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, 514 insertions, 0 deletions

diff --git a/drivers/base/devtmpfs.c b/drivers/base/devtmpfs.c
new file mode 100644
index 000000000..aef0fcc6a
--- /dev/null
+++ b/drivers/base/devtmpfs.c
@@ -0,0 +1,514 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * devtmpfs - kernel-maintained tmpfs-based /dev
+ *
+ * Copyright (C) 2009, Kay Sievers <kay.sievers@vrfy.org>
+ *
+ * During bootup, before any driver core device is registered,
+ * devtmpfs, a tmpfs-based filesystem is created. Every driver-core
+ * device which requests a device node, will add a node in this
+ * filesystem.
+ * By default, all devices are named after the name of the device,
+ * owned by root and have a default mode of 0600. Subsystems can
+ * overwrite the default setting if needed.
+ */
+
+#define pr_fmt(fmt) "devtmpfs: " fmt
+
+#include <linux/kernel.h>
+#include <linux/syscalls.h>
+#include <linux/mount.h>
+#include <linux/device.h>
+#include <linux/blkdev.h>
+#include <linux/namei.h>
+#include <linux/fs.h>
+#include <linux/shmem_fs.h>
+#include <linux/ramfs.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/kthread.h>
+#include <linux/init_syscalls.h>
+#include <uapi/linux/mount.h>
+#include "base.h"
+
+#ifdef CONFIG_DEVTMPFS_SAFE
+#define DEVTMPFS_MFLAGS       (MS_SILENT | MS_NOEXEC | MS_NOSUID)
+#else
+#define DEVTMPFS_MFLAGS       (MS_SILENT)
+#endif
+
+static struct task_struct *thread;
+
+static int __initdata mount_dev = IS_ENABLED(CONFIG_DEVTMPFS_MOUNT);
+
+static DEFINE_SPINLOCK(req_lock);
+
+static struct req {
+	struct req *next;
+	struct completion done;
+	int err;
+	const char *name;
+	umode_t mode;	/* 0 => delete */
+	kuid_t uid;
+	kgid_t gid;
+	struct device *dev;
+} *requests;
+
+static int __init mount_param(char *str)
+{
+	return kstrtoint(str, 0, &mount_dev) == 0;
+}
+__setup("devtmpfs.mount=", mount_param);
+
+static struct vfsmount *mnt;
+
+static struct file_system_type internal_fs_type = {
+	.name = "devtmpfs",
+#ifdef CONFIG_TMPFS
+	.init_fs_context = shmem_init_fs_context,
+#else
+	.init_fs_context = ramfs_init_fs_context,
+#endif
+	.kill_sb = kill_anon_super,
+};
+
+/* Simply take a ref on the existing mount */
+static int devtmpfs_get_tree(struct fs_context *fc)
+{
+	struct super_block *sb = mnt->mnt_sb;
+
+	atomic_inc(&sb->s_active);
+	down_write(&sb->s_umount);
+	fc->root = dget(sb->s_root);
+	return 0;
+}
+
+/* Ops are filled in during init depending on underlying shmem or ramfs type */
+static struct fs_context_operations devtmpfs_context_ops = {};
+
+/* Call the underlying initialization and set to our ops */
+static int devtmpfs_init_fs_context(struct fs_context *fc)
+{
+	int ret;
+#ifdef CONFIG_TMPFS
+	ret = shmem_init_fs_context(fc);
+#else
+	ret = ramfs_init_fs_context(fc);
+#endif
+	if (ret < 0)
+		return ret;
+
+	fc->ops = &devtmpfs_context_ops;
+
+	return 0;
+}
+
+static struct file_system_type dev_fs_type = {
+	.name = "devtmpfs",
+	.init_fs_context = devtmpfs_init_fs_context,
+};
+
+static int devtmpfs_submit_req(struct req *req, const char *tmp)
+{
+	init_completion(&req->done);
+
+	spin_lock(&req_lock);
+	req->next = requests;
+	requests = req;
+	spin_unlock(&req_lock);
+
+	wake_up_process(thread);
+	wait_for_completion(&req->done);
+
+	kfree(tmp);
+
+	return req->err;
+}
+
+int devtmpfs_create_node(struct device *dev)
+{
+	const char *tmp = NULL;
+	struct req req;
+
+	if (!thread)
+		return 0;
+
+	req.mode = 0;
+	req.uid = GLOBAL_ROOT_UID;
+	req.gid = GLOBAL_ROOT_GID;
+	req.name = device_get_devnode(dev, &req.mode, &req.uid, &req.gid, &tmp);
+	if (!req.name)
+		return -ENOMEM;
+
+	if (req.mode == 0)
+		req.mode = 0600;
+	if (is_blockdev(dev))
+		req.mode |= S_IFBLK;
+	else
+		req.mode |= S_IFCHR;
+
+	req.dev = dev;
+
+	return devtmpfs_submit_req(&req, tmp);
+}
+
+int devtmpfs_delete_node(struct device *dev)
+{
+	const char *tmp = NULL;
+	struct req req;
+
+	if (!thread)
+		return 0;
+
+	req.name = device_get_devnode(dev, NULL, NULL, NULL, &tmp);
+	if (!req.name)
+		return -ENOMEM;
+
+	req.mode = 0;
+	req.dev = dev;
+
+	return devtmpfs_submit_req(&req, tmp);
+}
+
+static int dev_mkdir(const char *name, umode_t mode)
+{
+	struct dentry *dentry;
+	struct path path;
+
+	dentry = start_creating_path(AT_FDCWD, name, &path, LOOKUP_DIRECTORY);
+	if (IS_ERR(dentry))
+		return PTR_ERR(dentry);
+
+	dentry = vfs_mkdir(&nop_mnt_idmap, d_inode(path.dentry), dentry, mode, NULL);
+	if (!IS_ERR(dentry))
+		/* mark as kernel-created inode */
+		d_inode(dentry)->i_private = &thread;
+	end_creating_path(&path, dentry);
+	return PTR_ERR_OR_ZERO(dentry);
+}
+
+static int create_path(const char *nodepath)
+{
+	char *path;
+	char *s;
+	int err = 0;
+
+	/* parent directories do not exist, create them */
+	path = kstrdup(nodepath, GFP_KERNEL);
+	if (!path)
+		return -ENOMEM;
+
+	s = path;
+	for (;;) {
+		s = strchr(s, '/');
+		if (!s)
+			break;
+		s[0] = '\0';
+		err = dev_mkdir(path, 0755);
+		if (err && err != -EEXIST)
+			break;
+		s[0] = '/';
+		s++;
+	}
+	kfree(path);
+	return err;
+}
+
+static int handle_create(const char *nodename, umode_t mode, kuid_t uid,
+			 kgid_t gid, struct device *dev)
+{
+	struct dentry *dentry;
+	struct path path;
+	int err;
+
+	dentry = start_creating_path(AT_FDCWD, nodename, &path, 0);
+	if (dentry == ERR_PTR(-ENOENT)) {
+		create_path(nodename);
+		dentry = start_creating_path(AT_FDCWD, nodename, &path, 0);
+	}
+	if (IS_ERR(dentry))
+		return PTR_ERR(dentry);
+
+	err = vfs_mknod(&nop_mnt_idmap, d_inode(path.dentry), dentry, mode,
+			dev->devt, NULL);
+	if (!err) {
+		struct iattr newattrs;
+
+		newattrs.ia_mode = mode;
+		newattrs.ia_uid = uid;
+		newattrs.ia_gid = gid;
+		newattrs.ia_valid = ATTR_MODE|ATTR_UID|ATTR_GID;
+		inode_lock(d_inode(dentry));
+		notify_change(&nop_mnt_idmap, dentry, &newattrs, NULL);
+		inode_unlock(d_inode(dentry));
+
+		/* mark as kernel-created inode */
+		d_inode(dentry)->i_private = &thread;
+	}
+	end_creating_path(&path, dentry);
+	return err;
+}
+
+static int dev_rmdir(const char *name)
+{
+	struct path parent;
+	struct dentry *dentry;
+	int err;
+
+	dentry = start_removing_path(name, &parent);
+	if (IS_ERR(dentry))
+		return PTR_ERR(dentry);
+	if (d_inode(dentry)->i_private == &thread)
+		err = vfs_rmdir(&nop_mnt_idmap, d_inode(parent.dentry),
+				dentry, NULL);
+	else
+		err = -EPERM;
+
+	end_removing_path(&parent, dentry);
+	return err;
+}
+
+static int delete_path(const char *nodepath)
+{
+	char *path;
+	int err = 0;
+
+	path = kstrdup(nodepath, GFP_KERNEL);
+	if (!path)
+		return -ENOMEM;
+
+	for (;;) {
+		char *base;
+
+		base = strrchr(path, '/');
+		if (!base)
+			break;
+		base[0] = '\0';
+		err = dev_rmdir(path);
+		if (err)
+			break;
+	}
+
+	kfree(path);
+	return err;
+}
+
+static int dev_mynode(struct device *dev, struct inode *inode)
+{
+	/* did we create it */
+	if (inode->i_private != &thread)
+		return 0;
+
+	/* does the dev_t match */
+	if (is_blockdev(dev)) {
+		if (!S_ISBLK(inode->i_mode))
+			return 0;
+	} else {
+		if (!S_ISCHR(inode->i_mode))
+			return 0;
+	}
+	if (inode->i_rdev != dev->devt)
+		return 0;
+
+	/* ours */
+	return 1;
+}
+
+static int handle_remove(const char *nodename, struct device *dev)
+{
+	struct path parent;
+	struct dentry *dentry;
+	struct inode *inode;
+	int deleted = 0;
+	int err = 0;
+
+	dentry = start_removing_path(nodename, &parent);
+	if (IS_ERR(dentry))
+		return PTR_ERR(dentry);
+
+	inode = d_inode(dentry);
+	if (dev_mynode(dev, inode)) {
+		struct iattr newattrs;
+		/*
+		 * before unlinking this node, reset permissions
+		 * of possible references like hardlinks
+		 */
+		newattrs.ia_uid = GLOBAL_ROOT_UID;
+		newattrs.ia_gid = GLOBAL_ROOT_GID;
+		newattrs.ia_mode = inode->i_mode & ~0777;
+		newattrs.ia_valid =
+			ATTR_UID|ATTR_GID|ATTR_MODE;
+		inode_lock(d_inode(dentry));
+		notify_change(&nop_mnt_idmap, dentry, &newattrs, NULL);
+		inode_unlock(d_inode(dentry));
+		err = vfs_unlink(&nop_mnt_idmap, d_inode(parent.dentry),
+				 dentry, NULL);
+		if (!err || err == -ENOENT)
+			deleted = 1;
+	}
+	end_removing_path(&parent, dentry);
+
+	if (deleted && strchr(nodename, '/'))
+		delete_path(nodename);
+	return err;
+}
+
+/*
+ * If configured, or requested by the commandline, devtmpfs will be
+ * auto-mounted after the kernel mounted the root filesystem.
+ */
+int __init devtmpfs_mount(void)
+{
+	int err;
+
+	if (!mount_dev)
+		return 0;
+
+	if (!thread)
+		return 0;
+
+	err = init_mount("devtmpfs", "dev", "devtmpfs", DEVTMPFS_MFLAGS, NULL);
+	if (err)
+		pr_info("error mounting %d\n", err);
+	else
+		pr_info("mounted\n");
+	return err;
+}
+
+static __initdata DECLARE_COMPLETION(setup_done);
+
+static int handle(const char *name, umode_t mode, kuid_t uid, kgid_t gid,
+		  struct device *dev)
+{
+	if (mode)
+		return handle_create(name, mode, uid, gid, dev);
+	else
+		return handle_remove(name, dev);
+}
+
+static void __noreturn devtmpfs_work_loop(void)
+{
+	while (1) {
+		spin_lock(&req_lock);
+		while (requests) {
+			struct req *req = requests;
+			requests = NULL;
+			spin_unlock(&req_lock);
+			while (req) {
+				struct req *next = req->next;
+				req->err = handle(req->name, req->mode,
+						  req->uid, req->gid, req->dev);
+				complete(&req->done);
+				req = next;
+			}
+			spin_lock(&req_lock);
+		}
+		__set_current_state(TASK_INTERRUPTIBLE);
+		spin_unlock(&req_lock);
+		schedule();
+	}
+}
+
+static noinline int __init devtmpfs_setup(void *p)
+{
+	int err;
+
+	err = ksys_unshare(UNSHARE_EMPTY_MNTNS);
+	if (err)
+		goto out;
+	err = init_mount("devtmpfs", "/", "devtmpfs", DEVTMPFS_MFLAGS, NULL);
+	if (err)
+		goto out;
+	init_chdir("/.."); /* will traverse into overmounted root */
+	init_chroot(".");
+out:
+	*(int *)p = err;
+	return err;
+}
+
+/*
+ * The __ref is because devtmpfs_setup needs to be __init for the routines it
+ * calls.  That call is done while devtmpfs_init, which is marked __init,
+ * synchronously waits for it to complete.
+ */
+static int __ref devtmpfsd(void *p)
+{
+	int err = devtmpfs_setup(p);
+
+	complete(&setup_done);
+	if (err)
+		return err;
+	devtmpfs_work_loop();
+	return 0;
+}
+
+/*
+ * Get the underlying (shmem/ramfs) context ops to build ours
+ */
+static int devtmpfs_configure_context(void)
+{
+	struct fs_context *fc;
+
+	fc = fs_context_for_reconfigure(mnt->mnt_root, mnt->mnt_sb->s_flags,
+					MS_RMT_MASK);
+	if (IS_ERR(fc))
+		return PTR_ERR(fc);
+
+	/* Set up devtmpfs_context_ops based on underlying type */
+	devtmpfs_context_ops.free	      = fc->ops->free;
+	devtmpfs_context_ops.dup	      = fc->ops->dup;
+	devtmpfs_context_ops.parse_param      = fc->ops->parse_param;
+	devtmpfs_context_ops.parse_monolithic = fc->ops->parse_monolithic;
+	devtmpfs_context_ops.get_tree	      = &devtmpfs_get_tree;
+	devtmpfs_context_ops.reconfigure      = fc->ops->reconfigure;
+
+	put_fs_context(fc);
+
+	return 0;
+}
+
+/*
+ * Create devtmpfs instance, driver-core devices will add their device
+ * nodes here.
+ */
+int __init devtmpfs_init(void)
+{
+	char opts[] = "mode=0755";
+	int err;
+
+	mnt = vfs_kern_mount(&internal_fs_type, 0, "devtmpfs", opts);
+	if (IS_ERR(mnt)) {
+		pr_err("unable to create devtmpfs %ld\n", PTR_ERR(mnt));
+		return PTR_ERR(mnt);
+	}
+
+	err = devtmpfs_configure_context();
+	if (err) {
+		pr_err("unable to configure devtmpfs type %d\n", err);
+		return err;
+	}
+
+	err = register_filesystem(&dev_fs_type);
+	if (err) {
+		pr_err("unable to register devtmpfs type %d\n", err);
+		return err;
+	}
+
+	thread = kthread_run(devtmpfsd, &err, "kdevtmpfs");
+	if (!IS_ERR(thread)) {
+		wait_for_completion(&setup_done);
+	} else {
+		err = PTR_ERR(thread);
+		thread = NULL;
+	}
+
+	if (err) {
+		pr_err("unable to create devtmpfs %d\n", err);
+		unregister_filesystem(&dev_fs_type);
+		thread = NULL;
+		return err;
+	}
+
+	pr_info("initialized\n");
+	return 0;
+}