xref: /xnu-11215.61.5/tools/vm_sanitize_enforcement.py (revision 4f1223e81cd707a65cc109d0b8ad6653699da3c4)
1import sys
2
3# This script's return values
4ERROR = 1
5SUCCESS = 0
6
7error_help_message = """Safe types were used in MIG in violation of VM API sanitization enforcement.
8\t- If you created a new MIG file that is not related to Virtual Memory, you will want to add it to the list of \
9exceptions in this Python script.
10\t- If your work is related to Virtual Memory, you need to use unsafe types for userspace inputs. Make sure you \
11define VM_KERNEL_SERVER in the MIG file, and that the safe type is associated to an unsafe type via VM_UNSAFE_TYPE/VM_TYPE_SAFE_UNSAFE. Learn more by reading doc/vm/sanitize.md."""
12
13# List of safe types that should not be used to represent userspace inputs.
14# Note that is may be counter-intuitive that we want to prevent the use of
15# safe types: they are safe after all, so wouldn't it be better to use them?
16# The source of the confusion is that the safety of the type really describes
17# the safety of the data. Since the data will be coming from userspace, it is
18# considered unsafe/untrusted, and we want to make sure we represent it with
19# unsafe types to force the programmer to sanitize it before it can be
20# represented with the (directly usable) safe type.
21# This is a list of all the safe types which have corresponding unsafe types
22# in vm_types_unsafe.h.
23safe_type_list = [
24	"mach_vm_address_t",
25	"mach_vm_offset_t",
26	"mach_vm_size_t",
27	"vm_address_t",
28	"vm_offset_t",
29	"vm_size_t",
30	"vm_map_address_t",
31	"vm_map_offset_t",
32	"vm_map_size_t",
33	"memory_object_offset_t",
34	"memory_object_size_t",
35	"vm_object_offset_t",
36	"vm_object_size_t",
37	"pointer_t",
38	"vm32_address_t",
39	"vm32_offset_t",
40	"vm32_size_t",
41	"vm_prot_t",
42	"vm_inherit_t",
43	"vm_behavior_t",
44	"caddr_t",
45	"user_addr_t",
46	"size_t",
47	"user_size_t",
48	"struct mach_vm_range",
49	"mach_vm_range_recipe_v1_t",
50]
51
52# Files that are considered outside the VM boundary and are thus not subject to enforcement.
53file_ignorelist = [
54	"arcade_register_server.c",
55	"clock_server.c",
56	"exc_server.c",
57	"mach_eventlink_server.c",
58	"mach_exc_server.c",
59	"mach_notify_server.c",
60	"mach_port_server.c",
61	"mach_voucher_server.c",
62	"memory_entry_server.c",
63	"processor_server.c",
64	"processor_set_server.c",
65	"restartable_server.c",
66	"task_server.c",
67	"thread_act_server.c",
68	"upl_server.c",
69]
70
71def print_error(*args, **kwargs):
72	print("error:", *args, file=sys.stderr, **kwargs)
73
74def is_type_used_in_line(safe_type, line):
75	# This is used by an autogenerated struct in MIG that isn't an argument to a MIG call
76	if "vm_address_t	reserved;	/* Reserved */" in line:
77		return False
78
79	# arguments to MIG functions are typically the first thing on the line in the generated header,
80	# but we search for the type elsewhere to be on the safe side. We still need to be careful not
81	# to trigger false positives by doing a naive search
82	# e.g. size_t is in "__Request__host_page_size_t __attribute__((unused));"
83	if safe_type in line.replace(':', ' ').replace(';', ' ').replace(',', ' ').split():
84		return True
85
86	return False
87
88def are_safe_types_used_in_file(filepath):
89	are_safe_types_used = False
90	lineno = 1
91	with open(filepath, "r") as file:
92		for line in file:
93			for safe_type in safe_type_list:
94				if is_type_used_in_line(safe_type, line):
95					print_error("Found safe type \"" + safe_type + "\" in " +filepath + ":" + str(lineno) + ". Line is \"" + line.strip() + "\"")
96					are_safe_types_used = True
97			lineno += 1
98	return are_safe_types_used
99
100def main():
101	if len(sys.argv) < 2:
102		print_error("usage: python vm_api_enforcement.py filename [extra_filename...]")
103		return ERROR
104
105	are_safe_types_used = False
106	for filename in sys.argv[1:]:
107		if filename in file_ignorelist:
108			continue
109
110		if not (filename.endswith(".c") or filename.endswith(".h")):
111			print_error("File should be a .c or .h file:", filename)
112			return ERROR
113		header = filename[:-1] + "h"
114
115		are_safe_types_used = are_safe_types_used_in_file(header)
116
117	if are_safe_types_used:
118		print_error("{}: {}".format(sys.argv[0], error_help_message))
119		return ERROR
120	return SUCCESS
121
122sys.exit(main())
123