Merge From CURRENT ("MFC") notification service ("MFCns") for the FreeBSD Project.
At the some point of FreeBSD development it became apparent that in order to help developers to perform timely features merge from the developent branch (aka FreeBSD-CURRENT) into stable branch (aka FreeBSD-STABLE) some form of automatic system, which will track outstanding merges and notify appropriate developer when the time for a merge has come, is necessary. Otherwise, developers are often stick with two extreme cases: some of them forget to merge changes, even those that are clearly safe to merge, while some performing merge shorly after committing the change into a development branch fearing that if the merge is delayed they will forget to do it at all. Both those cases introduce problems to the Project, because in the former case stable branch experiences feature stagnation, while the latter leads to stability degradation, as the features are often being merged without proper testing in the development branch.
To address the problem the MFCns service was created. It works as follows: when committing a change to a development branch a developer specifies a time period during which the appropriate change has to be tested before being merged into a stable branch. The system notes this request and after this time period is over, it notifies the developer about that fact. It is important to emphasise that the final decision as to whether to merge change in question or not is up to the developer, because at that time he might have received an information that makes a merge undesirable (e.g. bug report from user who experieces problem with the change, security advisory from a security officer, negative feedback from other developers and so on).
When designing the MFCns the following goals were attacked:
- The service should be triggered by the presence of the special field in the commit message;
- due to the fact that it will have to parse all commit messages it should be speed-efficient during initial selection of messages, when it have to separate few messages which cointain that special field from a much larger number of those that do not;
- it should be robust enough to minimise possibility of the accidental match and provide some form of resilence, so that if the service is halted for some reason (e.g. machine maintenance, program error etc.) already queued notifications aren't lost and will be sent once the operation of the service is resumed.
The service is implemented in two parts. First part written in C is a
simple mail filter, which reads mail message (one a time) from the standard
input, looks for a set of patterns and if all conditions are satisfied
writes copy of message into a specially designated spool folder. To avoid
using locks the filter initially writes message into a temporary folder and
then moves it into a spool folder using rename(2) system call, which is
guranteed to be atomic operation. Obviously this means that both temporary
and spool folders have to reside on the same filesystem, so please keep
this in mind if you are going to use the service elsewhere. The name of the
file in the spool directory is generated using value in the "Message-ID"
field of the message, allowing to avoid duplicated notification if for
some reason more than one copy of the same commit message will be received
by the filter. Since e-mail messages is considered as an "unsafe" medium,
the filter puts significant restrictions on the part of the "Message-ID"
field to be used as the filename, so that it is impossible to use forged
message to overwrite a file outside of the spool directory.
The second part of the service does most of the job, which includes spool and queue processing and sending reminders. Since this part is not speed-critical (it only invoked once a day), it was implemented in Python. Its operation consists of the following phases:
- spool processing. At this stage the service parses all messages in the
spoolfolder, identifies the date at which each notification is to be delivered and moves messages into thequeuedirectory. In thequeuefolder messages are placed into subfolders named by the date when each notification is due (YYYYMMDD), so that all notifications which should be delivered at the same date end up in the same subfolder. Such design simplifies future queue processing as well as gurantees that each message is parsed only twice - first time when it is moved from thespooltoqueueto calculate the date when notification have to be sent and second time when the service actually sends a notification to identify to whom send it; - queue processing. At this stage the service inspects the
queuedirectory and identifies which notifications are ready to be sent. It parses each request from thereadysubset, identifies recepient, generates outgoing message and sends it out. Once notification is sent the notification request is deleted from thequeue.
Since e-mail is widely known as a not very safe medium, the service was designed with security in mind, however no formal security audit was performed yet. The part of the service written in C doesn't use any statically allocated string buffers, so it should not be vulnerable to various smash-the-stack attacks, while part written in Python was specially tailored to not pass any of the shell metacharacters to a subshell invoked for sending notification. All this makes a potential attack pretty meaningless, because all that an attacker could do is to make service sending fake MFC notifications, but he can do this even without any help from the service.
https://github.com/sobomax/mfc_notifications
If you have any questions/suggestions/requests please don't hesitate to contact Maxim Sobolev [email protected]. Also, if possible, drop me a line if you will find this code useful somewhere outside of the FreeBSD Project.
MFC Notification Service Statistics
-----------------------------------
Period: Wed May 9 00:22:02 2001 - Thu May 9 00:23:03 2019
Total number of notifications sent: 41356
Committer MFCs % total
--------- ---- -------
kib 2998 7.25
jhb 1544 3.73
hselasky 1462 3.54
ngie 1426 3.45
rwatson 1217 2.94
mav 1007 2.43
trasz 953 2.30
markj 938 2.27
pjd 922 2.23
avg 902 2.18
delphij 861 2.08
dim 724 1.75
sephe 703 1.70
tuexen 665 1.61
bz 652 1.58
brueffer 581 1.40
bdrewery 564 1.36
pfg 540 1.31
ae 539 1.30
des 509 1.23
marius 474 1.15
rmacklem 469 1.13
emaste 439 1.06
np 430 1.04
asomers 429 1.04
arybchik 409 0.99
gjb 400 0.97
eadler 390 0.94
sam 385 0.93
ume 357 0.86
alc 345 0.83
maxim 332 0.80
mm 311 0.75
mjacob 271 0.66
nwhitehorn 260 0.63
sobomax 257 0.62
jilles 229 0.55
brooks 211 0.51
bapt 210 0.51
yar 204 0.49
ru 201 0.49
dteske 196 0.47
sbruno 192 0.46
smh 189 0.46
jhibbits 186 0.45
rpaulo 183 0.44
luigi 182 0.44
emax 180 0.44
gad 180 0.44
cy 174 0.42
dchagin 171 0.41
gavin 170 0.41
ken 167 0.40
brian 161 0.39
cperciva 161 0.39
trociny 161 0.39
kevans 157 0.38
melifaro 157 0.38
simon 157 0.38
julian 154 0.37
imp 152 0.37
andrew 150 0.36
pluknet 144 0.35
mjg 136 0.33
gnn 134 0.32
kmacy 134 0.32
njl 133 0.32
keramida 131 0.32
rnoland 131 0.32
ed 130 0.31
kientzle 128 0.31
gshapiro 126 0.30
marcel 125 0.30
jimharris 122 0.29
avos 120 0.29
scottl 119 0.29
dumbbell 117 0.28
rstone 116 0.28
bms 115 0.28
jkim 115 0.28
andre 113 0.27
manu 113 0.27
vangyzen 109 0.26
csjp 108 0.26
truckman 107 0.26
dwmalone 106 0.26
attilio 104 0.25
yongari 103 0.25
edwin 102 0.25
schweikh 101 0.24
rrs 99 0.24
gonzo 98 0.24
jmg 98 0.24
glebius 97 0.23
bschmidt 92 0.22
kp 91 0.22
joerg 90 0.22
mlaier 90 0.22
dillon 89 0.22
silby 88 0.21
thompsa 88 0.21
kris 86 0.21
mckusick 86 0.21
lstewart 85 0.21
mmel 83 0.20
allanjude 82 0.20
hrs 80 0.19
slavash 77 0.19
flz 73 0.18
philip 72 0.17
grehan 71 0.17
neel 71 0.17
davidcs 69 0.17
iedowse 69 0.17
wblock 69 0.17
kadesai 66 0.16
brucec 65 0.16
bryanv 64 0.15
cognet 64 0.15
dds 64 0.15
matteo 63 0.15
remko 63 0.15
sevan 62 0.15
will 62 0.15
ache 61 0.15
eugen 61 0.15
loos 61 0.15
nyan 60 0.15
fabient 58 0.14
ariff 57 0.14
gibbs 56 0.14
suz 56 0.14
jh 55 0.13
kensmith 55 0.13
ceri 53 0.13
jamie 53 0.13
stas 53 0.13
oleg 50 0.12
adrian 48 0.12
qingli 48 0.12
roam 47 0.11
royger 47 0.11
tjr 46 0.11
amdmi3 45 0.11
n_hibma 45 0.11
wulf 45 0.11
bcr 44 0.11
cjc 44 0.11
antoine 43 0.10
oshogbo 43 0.10
se 42 0.10
archie 41 0.10
erj 41 0.10
jfv 41 0.10
mikeh 41 0.10
pkelsey 41 0.10
danger 40 0.10
hiren 40 0.10
murray 40 0.10
zec 40 0.10
gabor 39 0.09
jkoshy 39 0.09
jpaetzel 39 0.09
mdf 39 0.09
tijl 38 0.09
grog 37 0.09
jdp 37 0.09
jmallett 37 0.09
jtl 37 0.09
rodrigc 37 0.09
iwasaki 36 0.09
mbr 36 0.09
thomas 36 0.09
pav 35 0.08
vmaffione 35 0.08
fsu 34 0.08
kaiw 33 0.08
sanpei 33 0.08
bmah 32 0.08
dab 32 0.08
gallatin 32 0.08
netchild 32 0.08
simokawa 32 0.08
uqs 32 0.08
dfr 31 0.07
ambrisko 30 0.07
mtm 30 0.07
ray 30 0.07
fjoe 29 0.07
krion 29 0.07
roberto 29 0.07
avatar 28 0.07
davidxu 28 0.07
rik 28 0.07
zont 28 0.07
dexuan 27 0.07
pdeuskar 27 0.07
araujo 26 0.06
bp 26 0.06
jeff 26 0.06
semenu 26 0.06
davidch 25 0.06
marck 25 0.06
mike 25 0.06
ups 24 0.06
wollman 24 0.06
lidl 23 0.06
mux 23 0.06
scf 23 0.06
damien 22 0.05
kan 22 0.05
markus 22 0.05
mmacy 22 0.05
shurd 22 0.05
wilko 22 0.05
daichi 21 0.05
lulf 21 0.05
rgrimes 21 0.05
rpokala 21 0.05
slm 21 0.05
das 20 0.05
garga 20 0.05
jmmv 20 0.05
bde 19 0.05
jhay 19 0.05
jlh 19 0.05
joe 19 0.05
mp 19 0.05
raj 19 0.05
theraven 19 0.05
trhodes 19 0.05
weongyo 19 0.05
cem 18 0.04
phk 18 0.04
yokota 18 0.04
alfred 17 0.04
guido 17 0.04
harti 17 0.04
blackend 16 0.04
hm 16 0.04
kevlo 16 0.04
pirzyk 16 0.04
stevek 16 0.04
jah 15 0.04
pho 15 0.04
ssouhlal 15 0.04
tsoome 15 0.04
issyl0 14 0.03
jesper 14 0.03
mr 14 0.03
rse 14 0.03
cokane 13 0.03
matusita 13 0.03
ps 13 0.03
rafan 13 0.03
akiyama 12 0.03
andreast 12 0.03
gj 12 0.03
jon 12 0.03
kbyanc 12 0.03
mdodd 12 0.03
orion 12 0.03
rink 12 0.03
sheldonh 12 0.03
vanhu 12 0.03
asmodai 11 0.03
brd 11 0.03
dcs 11 0.03
eri 11 0.03
ivoras 11 0.03
jcamou 11 0.03
jch 11 0.03
johan 11 0.03
le 11 0.03
sumikawa 11 0.03
wes 11 0.03
zack 11 0.03
gordon 10 0.02
ian 10 0.02
rmh 10 0.02
stefanf 10 0.02
wkoszek 10 0.02
ygy 10 0.02
0mp 9 0.02
davide 9 0.02
fanf 9 0.02
kuriyama 9 0.02
pb 9 0.02
rea 9 0.02
tmm 9 0.02
zeising 9 0.02
badger 8 0.02
benno 8 0.02
chuck 8 0.02
darrenr 8 0.02
dwhite 8 0.02
erwin 8 0.02
green 8 0.02
jkh 8 0.02
jonathan 8 0.02
jwd 8 0.02
marcus 8 0.02
marks 8 0.02
phantom 8 0.02
randi 8 0.02
anholt 7 0.02
bsd 7 0.02
deischen 7 0.02
eik 7 0.02
garys 7 0.02
karels 7 0.02
mmokhi 7 0.02
peterj 7 0.02
sepotvin 7 0.02
tobez 7 0.02
bjk 6 0.01
den 6 0.01
dg 6 0.01
dhartmei 6 0.01
feld 6 0.01
fenner 6 0.01
gahr 6 0.01
keichii 6 0.01
markm 6 0.01
msmith 6 0.01
mw 6 0.01
olgeni 6 0.01
olli 6 0.01
ram 6 0.01
sef 6 0.01
syrinx 6 0.01
whu 6 0.01
yuripv 6 0.01
bwidawsk 5 0.01
eric 5 0.01
ghelmer 5 0.01
greid 5 0.01
hmp 5 0.01
jlemon 5 0.01
mckay 5 0.01
mizhka 5 0.01
mnag 5 0.01
nsayer 5 0.01
peadar 5 0.01
rdivacky 5 0.01
robak 5 0.01
seanc 5 0.01
bruno 4 0.01
cel 4 0.01
dannyboy 4 0.01
grembo 4 0.01
imura 4 0.01
jceel 4 0.01
jgh 4 0.01
jinmei 4 0.01
knu 4 0.01
lwhsu 4 0.01
matthew 4 0.01
mjoras 4 0.01
nectar 4 0.01
olivier 4 0.01
piso 4 0.01
rsm 4 0.01
skra 4 0.01
takawata 4 0.01
ticso 4 0.01
tom 4 0.01
wpaul 4 0.01
assar 3 0.01
benjsc 3 0.01
bmilekic 3 0.01
chinsan 3 0.01
dmlb 3 0.01
emoore 3 0.01
groudier 3 0.01
jedgar 3 0.01
lioux 3 0.01
mini 3 0.01
phil 3 0.01
rakuco 3 0.01
rees 3 0.01
robert 3 0.01
ade 2 0.00
ale 2 0.00
amorita 2 0.00
babkin 2 0.00
billf 2 0.00
bland 2 0.00
brix 2 0.00
cg 2 0.00
chris 2 0.00
crees 2 0.00
dbaker 2 0.00
dd 2 0.00
gleb 2 0.00
jb 2 0.00
jmcneill 2 0.00
lme 2 0.00
luoqi 2 0.00
motoyuki 2 0.00
nork 2 0.00
osa 2 0.00
snb 2 0.00
sos 2 0.00
svn 2 0.00
tabthorpe 2 0.00
tmunro 2 0.00
zml 2 0.00
adamw 1 0.00
alex 1 0.00
anish 1 0.00
art 1 0.00
beat 1 0.00
ben 1 0.00
chm 1 0.00
dan 1 0.00
danny 1 0.00
davidn 1 0.00
dbaio 1 0.00
dru 1 0.00
edavis 1 0.00
gber 1 0.00
hsu 1 0.00
jasone 1 0.00
jim 1 0.00
jmb 1 0.00
jmz 1 0.00
johalun 1 0.00
kato 1 0.00
lawrance 1 0.00
lifanov 1 0.00
matk 1 0.00
mhorne 1 0.00
miwi 1 0.00
mpp 1 0.00
mwlucas 1 0.00
novel 1 0.00
pgj 1 0.00
pst 1 0.00
rene 1 0.00
riggs 1 0.00
rushani 1 0.00
sjg 1 0.00
skreuzer 1 0.00
tackerman 1 0.00
tanimura 1 0.00
thj 1 0.00
tomsoft 1 0.00
twinterg 1 0.00
ue 1 0.00
versus 1 0.00
vsevolod 1 0.00
xride 1 0.00
-- Maxim Sobolev [email protected]