1#!/usr/bin/env python
2#
3# Copyright (C) 2018 The Android Open Source Project
4#
5# Licensed under the Apache License, Version 2.0 (the "License");
6# you may not use this file except in compliance with the License.
7# You may obtain a copy of the License at
8#
9#      http://www.apache.org/licenses/LICENSE-2.0
10#
11# Unless required by applicable law or agreed to in writing, software
12# distributed under the License is distributed on an "AS IS" BASIS,
13# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14# See the License for the specific language governing permissions and
15# limitations under the License.
16"""
17Generate API lists for non-SDK API enforcement.
18"""
19import argparse
20from collections import defaultdict
21import os
22import sys
23import re
24import functools
25
26# Names of flags recognized by the `hiddenapi` tool.
27FLAG_WHITELIST = "whitelist"
28FLAG_GREYLIST = "greylist"
29FLAG_BLACKLIST = "blacklist"
30FLAG_GREYLIST_MAX_O = "greylist-max-o"
31FLAG_GREYLIST_MAX_P = "greylist-max-p"
32FLAG_CORE_PLATFORM_API = "core-platform-api"
33FLAG_PUBLIC_API = "public-api"
34FLAG_SYSTEM_API = "system-api"
35FLAG_TEST_API = "test-api"
36
37# List of all known flags.
38FLAGS_API_LIST = [
39    FLAG_WHITELIST,
40    FLAG_GREYLIST,
41    FLAG_BLACKLIST,
42    FLAG_GREYLIST_MAX_O,
43    FLAG_GREYLIST_MAX_P,
44]
45ALL_FLAGS = FLAGS_API_LIST + [
46    FLAG_CORE_PLATFORM_API,
47    FLAG_PUBLIC_API,
48    FLAG_SYSTEM_API,
49    FLAG_TEST_API,
50    ]
51
52FLAGS_API_LIST_SET = set(FLAGS_API_LIST)
53ALL_FLAGS_SET = set(ALL_FLAGS)
54
55# Suffix used in command line args to express that only known and
56# otherwise unassigned entries should be assign the given flag.
57# For example, the P dark greylist is checked in as it was in P,
58# but signatures have changes since then. The flag instructs this
59# script to skip any entries which do not exist any more.
60FLAG_IGNORE_CONFLICTS_SUFFIX = "-ignore-conflicts"
61
62# Suffix used in command line args to express that all apis within a given set
63# of packages should be assign the given flag.
64FLAG_PACKAGES_SUFFIX = "-packages"
65
66# Regex patterns of fields/methods used in serialization. These are
67# considered public API despite being hidden.
68SERIALIZATION_PATTERNS = [
69    r'readObject\(Ljava/io/ObjectInputStream;\)V',
70    r'readObjectNoData\(\)V',
71    r'readResolve\(\)Ljava/lang/Object;',
72    r'serialVersionUID:J',
73    r'serialPersistentFields:\[Ljava/io/ObjectStreamField;',
74    r'writeObject\(Ljava/io/ObjectOutputStream;\)V',
75    r'writeReplace\(\)Ljava/lang/Object;',
76]
77
78# Single regex used to match serialization API. It combines all the
79# SERIALIZATION_PATTERNS into a single regular expression.
80SERIALIZATION_REGEX = re.compile(r'.*->(' + '|'.join(SERIALIZATION_PATTERNS) + r')$')
81
82# Predicates to be used with filter_apis.
83HAS_NO_API_LIST_ASSIGNED = lambda api, flags: not FLAGS_API_LIST_SET.intersection(flags)
84IS_SERIALIZATION = lambda api, flags: SERIALIZATION_REGEX.match(api)
85
86def get_args():
87    """Parses command line arguments.
88
89    Returns:
90        Namespace: dictionary of parsed arguments
91    """
92    parser = argparse.ArgumentParser()
93    parser.add_argument('--output', required=True)
94    parser.add_argument('--csv', nargs='*', default=[], metavar='CSV_FILE',
95        help='CSV files to be merged into output')
96
97    for flag in ALL_FLAGS:
98        ignore_conflicts_flag = flag + FLAG_IGNORE_CONFLICTS_SUFFIX
99        packages_flag = flag + FLAG_PACKAGES_SUFFIX
100        parser.add_argument('--' + flag, dest=flag, nargs='*', default=[], metavar='TXT_FILE',
101            help='lists of entries with flag "' + flag + '"')
102        parser.add_argument('--' + ignore_conflicts_flag, dest=ignore_conflicts_flag, nargs='*',
103            default=[], metavar='TXT_FILE',
104            help='lists of entries with flag "' + flag +
105                 '". skip entry if missing or flag conflict.')
106        parser.add_argument('--' + packages_flag, dest=packages_flag, nargs='*',
107            default=[], metavar='TXT_FILE',
108            help='lists of packages to be added to ' + flag + ' list')
109
110    return parser.parse_args()
111
112def read_lines(filename):
113    """Reads entire file and return it as a list of lines.
114
115    Lines which begin with a hash are ignored.
116
117    Args:
118        filename (string): Path to the file to read from.
119
120    Returns:
121        Lines of the file as a list of string.
122    """
123    with open(filename, 'r') as f:
124        lines = f.readlines();
125    lines = filter(lambda line: not line.startswith('#'), lines)
126    lines = map(lambda line: line.strip(), lines)
127    return set(lines)
128
129def write_lines(filename, lines):
130    """Writes list of lines into a file, overwriting the file it it exists.
131
132    Args:
133        filename (string): Path to the file to be writting into.
134        lines (list): List of strings to write into the file.
135    """
136    lines = map(lambda line: line + '\n', lines)
137    with open(filename, 'w') as f:
138        f.writelines(lines)
139
140def extract_package(signature):
141    """Extracts the package from a signature.
142
143    Args:
144        signature (string): JNI signature of a method or field.
145
146    Returns:
147        The package name of the class containing the field/method.
148    """
149    full_class_name = signature.split(";->")[0]
150    package_name = full_class_name[1:full_class_name.rindex("/")]
151    return package_name.replace('/', '.')
152
153class FlagsDict:
154    def __init__(self):
155        self._dict_keyset = set()
156        self._dict = defaultdict(set)
157
158    def _check_entries_set(self, keys_subset, source):
159        assert isinstance(keys_subset, set)
160        assert keys_subset.issubset(self._dict_keyset), (
161            "Error processing: {}\n"
162            "The following entries were unexpected:\n"
163            "{}"
164            "Please visit go/hiddenapi for more information.").format(
165                source, "".join(map(lambda x: "  " + str(x), keys_subset - self._dict_keyset)))
166
167    def _check_flags_set(self, flags_subset, source):
168        assert isinstance(flags_subset, set)
169        assert flags_subset.issubset(ALL_FLAGS_SET), (
170            "Error processing: {}\n"
171            "The following flags were not recognized: \n"
172            "{}\n"
173            "Please visit go/hiddenapi for more information.").format(
174                source, "\n".join(flags_subset - ALL_FLAGS_SET))
175
176    def filter_apis(self, filter_fn):
177        """Returns APIs which match a given predicate.
178
179        This is a helper function which allows to filter on both signatures (keys) and
180        flags (values). The built-in filter() invokes the lambda only with dict's keys.
181
182        Args:
183            filter_fn : Function which takes two arguments (signature/flags) and returns a boolean.
184
185        Returns:
186            A set of APIs which match the predicate.
187        """
188        return set(filter(lambda x: filter_fn(x, self._dict[x]), self._dict_keyset))
189
190    def get_valid_subset_of_unassigned_apis(self, api_subset):
191        """Sanitizes a key set input to only include keys which exist in the dictionary
192        and have not been assigned any API list flags.
193
194        Args:
195            entries_subset (set/list): Key set to be sanitized.
196
197        Returns:
198            Sanitized key set.
199        """
200        assert isinstance(api_subset, set)
201        return api_subset.intersection(self.filter_apis(HAS_NO_API_LIST_ASSIGNED))
202
203    def generate_csv(self):
204        """Constructs CSV entries from a dictionary.
205
206        Returns:
207            List of lines comprising a CSV file. See "parse_and_merge_csv" for format description.
208        """
209        return sorted(map(lambda api: ",".join([api] + sorted(self._dict[api])), self._dict))
210
211    def parse_and_merge_csv(self, csv_lines, source = "<unknown>"):
212        """Parses CSV entries and merges them into a given dictionary.
213
214        The expected CSV format is:
215            <api signature>,<flag1>,<flag2>,...,<flagN>
216
217        Args:
218            csv_lines (list of strings): Lines read from a CSV file.
219            source (string): Origin of `csv_lines`. Will be printed in error messages.
220
221        Throws:
222            AssertionError if parsed flags are invalid.
223        """
224        # Split CSV lines into arrays of values.
225        csv_values = [ line.split(',') for line in csv_lines ]
226
227        # Update the full set of API signatures.
228        self._dict_keyset.update([ csv[0] for csv in csv_values ])
229
230        # Check that all flags are known.
231        csv_flags = set(functools.reduce(
232            lambda x, y: set(x).union(y),
233            [ csv[1:] for csv in csv_values ],
234            []))
235        self._check_flags_set(csv_flags, source)
236
237        # Iterate over all CSV lines, find entry in dict and append flags to it.
238        for csv in csv_values:
239            flags = csv[1:]
240            if (FLAG_PUBLIC_API in flags) or (FLAG_SYSTEM_API in flags):
241                flags.append(FLAG_WHITELIST)
242            elif FLAG_TEST_API in flags:
243                flags.append(FLAG_GREYLIST)
244            self._dict[csv[0]].update(flags)
245
246    def assign_flag(self, flag, apis, source="<unknown>"):
247        """Assigns a flag to given subset of entries.
248
249        Args:
250            flag (string): One of ALL_FLAGS.
251            apis (set): Subset of APIs to receive the flag.
252            source (string): Origin of `entries_subset`. Will be printed in error messages.
253
254        Throws:
255            AssertionError if parsed API signatures of flags are invalid.
256        """
257        # Check that all APIs exist in the dict.
258        self._check_entries_set(apis, source)
259
260        # Check that the flag is known.
261        self._check_flags_set(set([ flag ]), source)
262
263        # Iterate over the API subset, find each entry in dict and assign the flag to it.
264        for api in apis:
265            self._dict[api].add(flag)
266
267def main(argv):
268    # Parse arguments.
269    args = vars(get_args())
270
271    # Initialize API->flags dictionary.
272    flags = FlagsDict()
273
274    # Merge input CSV files into the dictionary.
275    # Do this first because CSV files produced by parsing API stubs will
276    # contain the full set of APIs. Subsequent additions from text files
277    # will be able to detect invalid entries, and/or filter all as-yet
278    # unassigned entries.
279    for filename in args["csv"]:
280        flags.parse_and_merge_csv(read_lines(filename), filename)
281
282    # Combine inputs which do not require any particular order.
283    # (1) Assign serialization API to whitelist.
284    flags.assign_flag(FLAG_WHITELIST, flags.filter_apis(IS_SERIALIZATION))
285
286    # (2) Merge text files with a known flag into the dictionary.
287    for flag in ALL_FLAGS:
288        for filename in args[flag]:
289            flags.assign_flag(flag, read_lines(filename), filename)
290
291    # Merge text files where conflicts should be ignored.
292    # This will only assign the given flag if:
293    # (a) the entry exists, and
294    # (b) it has not been assigned any other flag.
295    # Because of (b), this must run after all strict assignments have been performed.
296    for flag in ALL_FLAGS:
297        for filename in args[flag + FLAG_IGNORE_CONFLICTS_SUFFIX]:
298            valid_entries = flags.get_valid_subset_of_unassigned_apis(read_lines(filename))
299            flags.assign_flag(flag, valid_entries, filename)
300
301    # All members in the specified packages will be assigned the appropriate flag.
302    for flag in ALL_FLAGS:
303        for filename in args[flag + FLAG_PACKAGES_SUFFIX]:
304            packages_needing_list = set(read_lines(filename))
305            should_add_signature_to_list = lambda sig,lists: extract_package(
306                sig) in packages_needing_list and not lists
307            valid_entries = flags.filter_apis(should_add_signature_to_list)
308            flags.assign_flag(flag, valid_entries)
309
310    # Assign all remaining entries to the blacklist.
311    flags.assign_flag(FLAG_BLACKLIST, flags.filter_apis(HAS_NO_API_LIST_ASSIGNED))
312
313    # Write output.
314    write_lines(args["output"], flags.generate_csv())
315
316if __name__ == "__main__":
317    main(sys.argv)
318