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, 366 insertions, 0 deletions

diff --git a/fs/ubifs/scan.c b/fs/ubifs/scan.c
new file mode 100644
index 000000000..cbff2d800
--- /dev/null
+++ b/fs/ubifs/scan.c
@@ -0,0 +1,366 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * This file is part of UBIFS.
+ *
+ * Copyright (C) 2006-2008 Nokia Corporation
+ *
+ * Authors: Adrian Hunter
+ *          Artem Bityutskiy (Битюцкий Артём)
+ */
+
+/*
+ * This file implements the scan which is a general-purpose function for
+ * determining what nodes are in an eraseblock. The scan is used to replay the
+ * journal, to do garbage collection. for the TNC in-the-gaps method, and by
+ * debugging functions.
+ */
+
+#include "ubifs.h"
+
+/**
+ * scan_padding_bytes - scan for padding bytes.
+ * @buf: buffer to scan
+ * @len: length of buffer
+ *
+ * This function returns the number of padding bytes on success and
+ * %SCANNED_GARBAGE on failure.
+ */
+static int scan_padding_bytes(void *buf, int len)
+{
+	int pad_len = 0, max_pad_len = min_t(int, UBIFS_PAD_NODE_SZ, len);
+	uint8_t *p = buf;
+
+	dbg_scan("not a node");
+
+	while (pad_len < max_pad_len && *p++ == UBIFS_PADDING_BYTE)
+		pad_len += 1;
+
+	if (!pad_len || (pad_len & 7))
+		return SCANNED_GARBAGE;
+
+	dbg_scan("%d padding bytes", pad_len);
+
+	return pad_len;
+}
+
+/**
+ * ubifs_scan_a_node - scan for a node or padding.
+ * @c: UBIFS file-system description object
+ * @buf: buffer to scan
+ * @len: length of buffer
+ * @lnum: logical eraseblock number
+ * @offs: offset within the logical eraseblock
+ * @quiet: print no messages
+ *
+ * This function returns a scanning code to indicate what was scanned.
+ */
+int ubifs_scan_a_node(const struct ubifs_info *c, void *buf, int len, int lnum,
+		      int offs, int quiet)
+{
+	struct ubifs_ch *ch = buf;
+	uint32_t magic;
+
+	magic = le32_to_cpu(ch->magic);
+
+	if (magic == 0xFFFFFFFF) {
+		dbg_scan("hit empty space at LEB %d:%d", lnum, offs);
+		return SCANNED_EMPTY_SPACE;
+	}
+
+	if (magic != UBIFS_NODE_MAGIC)
+		return scan_padding_bytes(buf, len);
+
+	if (len < UBIFS_CH_SZ)
+		return SCANNED_GARBAGE;
+
+	dbg_scan("scanning %s at LEB %d:%d",
+		 dbg_ntype(ch->node_type), lnum, offs);
+
+	if (ubifs_check_node(c, buf, len, lnum, offs, quiet, 1))
+		return SCANNED_A_CORRUPT_NODE;
+
+	if (ch->node_type == UBIFS_PAD_NODE) {
+		struct ubifs_pad_node *pad = buf;
+		int pad_len = le32_to_cpu(pad->pad_len);
+		int node_len = le32_to_cpu(ch->len);
+
+		/* Validate the padding node */
+		if (pad_len < 0 ||
+		    offs + node_len + pad_len > c->leb_size) {
+			if (!quiet) {
+				ubifs_err(c, "bad pad node at LEB %d:%d",
+					  lnum, offs);
+				ubifs_dump_node(c, pad, len);
+			}
+			return SCANNED_A_BAD_PAD_NODE;
+		}
+
+		/* Make the node pads to 8-byte boundary */
+		if ((node_len + pad_len) & 7) {
+			if (!quiet)
+				ubifs_err(c, "bad padding length %d - %d",
+					  offs, offs + node_len + pad_len);
+			return SCANNED_A_BAD_PAD_NODE;
+		}
+
+		dbg_scan("%d bytes padded at LEB %d:%d, offset now %d", pad_len,
+			 lnum, offs, ALIGN(offs + node_len + pad_len, 8));
+
+		return node_len + pad_len;
+	}
+
+	return SCANNED_A_NODE;
+}
+
+/**
+ * ubifs_start_scan - create LEB scanning information at start of scan.
+ * @c: UBIFS file-system description object
+ * @lnum: logical eraseblock number
+ * @offs: offset to start at (usually zero)
+ * @sbuf: scan buffer (must be c->leb_size)
+ *
+ * This function returns the scanned information on success and a negative error
+ * code on failure.
+ */
+struct ubifs_scan_leb *ubifs_start_scan(const struct ubifs_info *c, int lnum,
+					int offs, void *sbuf)
+{
+	struct ubifs_scan_leb *sleb;
+	int err;
+
+	dbg_scan("scan LEB %d:%d", lnum, offs);
+
+	sleb = kzalloc_obj(struct ubifs_scan_leb, GFP_NOFS);
+	if (!sleb)
+		return ERR_PTR(-ENOMEM);
+
+	sleb->lnum = lnum;
+	INIT_LIST_HEAD(&sleb->nodes);
+	sleb->buf = sbuf;
+
+	err = ubifs_leb_read(c, lnum, sbuf + offs, offs, c->leb_size - offs, 0);
+	if (err && err != -EBADMSG) {
+		ubifs_err(c, "cannot read %d bytes from LEB %d:%d, error %d",
+			  c->leb_size - offs, lnum, offs, err);
+		kfree(sleb);
+		return ERR_PTR(err);
+	}
+
+	/*
+	 * Note, we ignore integrity errors (EBASMSG) because all the nodes are
+	 * protected by CRC checksums.
+	 */
+	return sleb;
+}
+
+/**
+ * ubifs_end_scan - update LEB scanning information at end of scan.
+ * @c: UBIFS file-system description object
+ * @sleb: scanning information
+ * @lnum: logical eraseblock number
+ * @offs: offset to start at (usually zero)
+ */
+void ubifs_end_scan(const struct ubifs_info *c, struct ubifs_scan_leb *sleb,
+		    int lnum, int offs)
+{
+	dbg_scan("stop scanning LEB %d at offset %d", lnum, offs);
+	ubifs_assert(c, offs % c->min_io_size == 0);
+
+	sleb->endpt = ALIGN(offs, c->min_io_size);
+}
+
+/**
+ * ubifs_add_snod - add a scanned node to LEB scanning information.
+ * @c: UBIFS file-system description object
+ * @sleb: scanning information
+ * @buf: buffer containing node
+ * @offs: offset of node on flash
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_add_snod(const struct ubifs_info *c, struct ubifs_scan_leb *sleb,
+		   void *buf, int offs)
+{
+	struct ubifs_ch *ch = buf;
+	struct ubifs_ino_node *ino = buf;
+	struct ubifs_scan_node *snod;
+
+	snod = kmalloc_obj(struct ubifs_scan_node, GFP_NOFS);
+	if (!snod)
+		return -ENOMEM;
+
+	snod->sqnum = le64_to_cpu(ch->sqnum);
+	snod->type = ch->node_type;
+	snod->offs = offs;
+	snod->len = le32_to_cpu(ch->len);
+	snod->node = buf;
+
+	switch (ch->node_type) {
+	case UBIFS_INO_NODE:
+	case UBIFS_DENT_NODE:
+	case UBIFS_XENT_NODE:
+	case UBIFS_DATA_NODE:
+		/*
+		 * The key is in the same place in all keyed
+		 * nodes.
+		 */
+		key_read(c, &ino->key, &snod->key);
+		break;
+	default:
+		invalid_key_init(c, &snod->key);
+		break;
+	}
+	list_add_tail(&snod->list, &sleb->nodes);
+	sleb->nodes_cnt += 1;
+	return 0;
+}
+
+/**
+ * ubifs_scanned_corruption - print information after UBIFS scanned corruption.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number of corruption
+ * @offs: offset of corruption
+ * @buf: buffer containing corruption
+ */
+void ubifs_scanned_corruption(const struct ubifs_info *c, int lnum, int offs,
+			      void *buf)
+{
+	int len;
+
+	ubifs_err(c, "corruption at LEB %d:%d", lnum, offs);
+	len = c->leb_size - offs;
+	if (len > 8192)
+		len = 8192;
+	ubifs_err(c, "first %d bytes from LEB %d:%d", len, lnum, offs);
+	print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 4, buf, len, 1);
+}
+
+/**
+ * ubifs_scan - scan a logical eraseblock.
+ * @c: UBIFS file-system description object
+ * @lnum: logical eraseblock number
+ * @offs: offset to start at (usually zero)
+ * @sbuf: scan buffer (must be of @c->leb_size bytes in size)
+ * @quiet: print no messages
+ *
+ * This function scans LEB number @lnum and returns complete information about
+ * its contents. Returns the scanned information in case of success and,
+ * %-EUCLEAN if the LEB neads recovery, and other negative error codes in case
+ * of failure.
+ *
+ * If @quiet is non-zero, this function does not print large and scary
+ * error messages and flash dumps in case of errors.
+ */
+struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
+				  int offs, void *sbuf, int quiet)
+{
+	void *buf = sbuf + offs;
+	int err, len = c->leb_size - offs;
+	struct ubifs_scan_leb *sleb;
+
+	sleb = ubifs_start_scan(c, lnum, offs, sbuf);
+	if (IS_ERR(sleb))
+		return sleb;
+
+	while (len >= 8) {
+		struct ubifs_ch *ch = buf;
+		int node_len, ret;
+
+		dbg_scan("look at LEB %d:%d (%d bytes left)",
+			 lnum, offs, len);
+
+		cond_resched();
+
+		ret = ubifs_scan_a_node(c, buf, len, lnum, offs, quiet);
+		if (ret > 0) {
+			/* Padding bytes or a valid padding node */
+			offs += ret;
+			buf += ret;
+			len -= ret;
+			continue;
+		}
+
+		if (ret == SCANNED_EMPTY_SPACE)
+			/* Empty space is checked later */
+			break;
+
+		switch (ret) {
+		case SCANNED_GARBAGE:
+			ubifs_err(c, "garbage");
+			goto corrupted;
+		case SCANNED_A_NODE:
+			break;
+		case SCANNED_A_CORRUPT_NODE:
+		case SCANNED_A_BAD_PAD_NODE:
+			ubifs_err(c, "bad node");
+			goto corrupted;
+		default:
+			ubifs_err(c, "unknown");
+			err = -EINVAL;
+			goto error;
+		}
+
+		err = ubifs_add_snod(c, sleb, buf, offs);
+		if (err)
+			goto error;
+
+		node_len = ALIGN(le32_to_cpu(ch->len), 8);
+		offs += node_len;
+		buf += node_len;
+		len -= node_len;
+	}
+
+	if (offs % c->min_io_size) {
+		if (!quiet)
+			ubifs_err(c, "empty space starts at non-aligned offset %d",
+				  offs);
+		goto corrupted;
+	}
+
+	ubifs_end_scan(c, sleb, lnum, offs);
+
+	for (; len > 4; offs += 4, buf = buf + 4, len -= 4)
+		if (*(uint32_t *)buf != 0xffffffff)
+			break;
+	for (; len; offs++, buf++, len--)
+		if (*(uint8_t *)buf != 0xff) {
+			if (!quiet)
+				ubifs_err(c, "corrupt empty space at LEB %d:%d",
+					  lnum, offs);
+			goto corrupted;
+		}
+
+	return sleb;
+
+corrupted:
+	if (!quiet) {
+		ubifs_scanned_corruption(c, lnum, offs, buf);
+		ubifs_err(c, "LEB %d scanning failed", lnum);
+	}
+	err = -EUCLEAN;
+	ubifs_scan_destroy(sleb);
+	return ERR_PTR(err);
+
+error:
+	ubifs_err(c, "LEB %d scanning failed, error %d", lnum, err);
+	ubifs_scan_destroy(sleb);
+	return ERR_PTR(err);
+}
+
+/**
+ * ubifs_scan_destroy - destroy LEB scanning information.
+ * @sleb: scanning information to free
+ */
+void ubifs_scan_destroy(struct ubifs_scan_leb *sleb)
+{
+	struct ubifs_scan_node *node;
+	struct list_head *head;
+
+	head = &sleb->nodes;
+	while (!list_empty(head)) {
+		node = list_entry(head->next, struct ubifs_scan_node, list);
+		list_del(&node->list);
+		kfree(node);
+	}
+	kfree(sleb);
+}