fixes and some improvements
[distro-setup] / mailtest-check
1 #!/bin/bash
2
3 # Usage: mail-test-check [slow] [anything]
4 #
5 # slow: do slow checks, like spamassassin
6 #
7 # anything: consider non-interactive, dont print unless something went
8 # wrong
9
10
11 source /b/errhandle/err
12
13 [[ $EUID == 0 ]] || exec sudo -E "${BASH_SOURCE[0]}" "$@"
14
15 shopt -s nullglob
16
17 e() { $int || return 0; printf "mailtest-check: %s\n" "$*"; }
18
19 getspamdpid() {
20 if [[ ! $spamdpid || ! -d /proc/$spamdpid ]]; then
21 # try twice in case we are restarting, it happens.
22 for i in 1 2; do
23 spamdpid=$(systemctl show --property MainPID --value spamassassin | sed 's/^[10]$//' ||:)
24 if [[ $spamdpid ]]; then
25 break
26 fi
27 sleep 30
28 done
29 fi
30 }
31 pr() {
32 if $doprom && [[ -e /var/lib/prometheus/node-exporter ]]; then
33 cat >>/var/lib/prometheus/node-exporter/mailtest-check.prom.$$
34 fi
35 }
36
37
38 #### begin arg processing ####
39 # spamassassin checking takes about 8 seconds.
40 slow=false
41 if [[ $1 == slow ]]; then
42 slow=true
43 shift
44 fi
45
46 int=false
47 if [[ $SUDO_USER || $SSH_CONNECTION ]]; then
48 int=true
49 fi
50
51 if [[ $1 == int ]]; then
52 int=true
53 fi
54
55 if [[ $1 == nonint ]]; then
56 int=false
57 fi
58 #### end arg processing ####
59
60 # we put this in to avoid dns errors that happen on reboot,
61 # but I want to debug them.
62 # if ! $int; then
63 # sleep 60
64 # fi
65
66
67 # TODO, get je to deliver the local mailbox: /m/md/INBOX
68 # dovecot appears to setup, i can t be sure.
69
70 source /a/bin/bash_unpublished/source-state
71
72 doprom=false
73 case $HOSTNAME in
74 $MAIL_HOST|bk|je)
75 doprom=true
76 ;;
77 *)
78 rm -f /var/lib/prometheus/node-exporter/mailtest-check.prom*
79 ;;
80 esac
81
82 main() {
83
84 case $HOSTNAME in
85 bk)
86 folders=(/m/md/{expertpathologyreview.com,amnimal.ninja}/testignore)
87 froms=(ian@iankelling.org z@zroe.org testignore@je.b8.nz iank@gnu.org)
88 ;;
89 je)
90 froms=(ian@iankelling.org z@zroe.org iank@gnu.org testignore@amnimal.ninja)
91 folders=(/m/md/je.b8.nz/testignore)
92 ;;
93 *)
94 folders=(/m/md/l/testignore)
95 froms=(testignore@je.b8.nz testignore@expertpathologyreview.com testignore@amnimal.ninja ian@iankelling.org z@zroe.org iank@gnu.org)
96 if ! $int; then
97 ### begin rsyncing fencepost email ###
98 # We dont want to exit if rsync fails, that will get caught by
99 # our later test by virtue of not having the latest email.
100 did_rsync=false
101 try_start_time=$EPOCHSECONDS
102 try_limit=140 # somewhat arbitrary value
103 while ! $did_rsync; do
104 try_left=$(( try_limit - ( EPOCHSECONDS - try_start_time) ))
105 timeout=120 # somewhat arbitrary value
106 if (( try_left < 0 )); then
107 echo "mailtest-check: failed to rsync fencepost > $try_limit seconds"
108 break
109 fi
110 if (( try_left < timeout )); then
111 timeout=$try_left
112 fi
113 if timeout $timeout rsync --chown iank:iank -e "ssh -oIdentitiesOnly=yes -F /dev/null -i /root/.ssh/jtuttle" -t --inplace -r 'jtuttle@fencepost.gnu.org:/home/j/jtuttle/Maildir/new/' /m/md/l/testignore/new; then
114 did_rsync=true
115 else
116 sleep 4
117 fi
118 done
119 if ! $did_rsync; then
120 echo mailtest-check: warning: fencepost rsync failed
121 fi
122 ### end rsyncing fencepost email ###
123 fi
124 ;;
125 esac
126
127
128 # avoid errors like this:
129 # Nov 8 08:16:05.439 [6080] warn: plugin: failed to parse plugin (from @INC): Can't locate Mail/SpamAssassin/Plugin/WLBLEval.pm: lib/Mail/SpamAssassin/Plugin/WLBLEval.pm: Permission denied at (eval 59) line 1.
130 #Nov 8 08:16:05.439 [6080] warn: plugin: failed to parse plugin (from @INC): Can't locate Mail/SpamAssassin/Plugin/VBounce.pm: lib/Mail/SpamAssassin/Plugin/VBounce.pm: Permission denied at (eval 60) line 1.
131 # i dont know why, i just found the solution online
132 cd /m/md
133
134
135 getspamdpid
136 # first time we write, overwrite anything existing
137 if [[ -e /var/lib/prometheus/node-exporter ]]; then
138 cat >/var/lib/prometheus/node-exporter/mailtest-check.prom.$$ <<EOF
139 mailtest_check_found_spamd_pid_bool $(( ${spamdpid:-0} > 0 ))
140 EOF
141 fi
142 e spamdpid: $spamdpid
143 if [[ ! $spamdpid ]]; then
144 echo mailtest spamd pid not found. systemctl status spamassassin:
145 systemctl status spamassassin
146 fi
147 tmpfile=$(mktemp)
148 declare -i unexpected=0
149 for folder in ${folders[@]}; do
150 for from in ${froms[@]}; do
151 declare -i missing_dnswl=0
152 declare -i dnsfail=0
153 declare -i unexpected=0
154 latest=
155 last_sec=0
156
157 if ! grep -rlFx "From: $from" $folder/{new,cur} >$tmpfile; then
158 echo "no message found from: $from"
159 continue
160 fi
161 # webmail sends them to cur it seems
162 while read -r file; do
163 file_sec=$(awk '/^Subject: / {print $4}' $file)
164 if [[ $file_sec ]] && (( file_sec > last_sec )); then
165 latest=$file
166 last_sec="$file_sec"
167 fi
168 done <$tmpfile
169 rm -f $tmpfile
170
171 to=$(awk '/^Envelope-to: / {print $2}' $latest)
172
173 if $slow; then
174 if ! $int; then
175 find $folder/new $folder/cur -type f -mmin +1080 -delete
176 fi
177 getspamdpid
178 if [[ $spamdpid ]]; then
179 if [[ $(readlink /proc/$$/ns/net) != "$(readlink /proc/$spamdpid/ns/net)" ]]; then
180 spamcpre="nsenter -t $spamdpid -n -m"
181 fi
182 unset results
183 declare -A results
184 # pyzor fails for our test message, so dont put useless load on their
185 # servers.
186 # example line that sed is parsing:
187 # (-0.1 / 5.0 requ) DKIM_SIGNED=0.1,DKIM_VALID=-0.1,DKIM_VALID_AU=-0.1,SPF_HELO_PASS=-0.001,SPF_PASS=-0.001,TVD_SPACE_RATIO=0.001 autolearn=_AUTOLEARN
188 resultfile=$(mktemp)
189 $spamcpre sudo -u Debian-exim spamassassin -D -t --cf='score PYZOR_CHECK 0' <"$latest" &>$resultfile
190
191 # note: on some mail, its 1 line after the send-test-forward, on others its 2 with a blank inbetween.
192 # I use the sed -n to filter this.
193 raw_results="$(tail $resultfile | grep -A2 -Fx /usr/local/bin/send-test-forward | tail -n+2 | sed -nr 's/^\([^)]*\) *//;s/=[^, ]*([, ]|$)/ /gp')"
194 for r in $raw_results; do
195 case $r in
196 # got this in an update 2022-01. dun care
197 T_SCC_BODY_TEXT_LINE|SCC_BODY_SINGLE_WORD) : ;;
198 # we have a new domain, ignore this.
199 # it seems like some versions of spamassassin do BODY_SINGLE_WORD, others dont, we dun care.
200 # bayes_00 is a new one indicating ham, we dont care if its missing.
201 BAYES_00|BODY_SINGLE_WORD|FROM_FMBLA_NEWDOM*|autolearn) : ;;
202 SPF_HELO_NEUTRAL)
203 # some of my domains use neutral spf, treat them the same.
204 results[SPF_HELO_PASS]=t
205 ;;
206 *)
207 results[$r]=t
208 ;;
209 esac
210 done
211 # debugging
212 # e results = ${!results[@]}
213 missing=()
214
215 keys=(DKIM_SIGNED DKIM_VALID{,_AU,_EF} SPF_HELO_PASS SPF_PASS TVD_SPACE_RATIO)
216 if [[ $to == *@gnu.org && $from == *@gnu.org ]]; then
217 keys=(ALL_TRUSTED TVD_SPACE_RATIO)
218 elif [[ $to == *@gnu.org ]]; then
219 # eggs has RCVD_IN_DNSWL_MED
220 keys+=(RCVD_IN_DNSWL_MED)
221 elif [[ $from == *@gnu.org ]]; then
222 # eggs has this. it used to have DKIMWL_WL_HIGH sometime in 2022
223 keys+=(RCVD_IN_DNSWL_MED)
224 fi
225
226 for t in ${keys[@]}; do
227 if [[ ${results[$t]} ]]; then
228 unset "results[$t]"
229 elif [[ $t == DKIM_VALID_EF && $from == *@[^.]*.[^.]*.[^.]* ]]; then
230 :
231 # third level domains dont hit this. its because
232 # /usr/share/perl5/Mail/SpamAssassin/Plugin/DKIM.pm checks
233 # if its signed with the registryboundaries domain. afaik:
234 # we need the actual domain to sign it, this would result in
235 # a second signature. I only use second level domains for
236 # testing atm, fsf doesnt use them for anything but the
237 # forum and I dont expect that to have any deliverability
238 # problems. So, not bothering atm.
239 else
240 missing+=($t)
241 fi
242 done
243 if (( ${#results[@]} || ${#missing[@]} )); then
244 printf "$HOSTNAME spamtest %s\n" "$latest"
245 if (( ${#results[@]} )); then
246 printf "unexpected %s" "${!results[*]} "
247 fi
248 if (( ${#missing[@]} )); then
249 printf "missing %s" "${missing[*]}"
250 fi
251 echo # ends our printf string buildup
252 cat $resultfile
253 echo mailtest-check: end of spam debug results
254 # lets just handle 1 failure at a time in interactive mode.
255 if $int; then
256 echo mailtest-check: from: $from, to: $to
257 exit 0
258 fi
259
260 # less verbose debug output, commented since I might want it another time.
261 # if $int; then
262 # echo mailtest-check: cat $latest:
263 # cat $latest
264 # echo mailtest-check: end of cat
265 # echo "$(tput setaf 5 2>/dev/null ||:)█$(tput sgr0 2>/dev/null||:)%.0s" $(eval echo "{1..${COLUMNS:-60}}")
266 #fi
267 fi
268 rm -f $resultfile
269 for r in ${results[@]}; do
270 case $r in
271 # iank: for when we want to handle dns errors differently
272 # DKIM_INVALID|T_SPF_TEMPERROR|T_SPF_HELO_TEMPERROR)
273 # dnsfail+=1
274 # ;;
275 *)
276 unexpected=$(( unexpected + 1 ))
277 ;;
278 esac
279 done
280 for miss in ${missing[@]}; do
281 # We expect dns failures from time to time, so
282 # we count them separately and alert differently.
283 case $miss in
284 # iank: dns fail
285 # DKIM_VALID|DKIM_VALID_AU|DKIM_VALID_EF|SPF_HELO_PASS|SPF_PASS|
286 RCVD_IN_DNSWL_MED|DKIMWL_WL_HIGH)
287 missing_dnswl+=1
288 ;;
289 *)
290 unexpected+=1
291 ;;
292 esac
293 done
294 pr <<EOF
295 mailtest_check_missing_dnswl{folder="$folder",from="$from"} $missing_dnswl
296 mailtest_check_unexpected_spamd_results{folder="$folder",from="$from"} $unexpected
297 EOF
298 fi # if spamdpid
299 fi # if $slow
300
301 now=$EPOCHSECONDS
302 age_sec=$(( now - last_sec ))
303 e $((age_sec / 60)):$(( age_sec % 60 )) ago. to:$to from:$from $latest
304
305 # usec = unix seconds
306 pr <<EOF
307 mailtest_check_last_usec{folder="$folder",from="$from"} $last_sec
308 EOF
309 done # end for from in ${froms[@]}
310 done # end for folder in ${folders[@]}
311
312 dir=/var/lib/prometheus/node-exporter
313 if [[ -e $dir ]]; then
314 mv $dir/mailtest-check.prom.$$ $dir/mailtest-check.prom
315 # note: node_textfile_mtime_seconds will tell us when this last happened. useful for debugging.
316 fi
317 }
318
319 loop-main() {
320 # When running under systemd, the system just started. Ve nice and
321 # give programs some time to finish their startup.
322 sleep 10
323 while true; do
324 premain_sec=$EPOCHSECONDS
325 main
326 sleep $(( 300 - ( EPOCHSECONDS - premain_sec ) ))
327 done
328 }
329
330
331 if [[ $INVOCATION_ID ]]; then
332 loop-main
333 else
334 main
335 fi
336
337 exit 0