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, 236 insertions, 0 deletions

diff --git a/arch/um/os-Linux/helper.c b/arch/um/os-Linux/helper.c
new file mode 100644
index 000000000..89c2ad2a4
--- /dev/null
+++ b/arch/um/os-Linux/helper.c
@@ -0,0 +1,236 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2002 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <sched.h>
+#include <pthread.h>
+#include <linux/limits.h>
+#include <sys/socket.h>
+#include <sys/wait.h>
+#include <kern_util.h>
+#include <os.h>
+#include <um_malloc.h>
+
+struct helper_data {
+	void (*pre_exec)(void*);
+	void *pre_data;
+	char **argv;
+	int fd;
+	char *buf;
+};
+
+static int helper_child(void *arg)
+{
+	struct helper_data *data = arg;
+	char **argv = data->argv;
+	int err, ret;
+
+	if (data->pre_exec != NULL)
+		(*data->pre_exec)(data->pre_data);
+	err = execvp_noalloc(data->buf, argv[0], argv);
+
+	/* If the exec succeeds, we don't get here */
+	CATCH_EINTR(ret = write(data->fd, &err, sizeof(err)));
+
+	return 0;
+}
+
+/* Returns either the pid of the child process we run or -E* on failure. */
+int run_helper(void (*pre_exec)(void *), void *pre_data, char **argv)
+{
+	struct helper_data data;
+	unsigned long stack, sp;
+	int pid, fds[2], ret, n;
+
+	stack = alloc_stack(0, __uml_cant_sleep());
+	if (stack == 0)
+		return -ENOMEM;
+
+	ret = socketpair(AF_UNIX, SOCK_STREAM, 0, fds);
+	if (ret < 0) {
+		ret = -errno;
+		printk(UM_KERN_ERR "run_helper : pipe failed, errno = %d\n",
+		       errno);
+		goto out_free;
+	}
+
+	ret = os_set_exec_close(fds[1]);
+	if (ret < 0) {
+		printk(UM_KERN_ERR "run_helper : setting FD_CLOEXEC failed, "
+		       "ret = %d\n", -ret);
+		goto out_close;
+	}
+
+	sp = stack + UM_KERN_PAGE_SIZE;
+	data.pre_exec = pre_exec;
+	data.pre_data = pre_data;
+	data.argv = argv;
+	data.fd = fds[1];
+	data.buf = __uml_cant_sleep() ? uml_kmalloc(PATH_MAX, UM_GFP_ATOMIC) :
+					uml_kmalloc(PATH_MAX, UM_GFP_KERNEL);
+	pid = clone(helper_child, (void *) sp, CLONE_VM, &data);
+	if (pid < 0) {
+		ret = -errno;
+		printk(UM_KERN_ERR "run_helper : clone failed, errno = %d\n",
+		       errno);
+		goto out_free2;
+	}
+
+	close(fds[1]);
+	fds[1] = -1;
+
+	/*
+	 * Read the errno value from the child, if the exec failed, or get 0 if
+	 * the exec succeeded because the pipe fd was set as close-on-exec.
+	 */
+	n = read(fds[0], &ret, sizeof(ret));
+	if (n == 0) {
+		ret = pid;
+	} else {
+		if (n < 0) {
+			n = -errno;
+			printk(UM_KERN_ERR "run_helper : read on pipe failed, "
+			       "ret = %d\n", -n);
+			ret = n;
+		}
+		CATCH_EINTR(waitpid(pid, NULL, __WALL));
+	}
+
+	if (ret < 0)
+		printk(UM_KERN_ERR "run_helper : failed to exec %s on host: %s\n",
+		       argv[0], strerror(-ret));
+
+out_free2:
+	kfree(data.buf);
+out_close:
+	if (fds[1] != -1)
+		close(fds[1]);
+	close(fds[0]);
+out_free:
+	free_stack(stack, 0);
+	return ret;
+}
+
+int run_helper_thread(int (*proc)(void *), void *arg, unsigned int flags,
+		      unsigned long *stack_out)
+{
+	unsigned long stack, sp;
+	int pid, status, err;
+
+	/* To share memory space, use os_run_helper_thread() instead. */
+	if (flags & CLONE_VM)
+		return -EINVAL;
+
+	stack = alloc_stack(0, __uml_cant_sleep());
+	if (stack == 0)
+		return -ENOMEM;
+
+	sp = stack + UM_KERN_PAGE_SIZE;
+	pid = clone(proc, (void *) sp, flags, arg);
+	if (pid < 0) {
+		err = -errno;
+		printk(UM_KERN_ERR "run_helper_thread : clone failed, "
+		       "errno = %d\n", errno);
+		return err;
+	}
+	if (stack_out == NULL) {
+		CATCH_EINTR(pid = waitpid(pid, &status, __WALL));
+		if (pid < 0) {
+			err = -errno;
+			printk(UM_KERN_ERR "run_helper_thread - wait failed, "
+			       "errno = %d\n", errno);
+			pid = err;
+		}
+		if (!WIFEXITED(status) || (WEXITSTATUS(status) != 0))
+			printk(UM_KERN_ERR "run_helper_thread - thread "
+			       "returned status 0x%x\n", status);
+		free_stack(stack, 0);
+	} else
+		*stack_out = stack;
+	return pid;
+}
+
+int helper_wait(int pid)
+{
+	int ret, status;
+	int wflags = __WALL;
+
+	CATCH_EINTR(ret = waitpid(pid, &status, wflags));
+	if (ret < 0) {
+		printk(UM_KERN_ERR "helper_wait : waitpid process %d failed, "
+		       "errno = %d\n", pid, errno);
+		return -errno;
+	} else if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+		printk(UM_KERN_ERR "helper_wait : process %d exited with "
+		       "status 0x%x\n", pid, status);
+		return -ECHILD;
+	} else
+		return 0;
+}
+
+struct os_helper_thread {
+	pthread_t handle;
+};
+
+int os_run_helper_thread(struct os_helper_thread **td_out,
+			 void *(*routine)(void *), void *arg)
+{
+	struct os_helper_thread *td;
+	sigset_t sigset, oset;
+	int err, flags;
+
+	flags = __uml_cant_sleep() ? UM_GFP_ATOMIC : UM_GFP_KERNEL;
+	td = uml_kmalloc(sizeof(*td), flags);
+	if (!td)
+		return -ENOMEM;
+
+	sigfillset(&sigset);
+	if (sigprocmask(SIG_SETMASK, &sigset, &oset) < 0) {
+		err = -errno;
+		kfree(td);
+		return err;
+	}
+
+	err = pthread_create(&td->handle, NULL, routine, arg);
+
+	if (sigprocmask(SIG_SETMASK, &oset, NULL) < 0)
+		panic("Failed to restore the signal mask: %d", errno);
+
+	if (err != 0)
+		kfree(td);
+	else
+		*td_out = td;
+
+	return -err;
+}
+
+void os_kill_helper_thread(struct os_helper_thread *td)
+{
+	pthread_cancel(td->handle);
+	pthread_join(td->handle, NULL);
+	kfree(td);
+}
+
+void os_fix_helper_thread_signals(void)
+{
+	sigset_t sigset;
+
+	sigemptyset(&sigset);
+
+	sigaddset(&sigset, SIGWINCH);
+	sigaddset(&sigset, SIGPIPE);
+	sigaddset(&sigset, SIGPROF);
+	sigaddset(&sigset, SIGINT);
+	sigaddset(&sigset, SIGTERM);
+	sigaddset(&sigset, SIGCHLD);
+	sigaddset(&sigset, SIGALRM);
+	sigaddset(&sigset, SIGIO);
+	sigaddset(&sigset, SIGUSR1);
+
+	pthread_sigmask(SIG_SETMASK, &sigset, NULL);
+}