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, 226 insertions, 0 deletions

diff --git a/arch/um/drivers/xterm.c b/arch/um/drivers/xterm.c
new file mode 100644
index 000000000..d05918e42
--- /dev/null
+++ b/arch/um/drivers/xterm.c
@@ -0,0 +1,226 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2001 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
+ */
+
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <errno.h>
+#include <string.h>
+#include <termios.h>
+#include "chan_user.h"
+#include <os.h>
+#include <um_malloc.h>
+#include "xterm.h"
+
+struct xterm_chan {
+	int pid;
+	int helper_pid;
+	int chan_fd;
+	char *title;
+	int device;
+	int raw;
+	struct termios tt;
+};
+
+static void *xterm_init(char *str, int device, const struct chan_opts *opts)
+{
+	struct xterm_chan *data;
+
+	data = uml_kmalloc(sizeof(*data), UM_GFP_KERNEL);
+	if (data == NULL)
+		return NULL;
+	*data = ((struct xterm_chan) { .pid 		= -1,
+				       .helper_pid 	= -1,
+				       .chan_fd		= -1,
+				       .device 		= device,
+				       .title 		= opts->xterm_title,
+				       .raw  		= opts->raw } );
+	return data;
+}
+
+/* Only changed by xterm_setup, which is a setup */
+static char *terminal_emulator = CONFIG_XTERM_CHAN_DEFAULT_EMULATOR;
+static char *title_switch = "-T";
+static char *exec_switch = "-e";
+
+static int __init xterm_setup(char *line, int *add)
+{
+	*add = 0;
+	terminal_emulator = line;
+
+	line = strchr(line, ',');
+	if (line == NULL)
+		return 0;
+
+	*line++ = '\0';
+	if (*line)
+		title_switch = line;
+
+	line = strchr(line, ',');
+	if (line == NULL)
+		return 0;
+
+	*line++ = '\0';
+	if (*line)
+		exec_switch = line;
+
+	return 0;
+}
+
+__uml_setup("xterm=", xterm_setup,
+"xterm=<terminal emulator>,<title switch>,<exec switch>\n"
+"    Specifies an alternate terminal emulator to use for the debugger,\n"
+"    consoles, and serial lines when they are attached to the xterm channel.\n"
+"    The values are the terminal emulator binary, the switch it uses to set\n"
+"    its title, and the switch it uses to execute a subprocess,\n"
+"    respectively.  The title switch must have the form '<switch> title',\n"
+"    not '<switch>=title'.  Similarly, the exec switch must have the form\n"
+"    '<switch> command arg1 arg2 ...'.\n"
+"    The default values are 'xterm=" CONFIG_XTERM_CHAN_DEFAULT_EMULATOR
+     ",-T,-e'.\n"
+"    Values for gnome-terminal are 'xterm=gnome-terminal,-t,--'.\n\n"
+);
+
+static int xterm_open(int input, int output, int primary, void *d,
+		      char **dev_out)
+{
+	struct xterm_chan *data = d;
+	int pid, fd, new, err;
+	char title[256], file[] = "/tmp/xterm-pipeXXXXXX";
+	char *argv[] = { terminal_emulator, title_switch, title, exec_switch,
+			 OS_LIB_PATH "/uml/port-helper", "-uml-socket",
+			 file, NULL };
+
+	if (access(argv[4], X_OK) < 0)
+		argv[4] = "port-helper";
+
+	/* Ensure we are running on Xorg or Wayland. */
+	if (!getenv("DISPLAY") && !getenv("WAYLAND_DISPLAY")) {
+		printk(UM_KERN_ERR "xterm_open : neither $DISPLAY nor $WAYLAND_DISPLAY is set.\n");
+		return -ENODEV;
+	}
+
+	/*
+	 * This business of getting a descriptor to a temp file,
+	 * deleting the file and closing the descriptor is just to get
+	 * a known-unused name for the Unix socket that we really
+	 * want.
+	 */
+	fd = mkstemp(file);
+	if (fd < 0) {
+		err = -errno;
+		printk(UM_KERN_ERR "xterm_open : mkstemp failed, errno = %d\n",
+		       errno);
+		return err;
+	}
+
+	if (unlink(file)) {
+		err = -errno;
+		printk(UM_KERN_ERR "xterm_open : unlink failed, errno = %d\n",
+		       errno);
+		close(fd);
+		return err;
+	}
+	close(fd);
+
+	fd = os_create_unix_socket(file, sizeof(file), 1);
+	if (fd < 0) {
+		printk(UM_KERN_ERR "xterm_open : create_unix_socket failed, "
+		       "errno = %d\n", -fd);
+		return fd;
+	}
+
+	sprintf(title, data->title, data->device);
+	pid = run_helper(NULL, NULL, argv);
+	if (pid < 0) {
+		err = pid;
+		printk(UM_KERN_ERR "xterm_open : run_helper failed, "
+		       "errno = %d\n", -err);
+		goto out_close1;
+	}
+
+	err = os_set_fd_block(fd, 0);
+	if (err < 0) {
+		printk(UM_KERN_ERR "xterm_open : failed to set descriptor "
+		       "non-blocking, err = %d\n", -err);
+		goto out_kill;
+	}
+
+	data->chan_fd = fd;
+	new = xterm_fd(fd, &data->helper_pid);
+	if (new < 0) {
+		err = new;
+		printk(UM_KERN_ERR "xterm_open : xterm_fd failed, err = %d\n",
+		       -err);
+		goto out_kill;
+	}
+
+	err = os_set_fd_block(new, 0);
+	if (err) {
+		printk(UM_KERN_ERR "xterm_open : failed to set xterm "
+		       "descriptor non-blocking, err = %d\n", -err);
+		goto out_close2;
+	}
+
+	CATCH_EINTR(err = tcgetattr(new, &data->tt));
+	if (err) {
+		new = err;
+		goto out_close2;
+	}
+
+	if (data->raw) {
+		err = raw(new);
+		if (err) {
+			new = err;
+			goto out_close2;
+		}
+	}
+
+	unlink(file);
+	data->pid = pid;
+	*dev_out = NULL;
+
+	return new;
+
+ out_close2:
+	close(new);
+ out_kill:
+	os_kill_process(pid, 1);
+ out_close1:
+	close(fd);
+
+	return err;
+}
+
+static void xterm_close(int fd, void *d)
+{
+	struct xterm_chan *data = d;
+
+	if (data->pid != -1)
+		os_kill_process(data->pid, 1);
+	data->pid = -1;
+
+	if (data->helper_pid != -1)
+		os_kill_process(data->helper_pid, 0);
+	data->helper_pid = -1;
+
+	if (data->chan_fd != -1)
+		os_close_file(data->chan_fd);
+	os_close_file(fd);
+}
+
+const struct chan_ops xterm_ops = {
+	.type		= "xterm",
+	.init		= xterm_init,
+	.open		= xterm_open,
+	.close		= xterm_close,
+	.read		= generic_read,
+	.write		= generic_write,
+	.console_write	= generic_console_write,
+	.window_size	= generic_window_size,
+	.free		= generic_free,
+	.winch		= 1,
+};