net *net = dev_net(skb->dev);
+	struct xfrm_state *x = NULL;
+	struct sec_path *sp;
+	int i = 0;
+
+	sp = secpath_set(skb);
+	if (!sp) {
+		XFRM_INC_STATS(net, LINUX_MIB_XFRMINERROR);
+		goto drop;
+	}
+
+	if (sp->len >= XFRM_MAX_DEPTH) {
+		XFRM_INC_STATS(net, LINUX_MIB_XFRMINBUFFERERROR);
+		goto drop;
+	}
+
+	for (i = 0; i < 3; i++) {
+		xfrm_address_t *dst, *src;
+
+		switch (i) {
+		case 0:
+			dst = daddr;
+			src = saddr;
+			break;
+		case 1:
+			/* lookup state with wild-card source address */
+			dst = daddr;
+			src = (xfrm_address_t *)&in6addr_any;
+			break;
+		default:
+			/* lookup state with wild-card addresses */
+			dst = (xfrm_address_t *)&in6addr_any;
+			src = (xfrm_address_t *)&in6addr_any;
+			break;
+		}
+
+		x = xfrm_state_lookup_byaddr(net, skb->mark, dst, src, proto, AF_INET6);
+		if (!x)
+			continue;
+
+		if (unlikely(x->dir && x->dir != XFRM_SA_DIR_IN)) {
+			XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEDIRERROR);
+			xfrm_state_put(x);
+			x = NULL;
+			continue;
+		}
+
+		spin_lock(&x->lock);
+
+		if ((!i || (x->props.flags & XFRM_STATE_WILDRECV)) &&
+		    likely(x->km.state == XFRM_STATE_VALID) &&
+		    !xfrm_state_check_expire(x)) {
+			spin_unlock(&x->lock);
+			if (x->type->input(x, skb) > 0) {
+				/* found a valid state */
+				break;
+			}
+		} else
+			spin_unlock(&x->lock);
+
+		xfrm_state_put(x);
+		x = NULL;
+	}
+
+	if (!x) {
+		XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
+		xfrm_audit_state_notfound_simple(skb, AF_INET6);
+		goto drop;
+	}
+
+	sp->xvec[sp->len++] = x;
+
+	spin_lock(&x->lock);
+
+	x->curlft.bytes += skb->len;
+	x->curlft.packets++;
+
+	spin_unlock(&x->lock);
+
+	return 1;
+
+drop:
+	return -1;
+}
+EXPORT_SYMBOL(xfrm6_input_addr);