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, 311 insertions, 0 deletions

diff --git a/net/mctp/test/utils.c b/net/mctp/test/utils.c
new file mode 100644
index 000000000..6b131084a
--- /dev/null
+++ b/net/mctp/test/utils.c
@@ -0,0 +1,311 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/netdevice.h>
+#include <linux/mctp.h>
+#include <linux/if_arp.h>
+
+#include <net/mctp.h>
+#include <net/mctpdevice.h>
+#include <net/pkt_sched.h>
+
+#include "utils.h"
+
+static netdev_tx_t mctp_test_dev_tx(struct sk_buff *skb,
+				    struct net_device *ndev)
+{
+	struct mctp_test_dev *dev = netdev_priv(ndev);
+
+	skb_queue_tail(&dev->pkts, skb);
+
+	return NETDEV_TX_OK;
+}
+
+static const struct net_device_ops mctp_test_netdev_ops = {
+	.ndo_start_xmit = mctp_test_dev_tx,
+};
+
+static void mctp_test_dev_setup(struct net_device *ndev)
+{
+	ndev->type = ARPHRD_MCTP;
+	ndev->mtu = MCTP_DEV_TEST_MTU;
+	ndev->hard_header_len = 0;
+	ndev->tx_queue_len = 0;
+	ndev->flags = IFF_NOARP;
+	ndev->netdev_ops = &mctp_test_netdev_ops;
+	ndev->needs_free_netdev = true;
+}
+
+static struct mctp_test_dev *__mctp_test_create_dev(unsigned short lladdr_len,
+						    const unsigned char *lladdr)
+{
+	struct mctp_test_dev *dev;
+	struct net_device *ndev;
+	int rc;
+
+	if (WARN_ON(lladdr_len > MAX_ADDR_LEN))
+		return NULL;
+
+	ndev = alloc_netdev(sizeof(*dev), "mctptest%d", NET_NAME_ENUM,
+			    mctp_test_dev_setup);
+	if (!ndev)
+		return NULL;
+
+	dev = netdev_priv(ndev);
+	dev->ndev = ndev;
+	ndev->addr_len = lladdr_len;
+	dev_addr_set(ndev, lladdr);
+	skb_queue_head_init(&dev->pkts);
+
+	rc = register_netdev(ndev);
+	if (rc) {
+		free_netdev(ndev);
+		return NULL;
+	}
+
+	rcu_read_lock();
+	dev->mdev = __mctp_dev_get(ndev);
+	dev->mdev->net = mctp_default_net(dev_net(ndev));
+	rcu_read_unlock();
+
+	/* bring the device up; we want to be able to TX immediately */
+	rtnl_lock();
+	dev_open(ndev, NULL);
+	rtnl_unlock();
+
+	return dev;
+}
+
+struct mctp_test_dev *mctp_test_create_dev(void)
+{
+	return __mctp_test_create_dev(0, NULL);
+}
+
+struct mctp_test_dev *mctp_test_create_dev_with_addr(mctp_eid_t addr)
+{
+	struct mctp_test_dev *dev;
+
+	dev = __mctp_test_create_dev(0, NULL);
+	if (!dev)
+		return NULL;
+
+	dev->mdev->addrs = kmalloc_objs(u8, 1);
+	if (!dev->mdev->addrs) {
+		mctp_test_destroy_dev(dev);
+		return NULL;
+	}
+
+	dev->mdev->num_addrs = 1;
+	dev->mdev->addrs[0] = addr;
+
+	return dev;
+}
+
+struct mctp_test_dev *mctp_test_create_dev_lladdr(unsigned short lladdr_len,
+						  const unsigned char *lladdr)
+{
+	return __mctp_test_create_dev(lladdr_len, lladdr);
+}
+
+void mctp_test_destroy_dev(struct mctp_test_dev *dev)
+{
+	skb_queue_purge(&dev->pkts);
+	mctp_dev_put(dev->mdev);
+	unregister_netdev(dev->ndev);
+}
+
+static int mctp_test_dst_output(struct mctp_dst *dst, struct sk_buff *skb)
+{
+	skb->dev = dst->dev->dev;
+	dev_direct_xmit(skb, 0);
+
+	return 0;
+}
+
+/* local version of mctp_route_alloc() */
+static struct mctp_test_route *mctp_route_test_alloc(void)
+{
+	struct mctp_test_route *rt;
+
+	rt = kzalloc_obj(*rt);
+	if (!rt)
+		return NULL;
+
+	INIT_LIST_HEAD(&rt->rt.list);
+	refcount_set(&rt->rt.refs, 1);
+	rt->rt.output = mctp_test_dst_output;
+
+	return rt;
+}
+
+struct mctp_test_route *mctp_test_create_route_direct(struct net *net,
+						      struct mctp_dev *dev,
+						      mctp_eid_t eid,
+						      unsigned int mtu)
+{
+	struct mctp_test_route *rt;
+
+	rt = mctp_route_test_alloc();
+	if (!rt)
+		return NULL;
+
+	rt->rt.min = eid;
+	rt->rt.max = eid;
+	rt->rt.mtu = mtu;
+	rt->rt.type = RTN_UNSPEC;
+	rt->rt.dst_type = MCTP_ROUTE_DIRECT;
+	if (dev)
+		mctp_dev_hold(dev);
+	rt->rt.dev = dev;
+
+	list_add_rcu(&rt->rt.list, &net->mctp.routes);
+
+	return rt;
+}
+
+struct mctp_test_route *mctp_test_create_route_gw(struct net *net,
+						  unsigned int netid,
+						  mctp_eid_t eid,
+						  mctp_eid_t gw,
+						  unsigned int mtu)
+{
+	struct mctp_test_route *rt;
+
+	rt = mctp_route_test_alloc();
+	if (!rt)
+		return NULL;
+
+	rt->rt.min = eid;
+	rt->rt.max = eid;
+	rt->rt.mtu = mtu;
+	rt->rt.type = RTN_UNSPEC;
+	rt->rt.dst_type = MCTP_ROUTE_GATEWAY;
+	rt->rt.gateway.eid = gw;
+	rt->rt.gateway.net = netid;
+
+	list_add_rcu(&rt->rt.list, &net->mctp.routes);
+
+	return rt;
+}
+
+/* Convenience function for our test dst; release with mctp_dst_release() */
+void mctp_test_dst_setup(struct kunit *test, struct mctp_dst *dst,
+			 struct mctp_test_dev *dev, unsigned int mtu)
+{
+	unsigned long flags;
+
+	KUNIT_EXPECT_NOT_ERR_OR_NULL(test, dev);
+
+	memset(dst, 0, sizeof(*dst));
+
+	dst->dev = dev->mdev;
+	__mctp_dev_get(dst->dev->dev);
+	dst->mtu = mtu;
+	dst->output = mctp_test_dst_output;
+	dst->saddr = MCTP_ADDR_NULL;
+	spin_lock_irqsave(&dev->mdev->addrs_lock, flags);
+	if (dev->mdev->num_addrs)
+		dst->saddr = dev->mdev->addrs[0];
+	spin_unlock_irqrestore(&dev->mdev->addrs_lock, flags);
+}
+
+void mctp_test_route_destroy(struct kunit *test, struct mctp_test_route *rt)
+{
+	unsigned int refs;
+
+	rtnl_lock();
+	list_del_rcu(&rt->rt.list);
+	rtnl_unlock();
+
+	if (rt->rt.dst_type == MCTP_ROUTE_DIRECT && rt->rt.dev)
+		mctp_dev_put(rt->rt.dev);
+
+	refs = refcount_read(&rt->rt.refs);
+	KUNIT_ASSERT_EQ_MSG(test, refs, 1, "route ref imbalance");
+
+	kfree_rcu(&rt->rt, rcu);
+}
+
+void mctp_test_skb_set_dev(struct sk_buff *skb, struct mctp_test_dev *dev)
+{
+	struct mctp_skb_cb *cb;
+
+	cb = mctp_cb(skb);
+	cb->net = READ_ONCE(dev->mdev->net);
+	skb->dev = dev->ndev;
+}
+
+struct sk_buff *mctp_test_create_skb(const struct mctp_hdr *hdr,
+				     unsigned int data_len)
+{
+	size_t hdr_len = sizeof(*hdr);
+	struct sk_buff *skb;
+	unsigned int i;
+	u8 *buf;
+
+	skb = alloc_skb(hdr_len + data_len, GFP_KERNEL);
+	if (!skb)
+		return NULL;
+
+	__mctp_cb(skb);
+	memcpy(skb_put(skb, hdr_len), hdr, hdr_len);
+
+	buf = skb_put(skb, data_len);
+	for (i = 0; i < data_len; i++)
+		buf[i] = i & 0xff;
+
+	return skb;
+}
+
+struct sk_buff *__mctp_test_create_skb_data(const struct mctp_hdr *hdr,
+					    const void *data, size_t data_len)
+{
+	size_t hdr_len = sizeof(*hdr);
+	struct sk_buff *skb;
+
+	skb = alloc_skb(hdr_len + data_len, GFP_KERNEL);
+	if (!skb)
+		return NULL;
+
+	__mctp_cb(skb);
+	memcpy(skb_put(skb, hdr_len), hdr, hdr_len);
+	memcpy(skb_put(skb, data_len), data, data_len);
+
+	return skb;
+}
+
+void mctp_test_bind_run(struct kunit *test,
+			const struct mctp_test_bind_setup *setup,
+			int *ret_bind_errno, struct socket **sock)
+{
+	struct sockaddr_mctp addr;
+	int rc;
+
+	*ret_bind_errno = -EIO;
+
+	rc = sock_create_kern(&init_net, AF_MCTP, SOCK_DGRAM, 0, sock);
+	KUNIT_ASSERT_EQ(test, rc, 0);
+
+	/* connect() if requested */
+	if (setup->have_peer) {
+		memset(&addr, 0x0, sizeof(addr));
+		addr.smctp_family = AF_MCTP;
+		addr.smctp_network = setup->peer_net;
+		addr.smctp_addr.s_addr = setup->peer_addr;
+		/* connect() type must match bind() type */
+		addr.smctp_type = setup->bind_type;
+		rc = kernel_connect(*sock, (struct sockaddr_unsized *)&addr,
+				    sizeof(addr), 0);
+		KUNIT_EXPECT_EQ(test, rc, 0);
+	}
+
+	/* bind() */
+	memset(&addr, 0x0, sizeof(addr));
+	addr.smctp_family = AF_MCTP;
+	addr.smctp_network = setup->bind_net;
+	addr.smctp_addr.s_addr = setup->bind_addr;
+	addr.smctp_type = setup->bind_type;
+
+	*ret_bind_errno =
+		kernel_bind(*sock, (struct sockaddr_unsized *)&addr,
+			    sizeof(addr));
+}