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, 162 insertions, 0 deletions

diff --git a/fs/udf/dir.c b/fs/udf/dir.c
new file mode 100644
index 000000000..425e3c162
--- /dev/null
+++ b/fs/udf/dir.c
@@ -0,0 +1,162 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * dir.c
+ *
+ * PURPOSE
+ *  Directory handling routines for the OSTA-UDF(tm) filesystem.
+ *
+ * COPYRIGHT
+ *  (C) 1998-2004 Ben Fennema
+ *
+ * HISTORY
+ *
+ *  10/05/98 dgb  Split directory operations into its own file
+ *                Implemented directory reads via do_udf_readdir
+ *  10/06/98      Made directory operations work!
+ *  11/17/98      Rewrote directory to support ICBTAG_FLAG_AD_LONG
+ *  11/25/98 blf  Rewrote directory handling (readdir+lookup) to support reading
+ *                across blocks.
+ *  12/12/98      Split out the lookup code to namei.c. bulk of directory
+ *                code now in directory.c:udf_fileident_read.
+ */
+
+#include "udfdecl.h"
+
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/filelock.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/bio.h>
+#include <linux/iversion.h>
+
+#include "udf_i.h"
+#include "udf_sb.h"
+
+static int udf_readdir(struct file *file, struct dir_context *ctx)
+{
+	struct inode *dir = file_inode(file);
+	loff_t nf_pos, emit_pos = 0;
+	int flen;
+	unsigned char *fname = NULL;
+	int ret = 0;
+	struct super_block *sb = dir->i_sb;
+	bool pos_valid = false;
+	struct udf_fileident_iter iter;
+
+	if (ctx->pos == 0) {
+		if (!dir_emit_dot(file, ctx))
+			return 0;
+		ctx->pos = 1;
+	}
+	nf_pos = (ctx->pos - 1) << 2;
+	if (nf_pos >= dir->i_size)
+		goto out;
+
+	/*
+	 * Something changed since last readdir (either lseek was called or dir
+	 * changed)?  We need to verify the position correctly points at the
+	 * beginning of some dir entry so that the directory parsing code does
+	 * not get confused. Since UDF does not have any reliable way of
+	 * identifying beginning of dir entry (names are under user control),
+	 * we need to scan the directory from the beginning.
+	 */
+	if (!inode_eq_iversion(dir, *(u64 *)file->private_data)) {
+		emit_pos = nf_pos;
+		nf_pos = 0;
+	} else {
+		pos_valid = true;
+	}
+
+	fname = kmalloc(UDF_NAME_LEN, GFP_KERNEL);
+	if (!fname) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	for (ret = udf_fiiter_init(&iter, dir, nf_pos);
+	     !ret && iter.pos < dir->i_size;
+	     ret = udf_fiiter_advance(&iter)) {
+		struct kernel_lb_addr tloc;
+		udf_pblk_t iblock;
+
+		/* Still not at offset where user asked us to read from? */
+		if (iter.pos < emit_pos)
+			continue;
+
+		/* Update file position only if we got past the current one */
+		pos_valid = true;
+		ctx->pos = (iter.pos >> 2) + 1;
+
+		if (iter.fi.fileCharacteristics & FID_FILE_CHAR_DELETED) {
+			if (!UDF_QUERY_FLAG(sb, UDF_FLAG_UNDELETE))
+				continue;
+		}
+
+		if (iter.fi.fileCharacteristics & FID_FILE_CHAR_HIDDEN) {
+			if (!UDF_QUERY_FLAG(sb, UDF_FLAG_UNHIDE))
+				continue;
+		}
+
+		if (iter.fi.fileCharacteristics & FID_FILE_CHAR_PARENT) {
+			if (!dir_emit_dotdot(file, ctx))
+				goto out_iter;
+			continue;
+		}
+
+		flen = udf_get_filename(sb, iter.name,
+				iter.fi.lengthFileIdent, fname, UDF_NAME_LEN);
+		if (flen < 0)
+			continue;
+
+		tloc = lelb_to_cpu(iter.fi.icb.extLocation);
+		iblock = udf_get_lb_pblock(sb, &tloc, 0);
+		if (!dir_emit(ctx, fname, flen, iblock, DT_UNKNOWN))
+			goto out_iter;
+	}
+
+	if (!ret) {
+		ctx->pos = (iter.pos >> 2) + 1;
+		pos_valid = true;
+	}
+out_iter:
+	udf_fiiter_release(&iter);
+out:
+	if (pos_valid)
+		*(u64 *)file->private_data = inode_query_iversion(dir);
+	kfree(fname);
+
+	return ret;
+}
+
+static int udf_dir_open(struct inode *inode, struct file *file)
+{
+	file->private_data = kzalloc(sizeof(u64), GFP_KERNEL);
+	if (!file->private_data)
+		return -ENOMEM;
+	return 0;
+}
+
+static int udf_dir_release(struct inode *inode, struct file *file)
+{
+	kfree(file->private_data);
+	return 0;
+}
+
+static loff_t udf_dir_llseek(struct file *file, loff_t offset, int whence)
+{
+	return generic_llseek_cookie(file, offset, whence,
+				     (u64 *)file->private_data);
+}
+
+/* readdir and lookup functions */
+const struct file_operations udf_dir_operations = {
+	.open			= udf_dir_open,
+	.release		= udf_dir_release,
+	.llseek			= udf_dir_llseek,
+	.read			= generic_read_dir,
+	.iterate_shared		= udf_readdir,
+	.unlocked_ioctl		= udf_ioctl,
+	.fsync			= simple_fsync,
+	.setlease		= generic_setlease,
+};