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, 291 insertions, 0 deletions

diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
new file mode 100644
index 000000000..58cbc5e54
--- /dev/null
+++ b/drivers/md/dm-exception-store.c
@@ -0,0 +1,291 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2006-2008 Red Hat GmbH
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-exception-store.h"
+
+#include <linux/ctype.h>
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/vmalloc.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#define DM_MSG_PREFIX "snapshot exception stores"
+
+static LIST_HEAD(_exception_store_types);
+static DEFINE_SPINLOCK(_lock);
+
+static struct dm_exception_store_type *__find_exception_store_type(const char *name)
+{
+	struct dm_exception_store_type *type;
+
+	list_for_each_entry(type, &_exception_store_types, list)
+		if (!strcmp(name, type->name))
+			return type;
+
+	return NULL;
+}
+
+static struct dm_exception_store_type *_get_exception_store_type(const char *name)
+{
+	struct dm_exception_store_type *type;
+
+	spin_lock(&_lock);
+
+	type = __find_exception_store_type(name);
+
+	if (type && !try_module_get(type->module))
+		type = NULL;
+
+	spin_unlock(&_lock);
+
+	return type;
+}
+
+/*
+ * get_type
+ * @type_name
+ *
+ * Attempt to retrieve the dm_exception_store_type by name.  If not already
+ * available, attempt to load the appropriate module.
+ *
+ * Exstore modules are named "dm-exstore-" followed by the 'type_name'.
+ * Modules may contain multiple types.
+ * This function will first try the module "dm-exstore-<type_name>",
+ * then truncate 'type_name' on the last '-' and try again.
+ *
+ * For example, if type_name was "clustered-shared", it would search
+ * 'dm-exstore-clustered-shared' then 'dm-exstore-clustered'.
+ *
+ * 'dm-exception-store-<type_name>' is too long of a name in my
+ * opinion, which is why I've chosen to have the files
+ * containing exception store implementations be 'dm-exstore-<type_name>'.
+ * If you want your module to be autoloaded, you will follow this
+ * naming convention.
+ *
+ * Returns: dm_exception_store_type* on success, NULL on failure
+ */
+static struct dm_exception_store_type *get_type(const char *type_name)
+{
+	char *p, *type_name_dup;
+	struct dm_exception_store_type *type;
+
+	type = _get_exception_store_type(type_name);
+	if (type)
+		return type;
+
+	type_name_dup = kstrdup(type_name, GFP_KERNEL);
+	if (!type_name_dup) {
+		DMERR("No memory left to attempt load for \"%s\"", type_name);
+		return NULL;
+	}
+
+	while (request_module("dm-exstore-%s", type_name_dup) ||
+	       !(type = _get_exception_store_type(type_name))) {
+		p = strrchr(type_name_dup, '-');
+		if (!p)
+			break;
+		p[0] = '\0';
+	}
+
+	if (!type)
+		DMWARN("Module for exstore type \"%s\" not found.", type_name);
+
+	kfree(type_name_dup);
+
+	return type;
+}
+
+static void put_type(struct dm_exception_store_type *type)
+{
+	spin_lock(&_lock);
+	module_put(type->module);
+	spin_unlock(&_lock);
+}
+
+int dm_exception_store_type_register(struct dm_exception_store_type *type)
+{
+	int r = 0;
+
+	spin_lock(&_lock);
+	if (!__find_exception_store_type(type->name))
+		list_add(&type->list, &_exception_store_types);
+	else
+		r = -EBUSY;
+	spin_unlock(&_lock);
+
+	return r;
+}
+EXPORT_SYMBOL(dm_exception_store_type_register);
+
+int dm_exception_store_type_unregister(struct dm_exception_store_type *type)
+{
+	spin_lock(&_lock);
+
+	if (!__find_exception_store_type(type->name)) {
+		spin_unlock(&_lock);
+		return -EINVAL;
+	}
+
+	list_del(&type->list);
+
+	spin_unlock(&_lock);
+
+	return 0;
+}
+EXPORT_SYMBOL(dm_exception_store_type_unregister);
+
+static int set_chunk_size(struct dm_exception_store *store,
+			  const char *chunk_size_arg, char **error)
+{
+	unsigned int chunk_size;
+
+	if (kstrtouint(chunk_size_arg, 10, &chunk_size)) {
+		*error = "Invalid chunk size";
+		return -EINVAL;
+	}
+
+	if (!chunk_size) {
+		store->chunk_size = store->chunk_mask = store->chunk_shift = 0;
+		return 0;
+	}
+
+	return dm_exception_store_set_chunk_size(store, chunk_size, error);
+}
+
+int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
+				      unsigned int chunk_size,
+				      char **error)
+{
+	/* Check chunk_size is a power of 2 */
+	if (!is_power_of_2(chunk_size)) {
+		*error = "Chunk size is not a power of 2";
+		return -EINVAL;
+	}
+
+	/* Validate the chunk size against the device block size */
+	if (chunk_size %
+	    (bdev_logical_block_size(dm_snap_cow(store->snap)->bdev) >> 9) ||
+	    chunk_size %
+	    (bdev_logical_block_size(dm_snap_origin(store->snap)->bdev) >> 9)) {
+		*error = "Chunk size is not a multiple of device blocksize";
+		return -EINVAL;
+	}
+
+	if (chunk_size > INT_MAX >> SECTOR_SHIFT) {
+		*error = "Chunk size is too high";
+		return -EINVAL;
+	}
+
+	store->chunk_size = chunk_size;
+	store->chunk_mask = chunk_size - 1;
+	store->chunk_shift = __ffs(chunk_size);
+
+	return 0;
+}
+
+int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
+			      unsigned int *args_used,
+			      struct dm_exception_store **store)
+{
+	int r = 0;
+	struct dm_exception_store_type *type = NULL;
+	struct dm_exception_store *tmp_store;
+	char persistent;
+
+	if (argc < 2) {
+		ti->error = "Insufficient exception store arguments";
+		return -EINVAL;
+	}
+
+	tmp_store = kzalloc_obj(*tmp_store);
+	if (!tmp_store) {
+		ti->error = "Exception store allocation failed";
+		return -ENOMEM;
+	}
+
+	persistent = toupper(*argv[0]);
+	if (persistent == 'P')
+		type = get_type("P");
+	else if (persistent == 'N')
+		type = get_type("N");
+	else {
+		ti->error = "Exception store type is not P or N";
+		r = -EINVAL;
+		goto bad_type;
+	}
+
+	if (!type) {
+		ti->error = "Exception store type not recognised";
+		r = -EINVAL;
+		goto bad_type;
+	}
+
+	tmp_store->type = type;
+	tmp_store->snap = snap;
+
+	r = set_chunk_size(tmp_store, argv[1], &ti->error);
+	if (r)
+		goto bad;
+
+	r = type->ctr(tmp_store, (strlen(argv[0]) > 1 ? &argv[0][1] : NULL));
+	if (r) {
+		ti->error = "Exception store type constructor failed";
+		goto bad;
+	}
+
+	*args_used = 2;
+	*store = tmp_store;
+	return 0;
+
+bad:
+	put_type(type);
+bad_type:
+	kfree(tmp_store);
+	return r;
+}
+EXPORT_SYMBOL(dm_exception_store_create);
+
+void dm_exception_store_destroy(struct dm_exception_store *store)
+{
+	store->type->dtr(store);
+	put_type(store->type);
+	kfree(store);
+}
+EXPORT_SYMBOL(dm_exception_store_destroy);
+
+int dm_exception_store_init(void)
+{
+	int r;
+
+	r = dm_transient_snapshot_init();
+	if (r) {
+		DMERR("Unable to register transient exception store type.");
+		goto transient_fail;
+	}
+
+	r = dm_persistent_snapshot_init();
+	if (r) {
+		DMERR("Unable to register persistent exception store type");
+		goto persistent_fail;
+	}
+
+	return 0;
+
+persistent_fail:
+	dm_transient_snapshot_exit();
+transient_fail:
+	return r;
+}
+
+void dm_exception_store_exit(void)
+{
+	dm_persistent_snapshot_exit();
+	dm_transient_snapshot_exit();
+}