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

4 files changed, 432 insertions, 0 deletions

diff --git a/samples/connector/.gitignore b/samples/connector/.gitignore
new file mode 100644
index 000000000..0e26039f3
--- /dev/null
+++ b/samples/connector/.gitignore
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0-only
+/ucon
diff --git a/samples/connector/Makefile b/samples/connector/Makefile
new file mode 100644
index 000000000..d98a9e047
--- /dev/null
+++ b/samples/connector/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0
+obj-$(CONFIG_SAMPLE_CONNECTOR) += cn_test.o
+
+userprogs-always-$(CONFIG_CC_CAN_LINK) += ucon
+
+userccflags += -I usr/include
diff --git a/samples/connector/cn_test.c b/samples/connector/cn_test.c
new file mode 100644
index 000000000..73d50b4ae
--- /dev/null
+++ b/samples/connector/cn_test.c
@@ -0,0 +1,188 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * 	cn_test.c
+ * 
+ * 2004+ Copyright (c) Evgeniy Polyakov <zbr@ioremap.net>
+ * All rights reserved.
+ */
+
+#define pr_fmt(fmt) "cn_test: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/skbuff.h>
+#include <linux/slab.h>
+#include <linux/timer.h>
+
+#include <linux/connector.h>
+
+static struct cb_id cn_test_id = { CN_NETLINK_USERS + 3, 0x456 };
+static char cn_test_name[] = "cn_test";
+static struct sock *nls;
+static struct timer_list cn_test_timer;
+
+static void cn_test_callback(struct cn_msg *msg, struct netlink_skb_parms *nsp)
+{
+	pr_info("%s: %lu: idx=%x, val=%x, seq=%u, ack=%u, len=%d: %s.\n",
+	        __func__, jiffies, msg->id.idx, msg->id.val,
+	        msg->seq, msg->ack, msg->len,
+	        msg->len ? (char *)msg->data : "");
+}
+
+/*
+ * Do not remove this function even if no one is using it as
+ * this is an example of how to get notifications about new
+ * connector user registration
+ */
+#if 0
+static int cn_test_want_notify(void)
+{
+	struct cn_ctl_msg *ctl;
+	struct cn_notify_req *req;
+	struct cn_msg *msg = NULL;
+	int size, size0;
+	struct sk_buff *skb;
+	struct nlmsghdr *nlh;
+	u32 group = 1;
+
+	size0 = sizeof(*msg) + sizeof(*ctl) + 3 * sizeof(*req);
+
+	size = NLMSG_SPACE(size0);
+
+	skb = alloc_skb(size, GFP_ATOMIC);
+	if (!skb) {
+		pr_err("failed to allocate new skb with size=%u\n", size);
+		return -ENOMEM;
+	}
+
+	nlh = nlmsg_put(skb, 0, 0x123, NLMSG_DONE, size - sizeof(*nlh), 0);
+	if (!nlh) {
+		kfree_skb(skb);
+		return -EMSGSIZE;
+	}
+
+	msg = nlmsg_data(nlh);
+
+	memset(msg, 0, size0);
+
+	msg->id.idx = -1;
+	msg->id.val = -1;
+	msg->seq = 0x123;
+	msg->ack = 0x345;
+	msg->len = size0 - sizeof(*msg);
+
+	ctl = (struct cn_ctl_msg *)(msg + 1);
+
+	ctl->idx_notify_num = 1;
+	ctl->val_notify_num = 2;
+	ctl->group = group;
+	ctl->len = msg->len - sizeof(*ctl);
+
+	req = (struct cn_notify_req *)(ctl + 1);
+
+	/*
+	 * Idx.
+	 */
+	req->first = cn_test_id.idx;
+	req->range = 10;
+
+	/*
+	 * Val 0.
+	 */
+	req++;
+	req->first = cn_test_id.val;
+	req->range = 10;
+
+	/*
+	 * Val 1.
+	 */
+	req++;
+	req->first = cn_test_id.val + 20;
+	req->range = 10;
+
+	NETLINK_CB(skb).dst_group = ctl->group;
+	//netlink_broadcast(nls, skb, 0, ctl->group, GFP_ATOMIC);
+	netlink_unicast(nls, skb, 0, 0);
+
+	pr_info("request was sent: group=0x%x\n", ctl->group);
+
+	return 0;
+}
+#endif
+
+static u32 cn_test_timer_counter;
+static void cn_test_timer_func(struct timer_list *unused)
+{
+	struct cn_msg *m;
+	char data[32];
+
+	pr_debug("%s: timer fired\n", __func__);
+
+	m = kzalloc(sizeof(*m) + sizeof(data), GFP_ATOMIC);
+	if (m) {
+
+		memcpy(&m->id, &cn_test_id, sizeof(m->id));
+		m->seq = cn_test_timer_counter;
+		m->len = sizeof(data);
+
+		m->len =
+		    scnprintf(data, sizeof(data), "counter = %u",
+			      cn_test_timer_counter) + 1;
+
+		memcpy(m + 1, data, m->len);
+
+		cn_netlink_send(m, 0, 0, GFP_ATOMIC);
+		kfree(m);
+	}
+
+	cn_test_timer_counter++;
+
+	mod_timer(&cn_test_timer, jiffies + msecs_to_jiffies(1000));
+}
+
+static int cn_test_init(void)
+{
+	int err;
+
+	err = cn_add_callback(&cn_test_id, cn_test_name, cn_test_callback);
+	if (err)
+		goto err_out;
+	cn_test_id.val++;
+	err = cn_add_callback(&cn_test_id, cn_test_name, cn_test_callback);
+	if (err) {
+		cn_del_callback(&cn_test_id);
+		goto err_out;
+	}
+
+	timer_setup(&cn_test_timer, cn_test_timer_func, 0);
+	mod_timer(&cn_test_timer, jiffies + msecs_to_jiffies(1000));
+
+	pr_info("initialized with id={%u.%u}\n",
+		cn_test_id.idx, cn_test_id.val);
+
+	return 0;
+
+      err_out:
+	if (nls && nls->sk_socket)
+		sock_release(nls->sk_socket);
+
+	return err;
+}
+
+static void cn_test_fini(void)
+{
+	timer_delete_sync(&cn_test_timer);
+	cn_del_callback(&cn_test_id);
+	cn_test_id.val--;
+	cn_del_callback(&cn_test_id);
+	if (nls && nls->sk_socket)
+		sock_release(nls->sk_socket);
+}
+
+module_init(cn_test_init);
+module_exit(cn_test_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Evgeniy Polyakov <zbr@ioremap.net>");
+MODULE_DESCRIPTION("Connector's test module");
diff --git a/samples/connector/ucon.c b/samples/connector/ucon.c
new file mode 100644
index 000000000..fa17f8642
--- /dev/null
+++ b/samples/connector/ucon.c
@@ -0,0 +1,236 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * 	ucon.c
+ *
+ * Copyright (c) 2004+ Evgeniy Polyakov <zbr@ioremap.net>
+ */
+
+#include <asm/types.h>
+
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <sys/poll.h>
+
+#include <linux/netlink.h>
+#include <linux/rtnetlink.h>
+
+#include <arpa/inet.h>
+
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+#include <errno.h>
+#include <time.h>
+#include <getopt.h>
+
+#include <linux/connector.h>
+
+#define DEBUG
+#define NETLINK_CONNECTOR 	11
+
+/* Hopefully your userspace connector.h matches this kernel */
+#define CN_TEST_IDX		CN_NETLINK_USERS + 3
+#define CN_TEST_VAL		0x456
+
+#ifdef DEBUG
+#define ulog(f, a...) fprintf(stdout, f, ##a)
+#else
+#define ulog(f, a...) do {} while (0)
+#endif
+
+static int need_exit;
+static __u32 seq;
+
+static int netlink_send(int s, struct cn_msg *msg)
+{
+	struct nlmsghdr *nlh;
+	unsigned int size;
+	int err;
+	char buf[128];
+	struct cn_msg *m;
+
+	size = NLMSG_SPACE(sizeof(struct cn_msg) + msg->len);
+
+	nlh = (struct nlmsghdr *)buf;
+	nlh->nlmsg_seq = seq++;
+	nlh->nlmsg_pid = getpid();
+	nlh->nlmsg_type = NLMSG_DONE;
+	nlh->nlmsg_len = size;
+	nlh->nlmsg_flags = 0;
+
+	m = NLMSG_DATA(nlh);
+#if 0
+	ulog("%s: [%08x.%08x] len=%u, seq=%u, ack=%u.\n",
+	       __func__, msg->id.idx, msg->id.val, msg->len, msg->seq, msg->ack);
+#endif
+	memcpy(m, msg, sizeof(*m) + msg->len);
+
+	err = send(s, nlh, size, 0);
+	if (err == -1)
+		ulog("Failed to send: %s [%d].\n",
+			strerror(errno), errno);
+
+	return err;
+}
+
+static void usage(void)
+{
+	printf(
+		"Usage: ucon [options] [output file]\n"
+		"\n"
+		"\t-h\tthis help screen\n"
+		"\t-s\tsend buffers to the test module\n"
+		"\n"
+		"The default behavior of ucon is to subscribe to the test module\n"
+		"and wait for state messages.  Any ones received are dumped to the\n"
+		"specified output file (or stdout).  The test module is assumed to\n"
+		"have an id of {%u.%u}\n"
+		"\n"
+		"If you get no output, then verify the cn_test module id matches\n"
+		"the expected id above.\n"
+		, CN_TEST_IDX, CN_TEST_VAL
+	);
+}
+
+int main(int argc, char *argv[])
+{
+	int s;
+	char buf[1024];
+	int len;
+	struct nlmsghdr *reply;
+	struct sockaddr_nl l_local;
+	struct cn_msg *data;
+	FILE *out;
+	time_t tm;
+	struct pollfd pfd;
+	bool send_msgs = false;
+
+	while ((s = getopt(argc, argv, "hs")) != -1) {
+		switch (s) {
+		case 's':
+			send_msgs = true;
+			break;
+
+		case 'h':
+			usage();
+			return 0;
+
+		default:
+			/* getopt() outputs an error for us */
+			usage();
+			return 1;
+		}
+	}
+
+	if (argc != optind) {
+		out = fopen(argv[optind], "a+");
+		if (!out) {
+			ulog("Unable to open %s for writing: %s\n",
+				argv[1], strerror(errno));
+			out = stdout;
+		}
+	} else
+		out = stdout;
+
+	memset(buf, 0, sizeof(buf));
+
+	s = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_CONNECTOR);
+	if (s == -1) {
+		perror("socket");
+		return -1;
+	}
+
+	l_local.nl_family = AF_NETLINK;
+	l_local.nl_groups = -1; /* bitmask of requested groups */
+	l_local.nl_pid = 0;
+
+	ulog("subscribing to %u.%u\n", CN_TEST_IDX, CN_TEST_VAL);
+
+	if (bind(s, (struct sockaddr *)&l_local, sizeof(struct sockaddr_nl)) == -1) {
+		perror("bind");
+		close(s);
+		return -1;
+	}
+
+#if 0
+	{
+		int on = 0x57; /* Additional group number */
+		setsockopt(s, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &on, sizeof(on));
+	}
+#endif
+	if (send_msgs) {
+		int i, j;
+
+		memset(buf, 0, sizeof(buf));
+
+		data = (struct cn_msg *)buf;
+
+		data->id.idx = CN_TEST_IDX;
+		data->id.val = CN_TEST_VAL;
+		data->seq = seq++;
+		data->ack = 0;
+		data->len = 0;
+
+		for (j=0; j<10; ++j) {
+			for (i=0; i<1000; ++i) {
+				len = netlink_send(s, data);
+			}
+
+			ulog("%d messages have been sent to %08x.%08x.\n", i, data->id.idx, data->id.val);
+		}
+
+		return 0;
+	}
+
+
+	pfd.fd = s;
+
+	while (!need_exit) {
+		pfd.events = POLLIN;
+		pfd.revents = 0;
+		switch (poll(&pfd, 1, -1)) {
+			case 0:
+				need_exit = 1;
+				break;
+			case -1:
+				if (errno != EINTR) {
+					need_exit = 1;
+					break;
+				}
+				continue;
+		}
+		if (need_exit)
+			break;
+
+		memset(buf, 0, sizeof(buf));
+		len = recv(s, buf, sizeof(buf), 0);
+		if (len == -1) {
+			perror("recv buf");
+			close(s);
+			return -1;
+		}
+		reply = (struct nlmsghdr *)buf;
+
+		switch (reply->nlmsg_type) {
+		case NLMSG_ERROR:
+			fprintf(out, "Error message received.\n");
+			fflush(out);
+			break;
+		case NLMSG_DONE:
+			data = (struct cn_msg *)NLMSG_DATA(reply);
+
+			time(&tm);
+			fprintf(out, "%.24s : [%x.%x] [%08u.%08u].\n",
+				ctime(&tm), data->id.idx, data->id.val, data->seq, data->ack);
+			fflush(out);
+			break;
+		default:
+			break;
+		}
+	}
+
+	close(s);
+	return 0;
+}