2 # SPDX-License-Identifier: GPL-2.0-only
4 # Short version of this script:
5 # curl -f -o /var/cache/exim/opendmarc.tlds https://publicsuffix.org/list/public_suffix_list.dat
6 # but run as Exim runtime user, writing to a place it can write to, and with
7 # sanity checks and atomic replacement.
9 # For now, we deliberately leave the invalid file around for analysis
12 # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~8< cut here >8~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14 # Create a cron-job as the Exim run-time user to invoke this daily, with a
15 # single parameter, 'cron'. Eg:
17 # 3 4 * * * /usr/local/sbin/renew-opendmarc-tlds.sh cron
19 # That will, at 3 minutes past the 4th hour (in whatever timezone cron is
20 # running it) invoke this script with 'cron'; we will then sleep between 10 and
21 # 50 seconds, before continuing.
23 # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~8< cut here >8~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
25 # This should be "pretty portable"; the only things it depends upon are:
26 # * a POSIX shell which additionally implements 'local' (dash works)
27 # * the 'curl' command; change the fetch_candidate() function to replace that
28 # * the 'stat' command, to get the size of a file; else Perl
29 # + change size_of() if need be; it's defined per-OS
30 # * the 'hexdump' command and /dev/urandom existing
31 # + used when invoked with 'cron', to avoid retrieving on a minute boundary
32 # and contending with many other automated systems.
33 # + with bash/zsh, can replace with: $(( 10 + ( RANDOM % 40 ) ))
34 # + on Debian/Ubuntu systems, hexdump is in the 'bsdmainutils' package.
36 # Consider putting an email address inside the parentheses, something like
37 # noc@example.org or other reachable address, so that if something goes wrong
38 # and the server operators need to step in, they can see from logs who to
39 # contact instead of just blocking your IP:
40 readonly CurlUserAgent='renew-opendmarc-tlds/0.1 (distributed with Exim)'
42 # change this to your Exim run-time user (exim -n -bP exim_user) :
43 readonly RuntimeUser='_exim'
45 # Do not make this a directory which untrusted users can write to:
46 readonly StateDir='/var/cache/exim'
48 readonly URL='https://publicsuffix.org/list/public_suffix_list.dat'
50 readonly TargetShortFile='opendmarc.tlds'
52 # When replacing, new file must be at least this percentage the size of
53 # the old one or it's an error:
54 readonly MinNewSizeRation=90
56 # Each of these regexps must be matched by the file, or it's an error:
57 readonly MustExistRegexps='
63 # =======================8< end of configuration >8=======================
67 readonly FullTargetPath="${StateDir}/${TargetShortFile}"
68 readonly WorkingFile="${FullTargetPath}.$$"
70 progname="$(basename "$0")"
71 note() { printf >&2 '%s: %s\n' "$progname" "$*"; }
72 die() { note "$@"; exit 1; }
74 # guard against stomping on file-permissions
75 [ ".$(id -un)" = ".${RuntimeUser:?}" ] || \
76 die "must be invoked as ${RuntimeUser}"
79 curl --user-agent "$CurlUserAgent" -fSs -o "${WorkingFile}" "${URL}"
84 size_of() { stat -f %z "$1"; }
87 size_of() { stat -c %s "$1"; }
90 # why do we live in a world where Perl is the safe portable solution
91 # to getting the size of a file?
92 size_of() { perl -le 'print((stat($ARGV[0]))[7])' -- "$1"; }
96 sanity_check_candidate() {
97 local new_size prev_size re
98 new_size="$(size_of "$WorkingFile")"
100 for re in $MustExistRegexps; do
101 grep -qs "$re" -- "$WorkingFile" || \
102 die "regexp $re not found in $WorkingFile"
105 if ! prev_size="$(size_of "$FullTargetPath")"; then
106 note "missing previous file, can't size-compare: $FullTargetPath"
107 # We're sane by definition, probably initial fetch, and the
108 # stat failure and this note will be printed. That's fine; if
109 # a cron invocation is missing the file then something has gone
114 ratio=$(expr $new_size \* 100 / $prev_size)
115 if [ $ratio -lt $MinNewSizeRation ]; then
116 die "New $TargetShortFile candidate only ${ratio}% size of old; $new_size vs $prev_size"
120 if [ "${1:-.}" = "cron" ]; then
122 # Don't pull on-the-minute, wait for off-cycle-peak
123 sleep $(( ($(dd if=/dev/urandom bs=1 count=1 2>/dev/null | hexdump -e '1/1 "%u"') % 40) + 10))
128 sanity_check_candidate
129 mv -- "$WorkingFile" "$FullTargetPath"