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, 256 insertions, 0 deletions

diff --git a/fs/udf/file.c b/fs/udf/file.c
new file mode 100644
index 000000000..57d11606a
--- /dev/null
+++ b/fs/udf/file.c
@@ -0,0 +1,256 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * file.c
+ *
+ * PURPOSE
+ *  File handling routines for the OSTA-UDF(tm) filesystem.
+ *
+ * COPYRIGHT
+ *  (C) 1998-1999 Dave Boynton
+ *  (C) 1998-2004 Ben Fennema
+ *  (C) 1999-2000 Stelias Computing Inc
+ *
+ * HISTORY
+ *
+ *  10/02/98 dgb  Attempt to integrate into udf.o
+ *  10/07/98      Switched to using generic_readpage, etc., like isofs
+ *                And it works!
+ *  12/06/98 blf  Added udf_file_read. uses generic_file_read for all cases but
+ *                ICBTAG_FLAG_AD_IN_ICB.
+ *  04/06/99      64 bit file handling on 32 bit systems taken from ext2 file.c
+ *  05/12/99      Preliminary file write support
+ */
+
+#include "udfdecl.h"
+#include <linux/fs.h>
+#include <linux/uaccess.h>
+#include <linux/kernel.h>
+#include <linux/string.h> /* memset */
+#include <linux/capability.h>
+#include <linux/errno.h>
+#include <linux/filelock.h>
+#include <linux/pagemap.h>
+#include <linux/uio.h>
+
+#include "udf_i.h"
+#include "udf_sb.h"
+
+static vm_fault_t udf_page_mkwrite(struct vm_fault *vmf)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	struct inode *inode = file_inode(vma->vm_file);
+	struct address_space *mapping = inode->i_mapping;
+	struct folio *folio = page_folio(vmf->page);
+	loff_t size;
+	unsigned int end;
+	vm_fault_t ret = VM_FAULT_LOCKED;
+	int err;
+
+	sb_start_pagefault(inode->i_sb);
+	file_update_time(vma->vm_file);
+	filemap_invalidate_lock_shared(mapping);
+	folio_lock(folio);
+	size = i_size_read(inode);
+	if (folio->mapping != inode->i_mapping || folio_pos(folio) >= size) {
+		folio_unlock(folio);
+		ret = VM_FAULT_NOPAGE;
+		goto out_unlock;
+	}
+	/* Space is already allocated for in-ICB file */
+	if (UDF_I(inode)->i_alloc_type == ICBTAG_FLAG_AD_IN_ICB)
+		goto out_dirty;
+	if (folio->index == size >> PAGE_SHIFT)
+		end = size & ~PAGE_MASK;
+	else
+		end = PAGE_SIZE;
+	err = __block_write_begin(folio, 0, end, udf_get_block);
+	if (err) {
+		folio_unlock(folio);
+		ret = vmf_fs_error(err);
+		goto out_unlock;
+	}
+
+	block_commit_write(folio, 0, end);
+out_dirty:
+	folio_mark_dirty(folio);
+	folio_wait_stable(folio);
+out_unlock:
+	filemap_invalidate_unlock_shared(mapping);
+	sb_end_pagefault(inode->i_sb);
+	return ret;
+}
+
+static const struct vm_operations_struct udf_file_vm_ops = {
+	.fault		= filemap_fault,
+	.map_pages	= filemap_map_pages,
+	.page_mkwrite	= udf_page_mkwrite,
+};
+
+static ssize_t udf_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
+{
+	ssize_t retval;
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file_inode(file);
+	struct udf_inode_info *iinfo = UDF_I(inode);
+
+	inode_lock(inode);
+
+	retval = generic_write_checks(iocb, from);
+	if (retval <= 0)
+		goto out;
+
+	if (iinfo->i_alloc_type == ICBTAG_FLAG_AD_IN_ICB &&
+	    inode->i_sb->s_blocksize < (udf_file_entry_alloc_offset(inode) +
+				 iocb->ki_pos + iov_iter_count(from))) {
+		filemap_invalidate_lock(inode->i_mapping);
+		retval = udf_expand_file_adinicb(inode);
+		filemap_invalidate_unlock(inode->i_mapping);
+		if (retval)
+			goto out;
+	}
+
+	retval = __generic_file_write_iter(iocb, from);
+out:
+	if (iinfo->i_alloc_type == ICBTAG_FLAG_AD_IN_ICB && retval > 0) {
+		down_write(&iinfo->i_data_sem);
+		iinfo->i_lenAlloc = inode->i_size;
+		up_write(&iinfo->i_data_sem);
+	}
+	inode_unlock(inode);
+
+	if (retval > 0) {
+		mark_inode_dirty(inode);
+		retval = generic_write_sync(iocb, retval);
+	}
+
+	return retval;
+}
+
+long udf_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	struct inode *inode = file_inode(filp);
+	long old_block, new_block;
+	int result;
+
+	if (file_permission(filp, MAY_READ) != 0) {
+		udf_debug("no permission to access inode %llu\n", inode->i_ino);
+		return -EPERM;
+	}
+
+	if (!arg && ((cmd == UDF_GETVOLIDENT) || (cmd == UDF_GETEASIZE) ||
+		     (cmd == UDF_RELOCATE_BLOCKS) || (cmd == UDF_GETEABLOCK))) {
+		udf_debug("invalid argument to udf_ioctl\n");
+		return -EINVAL;
+	}
+
+	switch (cmd) {
+	case UDF_GETVOLIDENT:
+		if (copy_to_user((char __user *)arg,
+				 UDF_SB(inode->i_sb)->s_volume_ident, 32))
+			return -EFAULT;
+		return 0;
+	case UDF_RELOCATE_BLOCKS:
+		if (!capable(CAP_SYS_ADMIN))
+			return -EPERM;
+		if (get_user(old_block, (long __user *)arg))
+			return -EFAULT;
+		result = udf_relocate_blocks(inode->i_sb,
+						old_block, &new_block);
+		if (result == 0)
+			result = put_user(new_block, (long __user *)arg);
+		return result;
+	case UDF_GETEASIZE:
+		return put_user(UDF_I(inode)->i_lenEAttr, (int __user *)arg);
+	case UDF_GETEABLOCK:
+		return copy_to_user((char __user *)arg,
+				    UDF_I(inode)->i_data,
+				    UDF_I(inode)->i_lenEAttr) ? -EFAULT : 0;
+	default:
+		return -ENOIOCTLCMD;
+	}
+
+	return 0;
+}
+
+static int udf_release_file(struct inode *inode, struct file *filp)
+{
+	if (filp->f_mode & FMODE_WRITE &&
+	    atomic_read(&inode->i_writecount) == 1) {
+		/*
+		 * Grab i_mutex to avoid races with writes changing i_size
+		 * while we are running.
+		 */
+		inode_lock(inode);
+		down_write(&UDF_I(inode)->i_data_sem);
+		udf_discard_prealloc(inode);
+		udf_truncate_tail_extent(inode);
+		up_write(&UDF_I(inode)->i_data_sem);
+		inode_unlock(inode);
+	}
+	return 0;
+}
+
+static int udf_file_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	file_accessed(file);
+	vma->vm_ops = &udf_file_vm_ops;
+
+	return 0;
+}
+
+const struct file_operations udf_file_operations = {
+	.read_iter		= generic_file_read_iter,
+	.unlocked_ioctl		= udf_ioctl,
+	.open			= generic_file_open,
+	.mmap			= udf_file_mmap,
+	.write_iter		= udf_file_write_iter,
+	.release		= udf_release_file,
+	.fsync			= simple_fsync,
+	.splice_read		= filemap_splice_read,
+	.splice_write		= iter_file_splice_write,
+	.llseek			= generic_file_llseek,
+	.setlease		= generic_setlease,
+};
+
+static int udf_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
+		       struct iattr *attr)
+{
+	struct inode *inode = d_inode(dentry);
+	struct super_block *sb = inode->i_sb;
+	int error;
+
+	error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
+	if (error)
+		return error;
+
+	if ((attr->ia_valid & ATTR_UID) &&
+	    UDF_QUERY_FLAG(sb, UDF_FLAG_UID_SET) &&
+	    !uid_eq(attr->ia_uid, UDF_SB(sb)->s_uid))
+		return -EPERM;
+	if ((attr->ia_valid & ATTR_GID) &&
+	    UDF_QUERY_FLAG(sb, UDF_FLAG_GID_SET) &&
+	    !gid_eq(attr->ia_gid, UDF_SB(sb)->s_gid))
+		return -EPERM;
+
+	if ((attr->ia_valid & ATTR_SIZE) &&
+	    attr->ia_size != i_size_read(inode)) {
+		filemap_invalidate_lock(inode->i_mapping);
+		error = udf_setsize(inode, attr->ia_size);
+		filemap_invalidate_unlock(inode->i_mapping);
+		if (error)
+			return error;
+	}
+
+	if (attr->ia_valid & ATTR_MODE)
+		udf_update_extra_perms(inode, attr->ia_mode);
+
+	setattr_copy(&nop_mnt_idmap, inode, attr);
+	mark_inode_dirty(inode);
+	if (IS_SYNC(inode))
+		sync_inode_metadata(inode, 1);
+	return 0;
+}
+
+const struct inode_operations udf_file_inode_operations = {
+	.setattr		= udf_setattr,
+};