On 02/29/2016 08:50 AM, Burton, Ross wrote:
On 24 February 2016 at 15:27, <mariano.lopez@linux.intel.com> wrote:

 
+do_cve_check[depends] = "cve-check-tool-native:do_populate_cve_db"


And cve-check-tool-native:do_populate_sysroot.

cve-check-tool-native:do_populate_cve_db depends on cve-check-tool-native:do_populate_sysroot, so adding it there would be redundant.
 
+def get_patches_cves(d):
+    """
+    Get patches that solve CVEs using the "CVE: " tag.
+    """
+
+    import re
+
+    pn = d.getVar("PN", True)
+    cve_match = re.compile("CVE:( CVE\-\d+\-\d+)+")

How does this work as the backslashes are escaping the - and d and d?  Use r"" strings.

The backslashes just escape the "-", the "d" is the same as with the raw string. I don't really see the need to use r"" here.


+    patched_cves = set()
+    for url in src_patches(d):
+        patch_file = bb.fetch.decodeurl(url)[2]
+        with open(patch_file, "r") as f:
+            patch_text = f.read()
+
+        # Search for the "CVE: " line
+        match = cve_match.search(patch_text)
+        if match:
+            # Get only the CVEs without the "CVE: " tag
+            cves = patch_text[match.start()+5:match.end()]
+            for cve in cves.split():
+                patched_cves.add(cve)


Breaks for patches such as this in glibc:

meta/recipes-core/glibc/glibc/CVE-2015-9761_1.patch:CVE: CVE-2015-9761 patch #1

I'd probably look for a line that starts with "CVE:" and the use re.findall to find all strings matching r"CVE-\d{4}-\d+"

What do you mean by break? It does catch the CVE just fine, to test it just revert the glibc 2.23 update. I find cleaner to match the string in a single operation instead of searching for the tag line by line and then match the CVEs.

+def get_cve_info(d, cves):
+    """
+    Get CVE information from the database used by cve-check-tool.
+    """
+
+    try:
+        import sqlite3
+    except ImportError:
+        from pysqlite2 import dbapi2 as sqlite3

Isn't the output from cve-check-tool good enough?  Would it be nicer to extend the log instead of assuming that the database format won't ever change?

The output from cve-check-tool is only the CVE number, if that is good enough, the query to the database can be removed.


+def cve_write_data(d, patched, unpatched, cve_data):
+    """
+    Write CVE information in WORKDIR; and to CVE_CHECK_DIR, and
+    CVE manifest if enabled.
+    """
+
+    from bb.utils import mkdirhier
+
+    cve_file = d.getVar("CVE_CHECK_LOCAL_FILE", True)
+    nvd_link = "https://web.nvd.nist.gov/view/vuln/detail?vulnId="
+    write_string = ""
+    mkdirhier(d.getVar("CVE_CHECK_LOCAL_DIR", True))
+
+    for cve in sorted(cve_data):
+        write_string += "PACKAGE NAME: %s\n" % d.getVar("PN", True)
+        write_string += "PACKAGE VERSION: %s\n" % d.getVar("PV", True)
+        write_string += "CVE: %s\n" % cve
+        if cve in patched:
+            write_string += "CVE STATUS: Patched\n"
+        else:
+            write_string += "CVE STATUS: Unpatched\n"
+            bb.warn("Found unpatched CVE, for more information check %s" % cve_file)
+        write_string += "CVE SUMMARY: %s\n" % cve_data[cve]["summary"]
+        write_string += "CVSS v2 BASE SCORE: %s\n" % cve_data[cve]["score"]
+        write_string += "VECTOR: %s\n" % cve_data[cve]["vector"]
+        write_string += "MORE INFORMATION: %s%s\n\n" % (nvd_link, cve)
+
+    with open(cve_file, "w") as f:
+        f.write(write_string)

Just write to the file instead of to a temporary string.

The temporary string is used for other two files, one could be copied, but the other appends the string content.


Ross

I have implemented the rest of the comments, just need your input before sending a new version.

Mariano