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, 475 insertions, 0 deletions

diff --git a/fs/openpromfs/inode.c b/fs/openpromfs/inode.c
new file mode 100644
index 000000000..fb8d84bde
--- /dev/null
+++ b/fs/openpromfs/inode.c
@@ -0,0 +1,475 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* inode.c: /proc/openprom handling routines
+ *
+ * Copyright (C) 1996-1999 Jakub Jelinek  (jakub@redhat.com)
+ * Copyright (C) 1998      Eddie C. Dost  (ecd@skynet.be)
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/fs.h>
+#include <linux/fs_context.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/seq_file.h>
+#include <linux/magic.h>
+
+#include <asm/openprom.h>
+#include <asm/oplib.h>
+#include <asm/prom.h>
+#include <linux/uaccess.h>
+
+static DEFINE_MUTEX(op_mutex);
+
+#define OPENPROM_ROOT_INO	0
+
+enum op_inode_type {
+	op_inode_node,
+	op_inode_prop,
+};
+
+union op_inode_data {
+	struct device_node	*node;
+	struct property		*prop;
+};
+
+struct op_inode_info {
+	struct inode		vfs_inode;
+	enum op_inode_type	type;
+	union op_inode_data	u;
+};
+
+static struct inode *openprom_iget(struct super_block *sb, ino_t ino);
+
+static inline struct op_inode_info *OP_I(struct inode *inode)
+{
+	return container_of(inode, struct op_inode_info, vfs_inode);
+}
+
+static int is_string(unsigned char *p, int len)
+{
+	int i;
+
+	for (i = 0; i < len; i++) {
+		unsigned char val = p[i];
+
+		if ((i && !val) ||
+		    (val >= ' ' && val <= '~'))
+			continue;
+
+		return 0;
+	}
+
+	return 1;
+}
+
+static int property_show(struct seq_file *f, void *v)
+{
+	struct property *prop = f->private;
+	void *pval;
+	int len;
+
+	len = prop->length;
+	pval = prop->value;
+
+	if (is_string(pval, len)) {
+		while (len > 0) {
+			int n = strlen(pval);
+
+			seq_printf(f, "%s", (char *) pval);
+
+			/* Skip over the NULL byte too.  */
+			pval += n + 1;
+			len -= n + 1;
+
+			if (len > 0)
+				seq_printf(f, " + ");
+		}
+	} else {
+		if (len & 3) {
+			while (len) {
+				len--;
+				if (len)
+					seq_printf(f, "%02x.",
+						   *(unsigned char *) pval);
+				else
+					seq_printf(f, "%02x",
+						   *(unsigned char *) pval);
+				pval++;
+			}
+		} else {
+			while (len >= 4) {
+				len -= 4;
+
+				if (len)
+					seq_printf(f, "%08x.",
+						   *(unsigned int *) pval);
+				else
+					seq_printf(f, "%08x",
+						   *(unsigned int *) pval);
+				pval += 4;
+			}
+		}
+	}
+	seq_printf(f, "\n");
+
+	return 0;
+}
+
+static void *property_start(struct seq_file *f, loff_t *pos)
+{
+	if (*pos == 0)
+		return pos;
+	return NULL;
+}
+
+static void *property_next(struct seq_file *f, void *v, loff_t *pos)
+{
+	(*pos)++;
+	return NULL;
+}
+
+static void property_stop(struct seq_file *f, void *v)
+{
+	/* Nothing to do */
+}
+
+static const struct seq_operations property_op = {
+	.start		= property_start,
+	.next		= property_next,
+	.stop		= property_stop,
+	.show		= property_show
+};
+
+static int property_open(struct inode *inode, struct file *file)
+{
+	struct op_inode_info *oi = OP_I(inode);
+	int ret;
+
+	BUG_ON(oi->type != op_inode_prop);
+
+	ret = seq_open(file, &property_op);
+	if (!ret) {
+		struct seq_file *m = file->private_data;
+		m->private = oi->u.prop;
+	}
+	return ret;
+}
+
+static const struct file_operations openpromfs_prop_ops = {
+	.open		= property_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static int openpromfs_readdir(struct file *, struct dir_context *);
+
+static const struct file_operations openprom_operations = {
+	.read		= generic_read_dir,
+	.iterate_shared	= openpromfs_readdir,
+	.llseek		= generic_file_llseek,
+};
+
+static struct dentry *openpromfs_lookup(struct inode *, struct dentry *, unsigned int);
+
+static const struct inode_operations openprom_inode_operations = {
+	.lookup		= openpromfs_lookup,
+};
+
+static struct dentry *openpromfs_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
+{
+	struct op_inode_info *ent_oi, *oi = OP_I(dir);
+	struct device_node *dp, *child;
+	struct property *prop;
+	enum op_inode_type ent_type;
+	union op_inode_data ent_data;
+	const char *name;
+	struct inode *inode;
+	unsigned int ino;
+	int len;
+	
+	BUG_ON(oi->type != op_inode_node);
+
+	dp = oi->u.node;
+
+	name = dentry->d_name.name;
+	len = dentry->d_name.len;
+
+	mutex_lock(&op_mutex);
+
+	child = dp->child;
+	while (child) {
+		const char *node_name = kbasename(child->full_name);
+		int n = strlen(node_name);
+
+		if (len == n &&
+		    !strncmp(node_name, name, len)) {
+			ent_type = op_inode_node;
+			ent_data.node = child;
+			ino = child->unique_id;
+			goto found;
+		}
+		child = child->sibling;
+	}
+
+	prop = dp->properties;
+	while (prop) {
+		int n = strlen(prop->name);
+
+		if (len == n && !strncmp(prop->name, name, len)) {
+			ent_type = op_inode_prop;
+			ent_data.prop = prop;
+			ino = prop->unique_id;
+			goto found;
+		}
+
+		prop = prop->next;
+	}
+
+	mutex_unlock(&op_mutex);
+	return ERR_PTR(-ENOENT);
+
+found:
+	inode = openprom_iget(dir->i_sb, ino);
+	mutex_unlock(&op_mutex);
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
+	if (inode_state_read_once(inode) & I_NEW) {
+		simple_inode_init_ts(inode);
+		ent_oi = OP_I(inode);
+		ent_oi->type = ent_type;
+		ent_oi->u = ent_data;
+
+		switch (ent_type) {
+		case op_inode_node:
+			inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
+			inode->i_op = &openprom_inode_operations;
+			inode->i_fop = &openprom_operations;
+			set_nlink(inode, 2);
+			break;
+		case op_inode_prop:
+			if (of_node_name_eq(dp, "options") && (len == 17) &&
+			    !strncmp (name, "security-password", 17))
+				inode->i_mode = S_IFREG | S_IRUSR | S_IWUSR;
+			else
+				inode->i_mode = S_IFREG | S_IRUGO;
+			inode->i_fop = &openpromfs_prop_ops;
+			set_nlink(inode, 1);
+			inode->i_size = ent_oi->u.prop->length;
+			break;
+		}
+		unlock_new_inode(inode);
+	}
+
+	return d_splice_alias(inode, dentry);
+}
+
+static int openpromfs_readdir(struct file *file, struct dir_context *ctx)
+{
+	struct inode *inode = file_inode(file);
+	struct op_inode_info *oi = OP_I(inode);
+	struct device_node *dp = oi->u.node;
+	struct device_node *child;
+	struct property *prop;
+	int i;
+
+	mutex_lock(&op_mutex);
+	
+	if (ctx->pos == 0) {
+		if (!dir_emit(ctx, ".", 1, inode->i_ino, DT_DIR))
+			goto out;
+		ctx->pos = 1;
+	}
+	if (ctx->pos == 1) {
+		if (!dir_emit(ctx, "..", 2,
+			    (dp->parent == NULL ?
+			     OPENPROM_ROOT_INO :
+			     dp->parent->unique_id), DT_DIR))
+			goto out;
+		ctx->pos = 2;
+	}
+	i = ctx->pos - 2;
+
+	/* First, the children nodes as directories.  */
+	child = dp->child;
+	while (i && child) {
+		child = child->sibling;
+		i--;
+	}
+	while (child) {
+		if (!dir_emit(ctx,
+			    kbasename(child->full_name),
+			    strlen(kbasename(child->full_name)),
+			    child->unique_id, DT_DIR))
+			goto out;
+
+		ctx->pos++;
+		child = child->sibling;
+	}
+
+	/* Next, the properties as files.  */
+	prop = dp->properties;
+	while (i && prop) {
+		prop = prop->next;
+		i--;
+	}
+	while (prop) {
+		if (!dir_emit(ctx, prop->name, strlen(prop->name),
+			    prop->unique_id, DT_REG))
+			goto out;
+
+		ctx->pos++;
+		prop = prop->next;
+	}
+
+out:
+	mutex_unlock(&op_mutex);
+	return 0;
+}
+
+static struct kmem_cache *op_inode_cachep;
+
+static struct inode *openprom_alloc_inode(struct super_block *sb)
+{
+	struct op_inode_info *oi;
+
+	oi = alloc_inode_sb(sb, op_inode_cachep, GFP_KERNEL);
+	if (!oi)
+		return NULL;
+
+	return &oi->vfs_inode;
+}
+
+static void openprom_free_inode(struct inode *inode)
+{
+	kmem_cache_free(op_inode_cachep, OP_I(inode));
+}
+
+static struct inode *openprom_iget(struct super_block *sb, ino_t ino)
+{
+	struct inode *inode = iget_locked(sb, ino);
+	if (!inode)
+		inode = ERR_PTR(-ENOMEM);
+	return inode;
+}
+
+static int openpromfs_reconfigure(struct fs_context *fc)
+{
+	sync_filesystem(fc->root->d_sb);
+	fc->sb_flags |= SB_NOATIME;
+	return 0;
+}
+
+static const struct super_operations openprom_sops = {
+	.alloc_inode	= openprom_alloc_inode,
+	.free_inode	= openprom_free_inode,
+	.statfs		= simple_statfs,
+};
+
+static int openprom_fill_super(struct super_block *s, struct fs_context *fc)
+{
+	struct inode *root_inode;
+	struct op_inode_info *oi;
+	int ret;
+
+	s->s_flags |= SB_NOATIME;
+	s->s_blocksize = 1024;
+	s->s_blocksize_bits = 10;
+	s->s_magic = OPENPROM_SUPER_MAGIC;
+	s->s_op = &openprom_sops;
+	s->s_time_gran = 1;
+	root_inode = openprom_iget(s, OPENPROM_ROOT_INO);
+	if (IS_ERR(root_inode)) {
+		ret = PTR_ERR(root_inode);
+		goto out_no_root;
+	}
+
+	simple_inode_init_ts(root_inode);
+	root_inode->i_op = &openprom_inode_operations;
+	root_inode->i_fop = &openprom_operations;
+	root_inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
+	oi = OP_I(root_inode);
+	oi->type = op_inode_node;
+	oi->u.node = of_find_node_by_path("/");
+	unlock_new_inode(root_inode);
+
+	s->s_root = d_make_root(root_inode);
+	if (!s->s_root)
+		goto out_no_root_dentry;
+	return 0;
+
+out_no_root_dentry:
+	ret = -ENOMEM;
+out_no_root:
+	printk("openprom_fill_super: get root inode failed\n");
+	return ret;
+}
+
+static int openpromfs_get_tree(struct fs_context *fc)
+{
+	return get_tree_single(fc, openprom_fill_super);
+}
+
+static const struct fs_context_operations openpromfs_context_ops = {
+	.get_tree	= openpromfs_get_tree,
+	.reconfigure	= openpromfs_reconfigure,
+};
+
+static int openpromfs_init_fs_context(struct fs_context *fc)
+{
+	fc->ops = &openpromfs_context_ops;
+	return 0;
+}
+
+static struct file_system_type openprom_fs_type = {
+	.owner		= THIS_MODULE,
+	.name		= "openpromfs",
+	.init_fs_context = openpromfs_init_fs_context,
+	.kill_sb	= kill_anon_super,
+};
+MODULE_ALIAS_FS("openpromfs");
+
+static void op_inode_init_once(void *data)
+{
+	struct op_inode_info *oi = (struct op_inode_info *) data;
+
+	inode_init_once(&oi->vfs_inode);
+}
+
+static int __init init_openprom_fs(void)
+{
+	int err;
+
+	op_inode_cachep = kmem_cache_create("op_inode_cache",
+					    sizeof(struct op_inode_info),
+					    0,
+					    (SLAB_RECLAIM_ACCOUNT |
+					     SLAB_ACCOUNT),
+					    op_inode_init_once);
+	if (!op_inode_cachep)
+		return -ENOMEM;
+
+	err = register_filesystem(&openprom_fs_type);
+	if (err)
+		kmem_cache_destroy(op_inode_cachep);
+
+	return err;
+}
+
+static void __exit exit_openprom_fs(void)
+{
+	unregister_filesystem(&openprom_fs_type);
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
+	kmem_cache_destroy(op_inode_cachep);
+}
+
+module_init(init_openprom_fs)
+module_exit(exit_openprom_fs)
+MODULE_DESCRIPTION("OpenPROM filesystem support");
+MODULE_LICENSE("GPL");