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, 263 insertions, 0 deletions

diff --git a/drivers/md/persistent-data/dm-btree-spine.c b/drivers/md/persistent-data/dm-btree-spine.c
new file mode 100644
index 000000000..c46fc50c2
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree-spine.c
@@ -0,0 +1,263 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-btree-internal.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "btree spine"
+
+/*----------------------------------------------------------------*/
+
+#define BTREE_CSUM_XOR 121107
+
+static void node_prepare_for_write(const struct dm_block_validator *v,
+				   struct dm_block *b,
+				   size_t block_size)
+{
+	struct btree_node *n = dm_block_data(b);
+	struct node_header *h = &n->header;
+
+	h->blocknr = cpu_to_le64(dm_block_location(b));
+	h->csum = cpu_to_le32(dm_bm_checksum(&h->flags,
+					     block_size - sizeof(__le32),
+					     BTREE_CSUM_XOR));
+}
+
+static int node_check(const struct dm_block_validator *v,
+		      struct dm_block *b,
+		      size_t block_size)
+{
+	struct btree_node *n = dm_block_data(b);
+	struct node_header *h = &n->header;
+	size_t value_size;
+	__le32 csum_disk;
+	uint32_t flags, nr_entries, max_entries;
+
+	if (dm_block_location(b) != le64_to_cpu(h->blocknr)) {
+		DMERR_LIMIT("%s failed: blocknr %llu != wanted %llu", __func__,
+			    le64_to_cpu(h->blocknr), dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	csum_disk = cpu_to_le32(dm_bm_checksum(&h->flags,
+					       block_size - sizeof(__le32),
+					       BTREE_CSUM_XOR));
+	if (csum_disk != h->csum) {
+		DMERR_LIMIT("%s failed: csum %u != wanted %u", __func__,
+			    le32_to_cpu(csum_disk), le32_to_cpu(h->csum));
+		return -EILSEQ;
+	}
+
+	nr_entries = le32_to_cpu(h->nr_entries);
+	max_entries = le32_to_cpu(h->max_entries);
+	value_size = le32_to_cpu(h->value_size);
+
+	if (sizeof(struct node_header) +
+	    (sizeof(__le64) + value_size) * max_entries > block_size) {
+		DMERR_LIMIT("%s failed: max_entries too large", __func__);
+		return -EILSEQ;
+	}
+
+	if (nr_entries > max_entries) {
+		DMERR_LIMIT("%s failed: too many entries", __func__);
+		return -EILSEQ;
+	}
+
+	/*
+	 * The node must be either INTERNAL or LEAF.
+	 */
+	flags = le32_to_cpu(h->flags);
+	if (!(flags & INTERNAL_NODE) && !(flags & LEAF_NODE)) {
+		DMERR_LIMIT("%s failed: node is neither INTERNAL or LEAF", __func__);
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+const struct dm_block_validator btree_node_validator = {
+	.name = "btree_node",
+	.prepare_for_write = node_prepare_for_write,
+	.check = node_check
+};
+
+/*----------------------------------------------------------------*/
+
+int bn_read_lock(struct dm_btree_info *info, dm_block_t b,
+		 struct dm_block **result)
+{
+	return dm_tm_read_lock(info->tm, b, &btree_node_validator, result);
+}
+
+static int bn_shadow(struct dm_btree_info *info, dm_block_t orig,
+	      struct dm_btree_value_type *vt,
+	      struct dm_block **result)
+{
+	int r, inc;
+
+	r = dm_tm_shadow_block(info->tm, orig, &btree_node_validator,
+			       result, &inc);
+	if (!r && inc)
+		inc_children(info->tm, dm_block_data(*result), vt);
+
+	return r;
+}
+
+int new_block(struct dm_btree_info *info, struct dm_block **result)
+{
+	return dm_tm_new_block(info->tm, &btree_node_validator, result);
+}
+
+void unlock_block(struct dm_btree_info *info, struct dm_block *b)
+{
+	dm_tm_unlock(info->tm, b);
+}
+
+/*----------------------------------------------------------------*/
+
+void init_ro_spine(struct ro_spine *s, struct dm_btree_info *info)
+{
+	s->info = info;
+	s->count = 0;
+	s->nodes[0] = NULL;
+	s->nodes[1] = NULL;
+}
+
+void exit_ro_spine(struct ro_spine *s)
+{
+	int i;
+
+	for (i = 0; i < s->count; i++)
+		unlock_block(s->info, s->nodes[i]);
+}
+
+int ro_step(struct ro_spine *s, dm_block_t new_child)
+{
+	int r;
+
+	if (s->count == 2) {
+		unlock_block(s->info, s->nodes[0]);
+		s->nodes[0] = s->nodes[1];
+		s->count--;
+	}
+
+	r = bn_read_lock(s->info, new_child, s->nodes + s->count);
+	if (!r)
+		s->count++;
+
+	return r;
+}
+
+void ro_pop(struct ro_spine *s)
+{
+	BUG_ON(!s->count);
+	--s->count;
+	unlock_block(s->info, s->nodes[s->count]);
+}
+
+struct btree_node *ro_node(struct ro_spine *s)
+{
+	struct dm_block *block;
+
+	BUG_ON(!s->count);
+	block = s->nodes[s->count - 1];
+
+	return dm_block_data(block);
+}
+
+/*----------------------------------------------------------------*/
+
+void init_shadow_spine(struct shadow_spine *s, struct dm_btree_info *info)
+{
+	s->info = info;
+	s->count = 0;
+}
+
+void exit_shadow_spine(struct shadow_spine *s)
+{
+	int i;
+
+	for (i = 0; i < s->count; i++)
+		unlock_block(s->info, s->nodes[i]);
+}
+
+int shadow_step(struct shadow_spine *s, dm_block_t b,
+		struct dm_btree_value_type *vt)
+{
+	int r;
+
+	if (s->count == 2) {
+		unlock_block(s->info, s->nodes[0]);
+		s->nodes[0] = s->nodes[1];
+		s->count--;
+	}
+
+	r = bn_shadow(s->info, b, vt, s->nodes + s->count);
+	if (!r) {
+		if (!s->count)
+			s->root = dm_block_location(s->nodes[0]);
+
+		s->count++;
+	}
+
+	return r;
+}
+
+struct dm_block *shadow_current(struct shadow_spine *s)
+{
+	BUG_ON(!s->count);
+
+	return s->nodes[s->count - 1];
+}
+
+struct dm_block *shadow_parent(struct shadow_spine *s)
+{
+	BUG_ON(s->count != 2);
+
+	return s->count == 2 ? s->nodes[0] : NULL;
+}
+
+int shadow_has_parent(struct shadow_spine *s)
+{
+	return s->count >= 2;
+}
+
+dm_block_t shadow_root(struct shadow_spine *s)
+{
+	return s->root;
+}
+
+static void le64_inc(void *context, const void *value_le, unsigned int count)
+{
+	dm_tm_with_runs(context, value_le, count, dm_tm_inc_range);
+}
+
+static void le64_dec(void *context, const void *value_le, unsigned int count)
+{
+	dm_tm_with_runs(context, value_le, count, dm_tm_dec_range);
+}
+
+static int le64_equal(void *context, const void *value1_le, const void *value2_le)
+{
+	__le64 v1_le, v2_le;
+
+	memcpy(&v1_le, value1_le, sizeof(v1_le));
+	memcpy(&v2_le, value2_le, sizeof(v2_le));
+	return v1_le == v2_le;
+}
+
+void init_le64_type(struct dm_transaction_manager *tm,
+		    struct dm_btree_value_type *vt)
+{
+	vt->context = tm;
+	vt->size = sizeof(__le64);
+	vt->inc = le64_inc;
+	vt->dec = le64_dec;
+	vt->equal = le64_equal;
+}