|
1
|
|
|
from __future__ import absolute_import |
|
2
|
|
|
from __future__ import print_function |
|
3
|
|
|
|
|
4
|
|
|
import os |
|
5
|
|
|
import os.path |
|
6
|
|
|
import json |
|
7
|
|
|
import sys |
|
8
|
|
|
|
|
9
|
|
|
from .build_yaml import Rule, DocumentationNotComplete |
|
10
|
|
|
from .constants import MULTI_PLATFORM_LIST |
|
11
|
|
|
from .jinja import process_file_with_macros |
|
12
|
|
|
from .rule_yaml import parse_prodtype |
|
13
|
|
|
from .rules import get_rule_dir_id, get_rule_dir_sces, find_rule_dirs_in_paths |
|
14
|
|
|
from . import utils |
|
15
|
|
|
|
|
16
|
|
|
|
|
17
|
|
|
def load_sce_and_metadata(file_path, local_env_yaml): |
|
18
|
|
|
""" |
|
19
|
|
|
For the given SCE audit file (file_path) under the specified environment |
|
20
|
|
|
(local_env_yaml), parse the file while expanding Jinja macros and read any |
|
21
|
|
|
metadata headers the file contains. Note that the last keyword of a |
|
22
|
|
|
specified type is the recorded one. |
|
23
|
|
|
|
|
24
|
|
|
Returns (audit_content, metadata). |
|
25
|
|
|
""" |
|
26
|
|
|
|
|
27
|
|
|
raw_content = process_file_with_macros(file_path, local_env_yaml) |
|
28
|
|
|
return load_sce_and_metadata_parsed(raw_content) |
|
29
|
|
|
|
|
30
|
|
|
|
|
31
|
|
|
def load_sce_and_metadata_parsed(raw_content): |
|
32
|
|
|
metadata = dict() |
|
33
|
|
|
sce_content = [] |
|
34
|
|
|
|
|
35
|
|
|
for line in raw_content.split("\n"): |
|
36
|
|
|
found_metadata = False |
|
37
|
|
|
keywords = ['platform', 'check-import', 'check-export', 'complex-check'] |
|
38
|
|
|
for keyword in keywords: |
|
39
|
|
|
if line.startswith('# ' + keyword + ' = '): |
|
40
|
|
|
# Strip off the initial comment marker |
|
41
|
|
|
_, value = line[2:].split('=', maxsplit=1) |
|
42
|
|
|
values = value.strip() |
|
43
|
|
|
if ',' in values: |
|
44
|
|
|
values.split(',') |
|
45
|
|
|
metadata[keyword] = values |
|
46
|
|
|
found_metadata = True |
|
47
|
|
|
break |
|
48
|
|
|
if not found_metadata: |
|
49
|
|
|
sce_content.append(line) |
|
50
|
|
|
|
|
51
|
|
|
return "\n".join(sce_content), metadata |
|
52
|
|
|
|
|
53
|
|
|
|
|
54
|
|
|
def _check_is_applicable_for_product(metadata, product): |
|
55
|
|
|
""" |
|
56
|
|
|
Validates whether or not the specified check is applicable for this |
|
57
|
|
|
product. Different from build_ovals.py in that this operates directly |
|
58
|
|
|
on the parsed metadata and doesn't have to deal with matching XML |
|
59
|
|
|
elements. |
|
60
|
|
|
""" |
|
61
|
|
|
|
|
62
|
|
|
if 'platform' not in metadata: |
|
63
|
|
|
return True |
|
64
|
|
|
|
|
65
|
|
|
product, product_version = utils.parse_name(product) |
|
66
|
|
|
|
|
67
|
|
|
multi_product = 'multi_platform_{0}'.format(product) |
|
68
|
|
|
if product in ['macos', 'ubuntu']: |
|
69
|
|
|
product_version = product_version[:2] + "." + product_version[2:] |
|
70
|
|
|
|
|
71
|
|
|
return ('multi_platform_all' in metadata['platform'] or |
|
72
|
|
|
(multi_product in metadata['platform'] and |
|
73
|
|
|
product in MULTI_PLATFORM_LIST) or |
|
74
|
|
|
(product in metadata['platform'] and |
|
75
|
|
|
product_version in metadata['platform'])) |
|
76
|
|
|
|
|
77
|
|
|
|
|
78
|
|
|
def _check_is_loaded(already_loaded, filename): |
|
79
|
|
|
# Right now this check doesn't look at metadata or anything |
|
80
|
|
|
# else. Eventually we might add versions to the entry or |
|
81
|
|
|
# something. |
|
82
|
|
|
return filename in already_loaded |
|
83
|
|
|
|
|
84
|
|
|
|
|
85
|
|
|
def checks(env_yaml, yaml_path, sce_dirs, template_builder, output): |
|
86
|
|
|
""" |
|
87
|
|
|
Walks the build system and builds all SCE checks (and metadata entry) |
|
88
|
|
|
into the output directory. |
|
89
|
|
|
""" |
|
90
|
|
|
product = utils.required_key(env_yaml, "product") |
|
91
|
|
|
included_checks_count = 0 |
|
92
|
|
|
reversed_dirs = sce_dirs[::-1] |
|
93
|
|
|
already_loaded = dict() |
|
94
|
|
|
local_env_yaml = dict() |
|
95
|
|
|
local_env_yaml.update(env_yaml) |
|
96
|
|
|
|
|
97
|
|
|
# We maintain the same search structure as build_ovals.py even though we |
|
98
|
|
|
# don't currently have any content under shared/checks/sce. |
|
99
|
|
|
product_dir = os.path.dirname(yaml_path) |
|
100
|
|
|
relative_guide_dir = utils.required_key(env_yaml, "benchmark_root") |
|
101
|
|
|
guide_dir = os.path.abspath(os.path.join(product_dir, relative_guide_dir)) |
|
102
|
|
|
additional_content_directories = env_yaml.get("additional_content_directories", []) |
|
103
|
|
|
add_content_dirs = [ |
|
104
|
|
|
os.path.abspath(os.path.join(product_dir, rd)) |
|
105
|
|
|
for rd in additional_content_directories |
|
106
|
|
|
] |
|
107
|
|
|
|
|
108
|
|
|
# First walk all rules under the product. These have higher priority than any |
|
109
|
|
|
# out-of-tree SCE checks. |
|
110
|
|
|
for _dir_path in find_rule_dirs_in_paths([guide_dir] + add_content_dirs): |
|
111
|
|
|
rule_id = get_rule_dir_id(_dir_path) |
|
112
|
|
|
|
|
113
|
|
|
rule_path = os.path.join(_dir_path, "rule.yml") |
|
114
|
|
|
try: |
|
115
|
|
|
rule = Rule.from_yaml(rule_path, env_yaml) |
|
116
|
|
|
except DocumentationNotComplete: |
|
117
|
|
|
# Happens on non-debug builds when a rule isn't yet completed. We |
|
118
|
|
|
# don't want to build the SCE check for this rule yet so skip it |
|
119
|
|
|
# and move on. |
|
120
|
|
|
continue |
|
121
|
|
|
|
|
122
|
|
|
prodtypes = parse_prodtype(rule.prodtype) |
|
123
|
|
|
if prodtypes and 'all' not in prodtypes and product not in prodtypes: |
|
124
|
|
|
# The prodtype exists, isn't all and doesn't contain this current |
|
125
|
|
|
# product, so we're best to skip this rule altogether. |
|
126
|
|
|
continue |
|
127
|
|
|
|
|
128
|
|
|
local_env_yaml['rule_id'] = rule.id_ |
|
129
|
|
|
local_env_yaml['rule_title'] = rule.title |
|
130
|
|
|
local_env_yaml['products'] = prodtypes # default is all |
|
131
|
|
|
|
|
132
|
|
|
for _path in get_rule_dir_sces(_dir_path, product): |
|
133
|
|
|
# To be compatible with later checks, use the rule_id (i.e., the |
|
134
|
|
|
# value of _dir) to recreate the expected filename if this OVAL |
|
135
|
|
|
# was in a rule directory. However, note that unlike |
|
136
|
|
|
# build_oval.checks(...), we have to get this script's extension |
|
137
|
|
|
# first. |
|
138
|
|
|
_, ext = os.path.splitext(_path) |
|
139
|
|
|
filename = "{0}{1}".format(rule_id, ext) |
|
140
|
|
|
|
|
141
|
|
|
sce_content, metadata = load_sce_and_metadata(_path, local_env_yaml) |
|
142
|
|
|
metadata['filename'] = filename |
|
143
|
|
|
|
|
144
|
|
|
if not _check_is_applicable_for_product(metadata, product): |
|
145
|
|
|
continue |
|
146
|
|
|
if _check_is_loaded(already_loaded, rule_id): |
|
147
|
|
|
continue |
|
148
|
|
|
|
|
149
|
|
|
with open(os.path.join(output, filename), 'w') as output_file: |
|
150
|
|
|
print(sce_content, file=output_file) |
|
151
|
|
|
|
|
152
|
|
|
included_checks_count += 1 |
|
153
|
|
|
already_loaded[rule_id] = metadata |
|
154
|
|
|
|
|
155
|
|
|
if rule.template: |
|
156
|
|
|
langs = template_builder.get_resolved_langs_to_generate(rule) |
|
157
|
|
|
if 'sce-bash' in langs: |
|
158
|
|
|
# Here we know the specified rule has a template and this |
|
159
|
|
|
# template actually generates (bash) SCE content. We |
|
160
|
|
|
# prioritize bespoke SCE content over templated content, |
|
161
|
|
|
# however, while we add this to our metadata, we do not |
|
162
|
|
|
# bother (yet!) with generating the SCE content. This is done |
|
163
|
|
|
# at a later time by build-scripts/build_templated_content.py. |
|
164
|
|
|
if _check_is_loaded(already_loaded, rule_id): |
|
165
|
|
|
continue |
|
166
|
|
|
|
|
167
|
|
|
# While we don't _write_ it, we still need to parse SCE |
|
168
|
|
|
# metadata from the templated content. Render it internally. |
|
169
|
|
|
raw_sce_content = template_builder.get_lang_for_rule( |
|
170
|
|
|
rule_id, rule.title, rule.template, 'sce-bash') |
|
171
|
|
|
|
|
172
|
|
|
ext = '.sh' |
|
173
|
|
|
filename = rule_id + ext |
|
174
|
|
|
|
|
175
|
|
|
# Load metadata and infer correct file name. |
|
176
|
|
|
sce_content, metadata = load_sce_and_metadata_parsed(raw_sce_content) |
|
177
|
|
|
metadata['filename'] = filename |
|
178
|
|
|
|
|
179
|
|
|
# Skip the check if it isn't applicable for this product. |
|
180
|
|
|
if not _check_is_applicable_for_product(metadata, product): |
|
181
|
|
|
continue |
|
182
|
|
|
|
|
183
|
|
|
with open(os.path.join(output, filename), 'w') as output_file: |
|
184
|
|
|
print(sce_content, file=output_file) |
|
185
|
|
|
|
|
186
|
|
|
# Finally, include it in our loaded content |
|
187
|
|
|
included_checks_count += 1 |
|
188
|
|
|
already_loaded[rule_id] = metadata |
|
189
|
|
|
|
|
190
|
|
|
# Finally take any shared SCE checks and build them as well. Note that |
|
191
|
|
|
# there's no way for shorthand generation to include them if they do NOT |
|
192
|
|
|
# align with a particular rule_id, so it is suggested that the former |
|
193
|
|
|
# method be used. |
|
194
|
|
|
for sce_dir in reversed_dirs: |
|
195
|
|
|
if not os.path.isdir(sce_dir): |
|
196
|
|
|
continue |
|
197
|
|
|
|
|
198
|
|
|
for filename in sorted(os.listdir(sce_dir)): |
|
199
|
|
|
rule_id, _ = os.path.splitext(filename) |
|
200
|
|
|
|
|
201
|
|
|
sce_content, metadata = load_sce_and_metadata(filename, env_yaml) |
|
202
|
|
|
metadata['filename'] = filename |
|
203
|
|
|
|
|
204
|
|
|
if not _check_is_applicable_for_product(metadata, product): |
|
205
|
|
|
continue |
|
206
|
|
|
if _check_is_loaded(already_loaded, rule_id): |
|
207
|
|
|
continue |
|
208
|
|
|
|
|
209
|
|
|
with open(os.path.join(output, filename), 'w') as output_file: |
|
210
|
|
|
print(sce_content, file=output_file) |
|
211
|
|
|
|
|
212
|
|
|
included_checks_count += 1 |
|
213
|
|
|
already_loaded[rule_id] = metadata |
|
214
|
|
|
|
|
215
|
|
|
# Finally, write out our metadata to disk so that we can reference it in |
|
216
|
|
|
# later build stages (such as during building shorthand content). |
|
217
|
|
|
metadata_path = os.path.join(output, 'metadata.json') |
|
218
|
|
|
json.dump(already_loaded, open(metadata_path, 'w')) |
|
219
|
|
|
|
|
220
|
|
|
return already_loaded |
|
221
|
|
|
|