xref: /openbmc/openbmc/poky/meta/lib/oe/cve_check.py (revision f52e3dde)
1#
2# Copyright OpenEmbedded Contributors
3#
4# SPDX-License-Identifier: MIT
5#
6
7import collections
8import re
9import itertools
10import functools
11
12_Version = collections.namedtuple(
13    "_Version", ["release", "patch_l", "pre_l", "pre_v"]
14)
15
16@functools.total_ordering
17class Version():
18
19    def __init__(self, version, suffix=None):
20
21        suffixes = ["alphabetical", "patch"]
22
23        if str(suffix) == "alphabetical":
24            version_pattern =  r"""r?v?(?:(?P<release>[0-9]+(?:[-\.][0-9]+)*)(?P<patch>[-_\.]?(?P<patch_l>[a-z]))?(?P<pre>[-_\.]?(?P<pre_l>(rc|alpha|beta|pre|preview|dev))[-_\.]?(?P<pre_v>[0-9]+)?)?)(.*)?"""
25        elif str(suffix) == "patch":
26            version_pattern =  r"""r?v?(?:(?P<release>[0-9]+(?:[-\.][0-9]+)*)(?P<patch>[-_\.]?(p|patch)(?P<patch_l>[0-9]+))?(?P<pre>[-_\.]?(?P<pre_l>(rc|alpha|beta|pre|preview|dev))[-_\.]?(?P<pre_v>[0-9]+)?)?)(.*)?"""
27        else:
28            version_pattern =  r"""r?v?(?:(?P<release>[0-9]+(?:[-\.][0-9]+)*)(?P<pre>[-_\.]?(?P<pre_l>(rc|alpha|beta|pre|preview|dev))[-_\.]?(?P<pre_v>[0-9]+)?)?)(.*)?"""
29        regex = re.compile(r"^\s*" + version_pattern + r"\s*$", re.VERBOSE | re.IGNORECASE)
30
31        match = regex.search(version)
32        if not match:
33            raise Exception("Invalid version: '{0}'".format(version))
34
35        self._version = _Version(
36            release=tuple(int(i) for i in match.group("release").replace("-",".").split(".")),
37            patch_l=match.group("patch_l") if str(suffix) in suffixes and match.group("patch_l") else "",
38            pre_l=match.group("pre_l"),
39            pre_v=match.group("pre_v")
40        )
41
42        self._key = _cmpkey(
43            self._version.release,
44            self._version.patch_l,
45            self._version.pre_l,
46            self._version.pre_v
47        )
48
49    def __eq__(self, other):
50        if not isinstance(other, Version):
51            return NotImplemented
52        return self._key == other._key
53
54    def __gt__(self, other):
55        if not isinstance(other, Version):
56            return NotImplemented
57        return self._key > other._key
58
59def _cmpkey(release, patch_l, pre_l, pre_v):
60    # remove leading 0
61    _release = tuple(
62        reversed(list(itertools.dropwhile(lambda x: x == 0, reversed(release))))
63    )
64
65    _patch = patch_l.upper()
66
67    if pre_l is None and pre_v is None:
68        _pre = float('inf')
69    else:
70        _pre = float(pre_v) if pre_v else float('-inf')
71    return _release, _patch, _pre
72
73
74def get_patched_cves(d):
75    """
76    Get patches that solve CVEs using the "CVE: " tag.
77    """
78
79    import re
80    import oe.patch
81
82    cve_match = re.compile(r"CVE:( CVE-\d{4}-\d+)+")
83
84    # Matches the last "CVE-YYYY-ID" in the file name, also if written
85    # in lowercase. Possible to have multiple CVE IDs in a single
86    # file name, but only the last one will be detected from the file name.
87    # However, patch files contents addressing multiple CVE IDs are supported
88    # (cve_match regular expression)
89    cve_file_name_match = re.compile(r".*(CVE-\d{4}-\d+)", re.IGNORECASE)
90
91    patched_cves = set()
92    patches = oe.patch.src_patches(d)
93    bb.debug(2, "Scanning %d patches for CVEs" % len(patches))
94    for url in patches:
95        patch_file = bb.fetch.decodeurl(url)[2]
96
97        # Check patch file name for CVE ID
98        fname_match = cve_file_name_match.search(patch_file)
99        if fname_match:
100            cve = fname_match.group(1).upper()
101            patched_cves.add(cve)
102            bb.debug(2, "Found %s from patch file name %s" % (cve, patch_file))
103
104        # Remote patches won't be present and compressed patches won't be
105        # unpacked, so say we're not scanning them
106        if not os.path.isfile(patch_file):
107            bb.note("%s is remote or compressed, not scanning content" % patch_file)
108            continue
109
110        with open(patch_file, "r", encoding="utf-8") as f:
111            try:
112                patch_text = f.read()
113            except UnicodeDecodeError:
114                bb.debug(1, "Failed to read patch %s using UTF-8 encoding"
115                        " trying with iso8859-1" %  patch_file)
116                f.close()
117                with open(patch_file, "r", encoding="iso8859-1") as f:
118                    patch_text = f.read()
119
120        # Search for one or more "CVE: " lines
121        text_match = False
122        for match in cve_match.finditer(patch_text):
123            # Get only the CVEs without the "CVE: " tag
124            cves = patch_text[match.start()+5:match.end()]
125            for cve in cves.split():
126                bb.debug(2, "Patch %s solves %s" % (patch_file, cve))
127                patched_cves.add(cve)
128                text_match = True
129
130        if not fname_match and not text_match:
131            bb.debug(2, "Patch %s doesn't solve CVEs" % patch_file)
132
133    # Search for additional patched CVEs
134    for cve in (d.getVarFlags("CVE_STATUS") or {}):
135        decoded_status, _, _ = decode_cve_status(d, cve)
136        if decoded_status == "Patched":
137            bb.debug(2, "CVE %s is additionally patched" % cve)
138            patched_cves.add(cve)
139
140    return patched_cves
141
142
143def get_cpe_ids(cve_product, version):
144    """
145    Get list of CPE identifiers for the given product and version
146    """
147
148    version = version.split("+git")[0]
149
150    cpe_ids = []
151    for product in cve_product.split():
152        # CVE_PRODUCT in recipes may include vendor information for CPE identifiers. If not,
153        # use wildcard for vendor.
154        if ":" in product:
155            vendor, product = product.split(":", 1)
156        else:
157            vendor = "*"
158
159        cpe_id = 'cpe:2.3:*:{}:{}:{}:*:*:*:*:*:*:*'.format(vendor, product, version)
160        cpe_ids.append(cpe_id)
161
162    return cpe_ids
163
164def cve_check_merge_jsons(output, data):
165    """
166    Merge the data in the "package" property to the main data file
167    output
168    """
169    if output["version"] != data["version"]:
170        bb.error("Version mismatch when merging JSON outputs")
171        return
172
173    for product in output["package"]:
174        if product["name"] == data["package"][0]["name"]:
175            bb.error("Error adding the same package %s twice" % product["name"])
176            return
177
178    output["package"].append(data["package"][0])
179
180def update_symlinks(target_path, link_path):
181    """
182    Update a symbolic link link_path to point to target_path.
183    Remove the link and recreate it if exist and is different.
184    """
185    if link_path != target_path and os.path.exists(target_path):
186        if os.path.exists(os.path.realpath(link_path)):
187            os.remove(link_path)
188        os.symlink(os.path.basename(target_path), link_path)
189
190
191def convert_cve_version(version):
192    """
193    This function converts from CVE format to Yocto version format.
194    eg 8.3_p1 -> 8.3p1, 6.2_rc1 -> 6.2-rc1
195
196    Unless it is redefined using CVE_VERSION in the recipe,
197    cve_check uses the version in the name of the recipe (${PV})
198    to check vulnerabilities against a CVE in the database downloaded from NVD.
199
200    When the version has an update, i.e.
201    "p1" in OpenSSH 8.3p1,
202    "-rc1" in linux kernel 6.2-rc1,
203    the database stores the version as version_update (8.3_p1, 6.2_rc1).
204    Therefore, we must transform this version before comparing to the
205    recipe version.
206
207    In this case, the parameter of the function is 8.3_p1.
208    If the version uses the Release Candidate format, "rc",
209    this function replaces the '_' by '-'.
210    If the version uses the Update format, "p",
211    this function removes the '_' completely.
212    """
213    import re
214
215    matches = re.match('^([0-9.]+)_((p|rc)[0-9]+)$', version)
216
217    if not matches:
218        return version
219
220    version = matches.group(1)
221    update = matches.group(2)
222
223    if matches.group(3) == "rc":
224        return version + '-' + update
225
226    return version + update
227
228def decode_cve_status(d, cve):
229    """
230    Convert CVE_STATUS into status, detail and description.
231    """
232    status = d.getVarFlag("CVE_STATUS", cve)
233    if not status:
234        return ("", "", "")
235
236    status_split = status.split(':', 1)
237    detail = status_split[0]
238    description = status_split[1].strip() if (len(status_split) > 1) else ""
239
240    status_mapping = d.getVarFlag("CVE_CHECK_STATUSMAP", detail)
241    if status_mapping is None:
242        bb.warn('Invalid detail "%s" for CVE_STATUS[%s] = "%s", fallback to Unpatched' % (detail, cve, status))
243        status_mapping = "Unpatched"
244
245    return (status_mapping, detail, description)
246