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, 295 insertions, 0 deletions

diff --git a/samples/timers/hpet_example.c b/samples/timers/hpet_example.c
new file mode 100644
index 000000000..f1cb622f6
--- /dev/null
+++ b/samples/timers/hpet_example.c
@@ -0,0 +1,295 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <string.h>
+#include <memory.h>
+#include <malloc.h>
+#include <time.h>
+#include <ctype.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <signal.h>
+#include <errno.h>
+#include <sys/time.h>
+#include <linux/hpet.h>
+
+
+extern void hpet_open_close(int, const char **);
+extern void hpet_info(int, const char **);
+extern void hpet_poll(int, const char **);
+extern void hpet_fasync(int, const char **);
+extern void hpet_read(int, const char **);
+
+#include <sys/poll.h>
+#include <sys/ioctl.h>
+
+struct hpet_command {
+	char		*command;
+	void		(*func)(int argc, const char ** argv);
+} hpet_command[] = {
+	{
+		"open-close",
+		hpet_open_close
+	},
+	{
+		"info",
+		hpet_info
+	},
+	{
+		"poll",
+		hpet_poll
+	},
+	{
+		"fasync",
+		hpet_fasync
+	},
+};
+
+int
+main(int argc, const char ** argv)
+{
+	unsigned int	i;
+
+	argc--;
+	argv++;
+
+	if (!argc) {
+		fprintf(stderr, "-hpet: requires command\n");
+		return -1;
+	}
+
+
+	for (i = 0; i < (sizeof (hpet_command) / sizeof (hpet_command[0])); i++)
+		if (!strcmp(argv[0], hpet_command[i].command)) {
+			argc--;
+			argv++;
+			fprintf(stderr, "-hpet: executing %s\n",
+				hpet_command[i].command);
+			hpet_command[i].func(argc, argv);
+			return 0;
+		}
+
+	fprintf(stderr, "do_hpet: command %s not implemented\n", argv[0]);
+
+	return -1;
+}
+
+void
+hpet_open_close(int argc, const char **argv)
+{
+	int	fd;
+
+	if (argc != 1) {
+		fprintf(stderr, "hpet_open_close: device-name\n");
+		return;
+	}
+
+	fd = open(argv[0], O_RDONLY);
+	if (fd < 0)
+		fprintf(stderr, "hpet_open_close: open failed\n");
+	else
+		close(fd);
+
+	return;
+}
+
+void
+hpet_info(int argc, const char **argv)
+{
+	struct hpet_info	info;
+	int			fd;
+
+	if (argc != 1) {
+		fprintf(stderr, "hpet_info: device-name\n");
+		return;
+	}
+
+	fd = open(argv[0], O_RDONLY);
+	if (fd < 0) {
+		fprintf(stderr, "hpet_info: open of %s failed\n", argv[0]);
+		return;
+	}
+
+	if (ioctl(fd, HPET_INFO, &info) < 0) {
+		fprintf(stderr, "hpet_info: failed to get info\n");
+		goto out;
+	}
+
+	fprintf(stderr, "hpet_info: hi_irqfreq 0x%lx hi_flags 0x%lx ",
+		info.hi_ireqfreq, info.hi_flags);
+	fprintf(stderr, "hi_hpet %d hi_timer %d\n",
+		info.hi_hpet, info.hi_timer);
+
+out:
+	close(fd);
+	return;
+}
+
+void
+hpet_poll(int argc, const char **argv)
+{
+	unsigned long		freq;
+	int			iterations, i, fd;
+	struct pollfd		pfd;
+	struct hpet_info	info;
+	struct timeval		stv, etv;
+	struct timezone		tz;
+	long			usec;
+
+	if (argc != 3) {
+		fprintf(stderr, "hpet_poll: device-name freq iterations\n");
+		return;
+	}
+
+	freq = atoi(argv[1]);
+	iterations = atoi(argv[2]);
+
+	fd = open(argv[0], O_RDONLY);
+
+	if (fd < 0) {
+		fprintf(stderr, "hpet_poll: open of %s failed\n", argv[0]);
+		return;
+	}
+
+	if (ioctl(fd, HPET_IRQFREQ, freq) < 0) {
+		fprintf(stderr, "hpet_poll: HPET_IRQFREQ failed\n");
+		goto out;
+	}
+
+	if (ioctl(fd, HPET_INFO, &info) < 0) {
+		fprintf(stderr, "hpet_poll: failed to get info\n");
+		goto out;
+	}
+
+	fprintf(stderr, "hpet_poll: info.hi_flags 0x%lx\n", info.hi_flags);
+
+	if (info.hi_flags && (ioctl(fd, HPET_EPI, 0) < 0)) {
+		fprintf(stderr, "hpet_poll: HPET_EPI failed\n");
+		goto out;
+	}
+
+	if (ioctl(fd, HPET_IE_ON, 0) < 0) {
+		fprintf(stderr, "hpet_poll, HPET_IE_ON failed\n");
+		goto out;
+	}
+
+	pfd.fd = fd;
+	pfd.events = POLLIN;
+
+	for (i = 0; i < iterations; i++) {
+		pfd.revents = 0;
+		gettimeofday(&stv, &tz);
+		if (poll(&pfd, 1, -1) < 0)
+			fprintf(stderr, "hpet_poll: poll failed\n");
+		else {
+			long 	data;
+
+			gettimeofday(&etv, &tz);
+			usec = stv.tv_sec * 1000000 + stv.tv_usec;
+			usec = (etv.tv_sec * 1000000 + etv.tv_usec) - usec;
+
+			fprintf(stderr,
+				"hpet_poll: expired time = 0x%lx\n", usec);
+
+			fprintf(stderr, "hpet_poll: revents = 0x%x\n",
+				pfd.revents);
+
+			if (read(fd, &data, sizeof(data)) != sizeof(data)) {
+				fprintf(stderr, "hpet_poll: read failed\n");
+			}
+			else
+				fprintf(stderr, "hpet_poll: data 0x%lx\n",
+					data);
+		}
+	}
+
+out:
+	close(fd);
+	return;
+}
+
+static int hpet_sigio_count;
+
+static void
+hpet_sigio(int val)
+{
+	fprintf(stderr, "hpet_sigio: called\n");
+	hpet_sigio_count++;
+}
+
+void
+hpet_fasync(int argc, const char **argv)
+{
+	unsigned long		freq;
+	int			iterations, i, fd, value;
+	sig_t			oldsig;
+	struct hpet_info	info;
+
+	hpet_sigio_count = 0;
+	fd = -1;
+
+	if ((oldsig = signal(SIGIO, hpet_sigio)) == SIG_ERR) {
+		fprintf(stderr, "hpet_fasync: failed to set signal handler\n");
+		return;
+	}
+
+	if (argc != 3) {
+		fprintf(stderr, "hpet_fasync: device-name freq iterations\n");
+		goto out;
+	}
+
+	fd = open(argv[0], O_RDONLY);
+
+	if (fd < 0) {
+		fprintf(stderr, "hpet_fasync: failed to open %s\n", argv[0]);
+		return;
+	}
+
+
+	if ((fcntl(fd, F_SETOWN, getpid()) == 1) ||
+		((value = fcntl(fd, F_GETFL)) == 1) ||
+		(fcntl(fd, F_SETFL, value | O_ASYNC) == 1)) {
+		fprintf(stderr, "hpet_fasync: fcntl failed\n");
+		goto out;
+	}
+
+	freq = atoi(argv[1]);
+	iterations = atoi(argv[2]);
+
+	if (ioctl(fd, HPET_IRQFREQ, freq) < 0) {
+		fprintf(stderr, "hpet_fasync: HPET_IRQFREQ failed\n");
+		goto out;
+	}
+
+	if (ioctl(fd, HPET_INFO, &info) < 0) {
+		fprintf(stderr, "hpet_fasync: failed to get info\n");
+		goto out;
+	}
+
+	fprintf(stderr, "hpet_fasync: info.hi_flags 0x%lx\n", info.hi_flags);
+
+	if (info.hi_flags && (ioctl(fd, HPET_EPI, 0) < 0)) {
+		fprintf(stderr, "hpet_fasync: HPET_EPI failed\n");
+		goto out;
+	}
+
+	if (ioctl(fd, HPET_IE_ON, 0) < 0) {
+		fprintf(stderr, "hpet_fasync, HPET_IE_ON failed\n");
+		goto out;
+	}
+
+	for (i = 0; i < iterations; i++) {
+		(void) pause();
+		fprintf(stderr, "hpet_fasync: count = %d\n", hpet_sigio_count);
+	}
+
+out:
+	signal(SIGIO, oldsig);
+
+	if (fd >= 0)
+		close(fd);
+
+	return;
+}