coverity-email-report-parse

#!/usr/bin/env python3
# Copyright 2019 Kees Cook <keescook@chromium.org>
# License: GPLv2+
#
# Coverity produces way too many false positives for this automation to
# actually be useful. e.g. IS_ERR() constantly triggers false UAF warnings
# because coverity can't tell that non-NULL doesn't mean unallocated.
#
# Regardless, this will perform "git blame" analysis on reported Coverity
# failures and direct the warning to the identified authors and related
# people. (And is currently designed to only get run by me for linux-next.)
#
import os, sys, re
import subprocess, tempfile

# ** CID 1487378:  Null pointer dereferences  (REVERSE_INULL)
#
# _____
# *** CID nnnn: reason (name)
# /net/dsa/dsa2.c: 849 in dsa_switch_probe()
# nnn code

# Ignore "^** CID"
# Collect all CIDs/files between ^___ lines

if len(sys.argv) > 1:
	report = sys.argv[1]
	fd = open(report, "r")
else:
	report = "stdin"
	fd = sys.stdin

me = 'keescook+coverity-bot@chromium.org'

ready = False
path_next = False
hits = []
concat = ""

cid = None
reason = None


# curl -sS https://scan.coverity.com/projects/linux-next-weekly-scan | grep "Version:"
#  <p>Version: next-20191025</p>
proc = subprocess.Popen(['curl', '-sS', 'https://scan.coverity.com/projects/linux-next-weekly-scan'], stdout = subprocess.PIPE)
output = proc.stdout.read().decode("utf-8")
match = re.search(r'Version:\s+([^<]+)<', output)
if not match:
	raise ValueError("Bad version from Coverity project page")
branch = match.group(1)


def commit_from_line(file_name, code_line):
	proc = subprocess.Popen(['git', 'blame', '-lL', code_line+','+code_line, branch, '--', file_name], stdout = subprocess.PIPE)
	output = proc.stdout.read().decode("utf-8")
	return output.split(' ', 1)[0]

def parse_commit(commit):
	print(f"Parsing commit {commit} ...")
	proc = subprocess.Popen(['git', 'log', '-1', commit], stdout=subprocess.PIPE)
	output = proc.stdout.read().decode("utf-8")
	author = None
	subject = None
	date = None
	others = set()
	for line in output.splitlines():
		line = line.rstrip()
		if line.startswith('Author: '):
			author = line.split(" ", 1)[1].strip()
			continue
		if line.startswith('Date: ') or line.startswith('CommitDate: ') or line.startswith('AuthorDate: '):
			date = line.split(" ", 1)[1].strip()
			continue
		if subject == None:
			match = re.search(r'^    (.*)$', line)
			if match:
				subject = match.group(1)
				continue
		match = re.search(r'^    .*-by:\s+(.*@.*)$', line)
		if match:
			others.add(match.group(1))
		match = re.search(r'^    Cc:\s+(.*@.*)$', line)
		if match:
			others.add(match.group(1))

	# Gather full list of people based on get_maintainers.pl too.
	patch = subprocess.Popen(['git', 'show', commit], stdout=subprocess.PIPE)
	output = subprocess.check_output(['./scripts/get_maintainer.pl',
					  '--email',
					  '--no-rolestats'], stdin=patch.stdout).decode('utf-8')
	patch.wait()
	for line in output.splitlines():
		others.add(line.strip())

	if author in others:
		others.remove(author)
	return subject, date, author, others

def send_email():
	info = set()
	funcs = set()
	cids = set()
	issues = set()
	for file_name,file_offset,function,cid,reason in hits:
		info.add("%s %s in %s: %s (CID %s)" % (file_name, file_offset, function, reason, cid))

		funcs.add(function)
		reason = reason.strip()
		if reason.startswith("("):
			reason = reason[1:len(reason)-1]
		else:
			reason = reason.split(" (", 1)[0]
		reason = reason.strip()
		if len(reason) == 0:
			raise ValueError("empty Coverity reason")
		cids.add('%s ("%s")' % (cid, reason))

		issues.add('%s: %s' % (function, reason))

	if len(issues) > 1:
		subject = "Multiple issues in %s" % (", ".join(funcs))
	else:
		for issue in issues:
			subject = issue
	subject = "Coverity: %s" % (subject)

	body = "From: coverity-bot <%s>\n\n" % (me)
	body += """Hello!

This is an experimental semi-automated report about issues detected by
Coverity from a scan of %s as part of the linux-next scan project:
https://scan.coverity.com/projects/linux-next-weekly-scan

You're getting this email because you were associated with the identified
lines of code (noted below) that were touched by commits:

""" % (branch)

	seen = set()
	commits = set()
	for file_name,file_offset,function,cid,reason in hits:
		what = "%s %s" % (file_name, file_offset)
		if what in seen:
			continue
		seen.add(what)
		commit = commit_from_line(file_name, file_offset)
		commits.add(commit)

	to = set()
	cc = set()
	fixes = []
	for commit in commits:
		name, date, author, others = parse_commit(commit)
		to.add(author)
		cc |= others
		fix = '%s ("%s")' % (commit[0:12], name)
		fixes.append(fix)
		body += "  %s\n    %s\n" % (date, fix)

	body += "\nCoverity reported the following:\n\n"
	body += concat

	body += """
If this is a false positive, please let us know so we can mark it as
such, or teach the Coverity rules to be smarter. If not, please make
sure fixes get into linux-next. :) For patches fixing this, please
include these lines (but double-check the "Fixes" first):

Reported-by: coverity-bot <%s>
""" % (me)

	for cid in cids:
		body += 'Addresses-Coverity-ID: %s\n' % (cid)
	for fix in fixes:
		body += 'Fixes: %s\n' % (fix)

	body += """
Thanks for your attention!

-- 
Coverity-bot
"""

	cc = list(cc)
	cc.append("Gustavo A. R. Silva <gustavo@embeddedor.com>")
	cc.append("linux-next@vger.kernel.org")
	cc.append("linux-hardening@vger.kernel.org")

	template = tempfile.NamedTemporaryFile(mode='w', prefix='mutt-', encoding='utf-8')
	template.write(body)
	template.flush()

	cc_args = []
	for person in cc:
		cc_args += ['-c', person]
	cmd = ['mutt-batch'] + cc_args + ['-H', template.name, '-b', me, '-s', subject, '-e', 'unset signature'] + list(to)

	print("To: %s" % (",\n\t".join(to)))
	if len(cc) > 0:
		print("Cc: %s" % (",\n\t".join(cc)))
	print("Subject: %s\n" % (subject))
	print("%s\n" % (body))

	want = ''
	while not want in ['e','s','k','q']:
		print("(E)dit, (S)end, s(K)ip, (Q)uit?")
		want = sys.stdin.readline().strip().lower()
	if want == 'q':
		print("Quit")
		sys.exit(0)
	if want == 'e':
		# Send with edit:
		print("Edit email")
		os.spawnvp(os.P_WAIT, 'mutt', cmd)
		return
	if want == 'k':
		print("Skip email")
		return
	if want == 's':
		print("Send email")
		# To auto-send:
		proc = subprocess.Popen(cmd, stdout=subprocess.PIPE, stdin=open("/dev/null", "r"))
		output = proc.stdout.read().decode("utf-8")
		if len(output) != 0:
			print(output)
		return
	raise ValueError("Impossible want value: '%s'" % (want))

line_num = 0
for line in fd:
	line_num += 1
	line = line.rstrip()

	# Flush reporting queue!
	if line.startswith("_________"):
		if not ready:
			ready = True

		if len(hits) == 0:
			continue

		send_email()

		hits = []
		concat = ""
		continue

	if line.startswith('** CID '):
		ready = False
		continue

	if not ready:
		continue

	if len(line.strip()) == 0:
		continue

	# Avoid MUAs hiding lines starting with ">" as if they were quotes.
	if line.startswith('>>>'):
		line = line.replace(">>>", "vvv", 1)

	match = re.search(r'^\*\*\* CID (\d+):\s+(.*)$', line)
	if match:
		cid = match.group(1)
		reason = match.group(2)
		concat += line + "\n"
		continue

	match = re.search(r'^/(\S+):\s+(\d+)\s+in\s+(.*)$', line)
	if match:
		file_name = match.group(1)
		file_offset = match.group(2)
		function = match.group(3)
		hits.append( (file_name, file_offset, function, cid, reason) )
		concat += f"{file_name}:{file_offset} in {function}\n"
		continue

	concat += line + "\n"