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, 331 insertions, 0 deletions

diff --git a/fs/hpfs/dir.c b/fs/hpfs/dir.c
new file mode 100644
index 000000000..3bf11202e
--- /dev/null
+++ b/fs/hpfs/dir.c
@@ -0,0 +1,331 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/fs/hpfs/dir.c
+ *
+ *  Mikulas Patocka (mikulas@artax.karlin.mff.cuni.cz), 1998-1999
+ *
+ *  directory VFS functions
+ */
+
+#include <linux/slab.h>
+#include "hpfs_fn.h"
+
+static int hpfs_dir_release(struct inode *inode, struct file *filp)
+{
+	hpfs_lock(inode->i_sb);
+	hpfs_del_pos(inode, &filp->f_pos);
+	/*hpfs_write_if_changed(inode);*/
+	hpfs_unlock(inode->i_sb);
+	return 0;
+}
+
+/* This is slow, but it's not used often */
+
+static loff_t hpfs_dir_lseek(struct file *filp, loff_t off, int whence)
+{
+	loff_t new_off = off + (whence == 1 ? filp->f_pos : 0);
+	loff_t pos;
+	struct quad_buffer_head qbh;
+	struct inode *i = file_inode(filp);
+	struct hpfs_inode_info *hpfs_inode = hpfs_i(i);
+	struct super_block *s = i->i_sb;
+
+	/* Somebody else will have to figure out what to do here */
+	if (whence == SEEK_DATA || whence == SEEK_HOLE)
+		return -EINVAL;
+
+	inode_lock(i);
+	hpfs_lock(s);
+
+	/*pr_info("dir lseek\n");*/
+	if (new_off == 0 || new_off == 1 || new_off == 11 || new_off == 12 || new_off == 13) goto ok;
+	pos = ((loff_t) hpfs_de_as_down_as_possible(s, hpfs_inode->i_dno) << 4) + 1;
+	while (pos != new_off) {
+		if (map_pos_dirent(i, &pos, &qbh)) hpfs_brelse4(&qbh);
+		else goto fail;
+		if (pos == 12) goto fail;
+	}
+	if (unlikely(hpfs_add_pos(i, &filp->f_pos) < 0)) {
+		hpfs_unlock(s);
+		inode_unlock(i);
+		return -ENOMEM;
+	}
+ok:
+	filp->f_pos = new_off;
+	hpfs_unlock(s);
+	inode_unlock(i);
+	return new_off;
+fail:
+	/*pr_warn("illegal lseek: %016llx\n", new_off);*/
+	hpfs_unlock(s);
+	inode_unlock(i);
+	return -ESPIPE;
+}
+
+static int hpfs_readdir(struct file *file, struct dir_context *ctx)
+{
+	struct inode *inode = file_inode(file);
+	struct hpfs_inode_info *hpfs_inode = hpfs_i(inode);
+	struct quad_buffer_head qbh;
+	struct hpfs_dirent *de;
+	int lc;
+	loff_t next_pos;
+	unsigned char *tempname;
+	int c1, c2 = 0;
+	int ret = 0;
+
+	hpfs_lock(inode->i_sb);
+
+	if (hpfs_sb(inode->i_sb)->sb_chk) {
+		if (hpfs_chk_sectors(inode->i_sb, inode->i_ino, 1, "dir_fnode")) {
+			ret = -EFSERROR;
+			goto out;
+		}
+		if (hpfs_chk_sectors(inode->i_sb, hpfs_inode->i_dno, 4, "dir_dnode")) {
+			ret = -EFSERROR;
+			goto out;
+		}
+	}
+	if (hpfs_sb(inode->i_sb)->sb_chk >= 2) {
+		struct buffer_head *bh;
+		struct fnode *fno;
+		int e = 0;
+		if (!(fno = hpfs_map_fnode(inode->i_sb, inode->i_ino, &bh))) {
+			ret = -EIOERROR;
+			goto out;
+		}
+		if (!fnode_is_dir(fno)) {
+			e = 1;
+			hpfs_error(inode->i_sb, "not a directory, fnode %08llx",
+					inode->i_ino);
+		}
+		if (hpfs_inode->i_dno != le32_to_cpu(fno->u.external[0].disk_secno)) {
+			e = 1;
+			hpfs_error(inode->i_sb, "corrupted inode: i_dno == %08x, fnode -> dnode == %08x", hpfs_inode->i_dno, le32_to_cpu(fno->u.external[0].disk_secno));
+		}
+		brelse(bh);
+		if (e) {
+			ret = -EFSERROR;
+			goto out;
+		}
+	}
+	lc = hpfs_sb(inode->i_sb)->sb_lowercase;
+	if (ctx->pos == 12) { /* diff -r requires this (note, that diff -r */
+		ctx->pos = 13; /* also fails on msdos filesystem in 2.0) */
+		goto out;
+	}
+	if (ctx->pos == 13) {
+		ret = -ENOENT;
+		goto out;
+	}
+	
+	while (1) {
+		again:
+		/* This won't work when cycle is longer than number of dirents
+		   accepted by filldir, but what can I do?
+		   maybe killall -9 ls helps */
+		if (hpfs_sb(inode->i_sb)->sb_chk)
+			if (hpfs_stop_cycles(inode->i_sb, ctx->pos, &c1, &c2, "hpfs_readdir")) {
+				ret = -EFSERROR;
+				goto out;
+			}
+		if (ctx->pos == 12)
+			goto out;
+		if (ctx->pos == 3 || ctx->pos == 4 || ctx->pos == 5) {
+			pr_err("pos==%d\n", (int)ctx->pos);
+			goto out;
+		}
+		if (ctx->pos == 0) {
+			if (!dir_emit_dot(file, ctx))
+				goto out;
+			ctx->pos = 11;
+		}
+		if (ctx->pos == 11) {
+			if (!dir_emit(ctx, "..", 2, hpfs_inode->i_parent_dir, DT_DIR))
+				goto out;
+			ctx->pos = 1;
+		}
+		if (ctx->pos == 1) {
+			ret = hpfs_add_pos(inode, &file->f_pos);
+			if (unlikely(ret < 0))
+				goto out;
+			ctx->pos = ((loff_t) hpfs_de_as_down_as_possible(inode->i_sb, hpfs_inode->i_dno) << 4) + 1;
+		}
+		next_pos = ctx->pos;
+		if (!(de = map_pos_dirent(inode, &next_pos, &qbh))) {
+			ctx->pos = next_pos;
+			ret = -EIOERROR;
+			goto out;
+		}
+		if (de->first || de->last) {
+			if (hpfs_sb(inode->i_sb)->sb_chk) {
+				if (de->first && !de->last && (de->namelen != 2
+				    || de ->name[0] != 1 || de->name[1] != 1))
+					hpfs_error(inode->i_sb, "hpfs_readdir: bad ^A^A entry; pos = %08lx", (unsigned long)ctx->pos);
+				if (de->last && (de->namelen != 1 || de ->name[0] != 255))
+					hpfs_error(inode->i_sb, "hpfs_readdir: bad \\377 entry; pos = %08lx", (unsigned long)ctx->pos);
+			}
+			hpfs_brelse4(&qbh);
+			ctx->pos = next_pos;
+			goto again;
+		}
+		tempname = hpfs_translate_name(inode->i_sb, de->name, de->namelen, lc, de->not_8x3);
+		if (!dir_emit(ctx, tempname, de->namelen, le32_to_cpu(de->fnode), DT_UNKNOWN)) {
+			if (tempname != de->name) kfree(tempname);
+			hpfs_brelse4(&qbh);
+			goto out;
+		}
+		ctx->pos = next_pos;
+		if (tempname != de->name) kfree(tempname);
+		hpfs_brelse4(&qbh);
+	}
+out:
+	hpfs_unlock(inode->i_sb);
+	return ret;
+}
+
+/*
+ * lookup.  Search the specified directory for the specified name, set
+ * *result to the corresponding inode.
+ *
+ * lookup uses the inode number to tell read_inode whether it is reading
+ * the inode of a directory or a file -- file ino's are odd, directory
+ * ino's are even.  read_inode avoids i/o for file inodes; everything
+ * needed is up here in the directory.  (And file fnodes are out in
+ * the boondocks.)
+ *
+ *    - M.P.: this is over, sometimes we've got to read file's fnode for eas
+ *	      inode numbers are just fnode sector numbers; iget lock is used
+ *	      to tell read_inode to read fnode or not.
+ */
+
+struct dentry *hpfs_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
+{
+	const unsigned char *name = dentry->d_name.name;
+	unsigned len = dentry->d_name.len;
+	struct quad_buffer_head qbh;
+	struct hpfs_dirent *de;
+	ino_t ino;
+	int err;
+	struct inode *result = NULL;
+	struct hpfs_inode_info *hpfs_result;
+
+	hpfs_lock(dir->i_sb);
+	if ((err = hpfs_chk_name(name, &len))) {
+		if (err == -ENAMETOOLONG) {
+			hpfs_unlock(dir->i_sb);
+			return ERR_PTR(-ENAMETOOLONG);
+		}
+		goto end_add;
+	}
+
+	/*
+	 * '.' and '..' will never be passed here.
+	 */
+
+	de = map_dirent(dir, hpfs_i(dir)->i_dno, name, len, NULL, &qbh);
+
+	/*
+	 * This is not really a bailout, just means file not found.
+	 */
+
+	if (!de) goto end;
+
+	/*
+	 * Get inode number, what we're after.
+	 */
+
+	ino = le32_to_cpu(de->fnode);
+
+	/*
+	 * Go find or make an inode.
+	 */
+
+	result = iget_locked(dir->i_sb, ino);
+	if (!result) {
+		hpfs_error(dir->i_sb, "hpfs_lookup: can't get inode");
+		result = ERR_PTR(-ENOMEM);
+		goto bail1;
+	}
+	if (inode_state_read_once(result) & I_NEW) {
+		hpfs_init_inode(result);
+		if (de->directory)
+			hpfs_read_inode(result);
+		else if (le32_to_cpu(de->ea_size) && hpfs_sb(dir->i_sb)->sb_eas)
+			hpfs_read_inode(result);
+		else {
+			result->i_mode |= S_IFREG;
+			result->i_mode &= ~0111;
+			result->i_op = &hpfs_file_iops;
+			result->i_fop = &hpfs_file_ops;
+			set_nlink(result, 1);
+		}
+		unlock_new_inode(result);
+	}
+	hpfs_result = hpfs_i(result);
+	if (!de->directory) hpfs_result->i_parent_dir = dir->i_ino;
+
+	if (de->has_acl || de->has_xtd_perm) if (!sb_rdonly(dir->i_sb)) {
+		hpfs_error(result->i_sb, "ACLs or XPERM found. This is probably HPFS386. This driver doesn't support it now. Send me some info on these structures");
+		iput(result);
+		result = ERR_PTR(-EINVAL);
+		goto bail1;
+	}
+
+	/*
+	 * Fill in the info from the directory if this is a newly created
+	 * inode.
+	 */
+
+	if (!inode_get_ctime_sec(result)) {
+		time64_t csec = local_to_gmt(dir->i_sb, le32_to_cpu(de->creation_date));
+
+		inode_set_ctime(result, csec ? csec : 1, 0);
+		inode_set_mtime(result,
+				local_to_gmt(dir->i_sb, le32_to_cpu(de->write_date)),
+				0);
+		inode_set_atime(result,
+				local_to_gmt(dir->i_sb, le32_to_cpu(de->read_date)),
+				0);
+		hpfs_result->i_ea_size = le32_to_cpu(de->ea_size);
+		if (!hpfs_result->i_ea_mode && de->read_only)
+			result->i_mode &= ~0222;
+		if (!de->directory) {
+			if (result->i_size == -1) {
+				result->i_size = le32_to_cpu(de->file_size);
+				result->i_data.a_ops = &hpfs_aops;
+				hpfs_i(result)->mmu_private = result->i_size;
+			/*
+			 * i_blocks should count the fnode and any anodes.
+			 * We count 1 for the fnode and don't bother about
+			 * anodes -- the disk heads are on the directory band
+			 * and we want them to stay there.
+			 */
+				result->i_blocks = 1 + ((result->i_size + 511) >> 9);
+			}
+		}
+	}
+
+bail1:
+	hpfs_brelse4(&qbh);
+
+	/*
+	 * Made it.
+	 */
+
+end:
+end_add:
+	hpfs_unlock(dir->i_sb);
+	return d_splice_alias(result, dentry);
+}
+
+const struct file_operations hpfs_dir_ops =
+{
+	.llseek		= hpfs_dir_lseek,
+	.read		= generic_read_dir,
+	.iterate_shared	= hpfs_readdir,
+	.release	= hpfs_dir_release,
+	.fsync		= hpfs_file_fsync,
+	.unlocked_ioctl	= hpfs_ioctl,
+	.compat_ioctl	= compat_ptr_ioctl,
+};