mailprio
上传用户:xu_441
上传日期:2007-01-04
资源大小:1640k
文件大小:19k
- Received: from austin.bsdi.com (root{9l9gVDC7v8t3dlv0OtXTlby6X1zBWd56}@austin.BSDI.COM [205.230.224.49]) by knecht.Sendmail.ORG (8.8.2/8.8.2) with ESMTP id JAA05023 for <eric@sendmail.org>; Thu, 31 Oct 1996 09:29:47 -0800 (PST)
- Received: from austin.bsdi.com (localhost [127.0.0.1]) by austin.bsdi.com (8.7.4/8.7.3) with ESMTP id KAA19250; Thu, 31 Oct 1996 10:28:18 -0700 (MST)
- Message-Id: <199610311728.KAA19250@austin.bsdi.com>
- To: Eric Allman <eric@sendmail.org>
- cc: marc@xfree86.org
- Subject: Updated mailprio_0_93.shar
- From: Tony Sanders <sanders@earth.com>
- Organization: Berkeley Software Design, Inc.
- Date: Thu, 31 Oct 1996 10:28:14 -0700
- Sender: sanders@austin.bsdi.com
- Eric, please update contrib/mailprio in the sendmail distribution
- to this version at your convenience. Thanks.
- I've also made this available in:
- ftp://ftp.earth.com/pub/postmaster/
- mailprio_0_93.shar follows...
- #!/bin/sh
- # This is a shell archive (produced by GNU sharutils 4.1).
- # To extract the files from this archive, save it to some FILE, remove
- # everything before the `!/bin/sh' line above, then type `sh FILE'.
- #
- # Made on 1996-10-31 10:07 MST by <sanders@earth.com>.
- #
- # Existing files will *not* be overwritten unless `-c' is specified.
- #
- # This shar contains:
- # length mode name
- # ------ ---------- ------------------------------------------
- # 8260 -rwxr-xr-x mailprio
- # 3402 -rw-r--r-- mailprio.README
- # 4182 -rwxr-xr-x mailprio_mkdb
- #
- touch -am 1231235999 $$.touch >/dev/null 2>&1
- if test ! -f 1231235999 && test -f $$.touch; then
- shar_touch=touch
- else
- shar_touch=:
- echo
- echo 'WARNING: not restoring timestamps. Consider getting and'
- echo "installing GNU `touch', distributed in GNU File Utilities..."
- echo
- fi
- rm -f 1231235999 $$.touch
- #
- # ============= mailprio ==============
- if test -f 'mailprio' && test X"$1" != X"-c"; then
- echo 'x - skipping mailprio (file already exists)'
- else
- echo 'x - extracting mailprio (text)'
- sed 's/^X//' << 'SHAR_EOF' > 'mailprio' &&
- #!/usr/bin/perl
- #
- # mailprio,v 1.4 1996/10/31 17:03:52 sanders Exp
- # Version 0.93 -- Thu Oct 31 09:42:25 MST 1996
- #
- # mailprio -- setup mail priorities for a mailing list
- #
- # Copyright 1994, 1996, Tony Sanders <sanders@earth.com>
- # Rights are hereby granted to download, use, modify, sell, copy, and
- # redistribute this software so long as the original copyright notice
- # and this list of conditions remain intact and modified versions are
- # noted as such.
- #
- # I would also very much appreciate it if you could send me a copy of
- # any changes you make so I can possibly integrate them into my version.
- #
- # Options:
- # -p priority_database -- Specify database to use if not default
- # -q -- Process sendmail V8.8.X queue format files
- #
- # Sort mailing lists or sendmail queue files by mailprio database.
- # Files listed on the command line are locked and then sorted in place, in
- # the absence of any file arguments it will read STDIN and write STDOUT.
- #
- # Examples:
- # mailprio < mailing-list > sorted_list
- # mailprio mailing-list1 mailing-list2 mailing-list3 ...
- # mailprio -q /var/spool/mqueue/qf*
- # To double check results:
- # sort sorted_list > checkit; sort orig-mailing-list | diff - checkit
- #
- # To get the maximum value from a transaction delay based priority
- # function you need to reorder the distribution list (and the mail
- # queue files for that matter) fairly often; you could even have
- # your mailing list software reorder the list before each outgoing
- # message.
- #
- $usage = "Usage: mailprio [-p priodb] [-q] [mailinglists ...]n";
- $home = "/home/sanders/lists";
- $priodb = "$home/mailprio";
- $locking = "flock"; # "flock" or "fcntl"
- X
- # In shell, it would go more or less like this:
- # old_mailprio > /tmp/a
- # fgrep -f lists/inet-access /tmp/a | sed -e 's/^.......//' > /tmp/b
- # ; /tmp/b contains list of known users, faster delivery first
- # fgrep -v -f /tmp/b lists/inet-access > /tmp/c
- # ; put all unknown stuff at the top of new list for now
- # echo '# -----' >> /tmp/c
- # cat /tmp/b >> /tmp/c
- X
- $qflag = 0;
- while ($main'ARGV[0] =~ /^-/) {
- X $args = shift;
- X if ($args =~ m/?/) { print $usage; exit 0; }
- X if ($args =~ m/q/) { $qflag = 1; }
- X if ($args =~ m/p/) {
- X $priodb = shift || die $usage, "-p requires argumentn"; }
- }
- X
- push(@main'ARGV, '-') if ($#ARGV < 0);
- while ($file = shift @ARGV) {
- X if ($file eq "-") {
- X $source = "main'STDIN";
- X $sink = "main'STDOUT";
- X } else {
- X $sink = $source = "FH";
- X open($source, "+< $file") || do { warn "$file: $!n"; next; };
- X if (!defined &seize($source, &LOCK_EX | &LOCK_NB)) {
- X # couldn't get lock, just skip it
- X close($source);
- X next;
- X }
- X }
- X
- X local(*list);
- X &process($source, *list);
- X
- X # setup to write output
- X if ($file ne "-") {
- X # zero the file (FH is hardcoded because truncate requires it, sigh)
- X seek(FH, 0, 0) || die "$file: seek: $!n";
- X truncate(FH, 0) || die "$file: truncate: $!n";
- X }
- X
- X # do the dirty work
- X &output($sink, *list);
- X
- X close($sink) || warn "$file: $!n"; # close clears the lock
- X close($source);
- }
- X
- sub process {
- X # Setup %list and @list
- X local($source, *list) = @_;
- X local($addr, $canon);
- X while ($addr = <$source>) {
- X chop $addr;
- X next if $addr =~ /^# ----- /; # that's our line
- X push(@list, $addr), next if $addr =~ /^s*#/; # save comments
- X if ($qflag) {
- X next if $addr =~ m/^./;
- X push(@list, $addr), next if !($addr =~ s/^(R[^:]*:)//);
- X $Rflags = $1;
- X }
- X $canon = &canonicalize((&simplify_address($addr))[0]);
- X unless (defined $canon) {
- X warn "$file: no address found: $addrn";
- X push(@list, ($qflag?$Rflags:'') . $addr); # save it as is
- X next;
- X }
- X if (defined $list{$canon}) {
- X warn "$file: duplicate: ``$addr -> $canon''n";
- X push(@list, ($qflag?$Rflags:'') . $addr); # save it as is
- X next;
- X }
- X $list{$canon} = $addr;
- X }
- }
- X
- sub output {
- X local($sink, *list) = @_;
- X
- X local($to, *prio, *userprio, *useracct);
- X dbmopen(%prio, $priodb, 0644) || die "$priodb: $!n";
- X foreach $to (keys %list) {
- X if (defined $prio{$to}) {
- X # add to list of found users (%userprio) and remove from %list
- X # so that we know what users were not yet prioritized
- X $userprio{$to} = $prio{$to}; # priority
- X $useracct{$to} = $list{$to}; # string
- X delete $list{$to};
- X }
- X }
- X dbmclose(%prio);
- X
- X # Put all the junk we found at the very top
- X # (this might not always be a feature)
- X print $sink join("n", @list), "n" if int(@list);
- X
- X # prioritized list of users
- X if (int(keys %userprio)) {
- X print $sink '# ----- prioritized users', "n" unless $qflag;
- X foreach $to (sort by_userprio keys %userprio) {
- X die "Opps! Something is seriously wrong with useracct: $ton"
- X unless defined $useracct{$to};
- X print $sink 'RFD:' if $qflag;
- X print $sink $useracct{$to}, "n";
- X }
- X }
- X
- X # unprioritized users go last, fast accounts will get moved up eventually
- X # XXX: should go before the "really slow" prioritized users?
- X if (int(keys %list)) {
- X print $sink '# ----- unprioritized users', "n" unless $qflag;
- X foreach $to (keys %list) {
- X print $sink 'RFD:' if $qflag;
- X print $sink $list{$to}, "n";
- X }
- X }
- X
- X print $sink ".n" if $qflag;
- }
- X
- sub by_userprio {
- X # sort first by priority, then by key.
- X $userprio{$a} <=> $userprio{$b} || $a cmp $b;
- }
- X
- # REPL-LIB ---------------------------------------------------------------
- X
- sub canonicalize {
- X local($addr) = @_;
- X # lowercase, strip leading/trailing whitespace
- X $addr =~ y/A-Z/a-z/; $addr =~ s/^s+//; $addr =~ s/s+$//; $addr;
- }
- X
- # @addrs = simplify_address($addr);
- sub simplify_address {
- X local($_) = shift;
- X 1 while s/([^()]*)//g; # strip comments
- X 1 while s/"[^"]*"//g; # strip comments
- X split(/,/); # split into parts
- X foreach (@_) {
- X 1 while s/.*<(.*)>.*/1/;
- X s/^s+//;
- X s/s+$//;
- X }
- X @_;
- }
- X
- ### ---- ###
- #
- # Error codes
- #
- do 'errno.ph';
- eval 'sub ENOENT {2;}' unless defined &ENOENT;
- eval 'sub EINTR {4;}' unless defined &EINTR;
- eval 'sub EINVAL {22;}' unless defined &EINVAL;
- X
- #
- # File locking
- #
- do 'sys/unistd.ph';
- eval 'sub SEEK_SET {0;}' unless defined &SEEK_SET;
- X
- do 'sys/file.ph';
- eval 'sub LOCK_SH {0x01;}' unless defined &LOCK_SH;
- eval 'sub LOCK_EX {0x02;}' unless defined &LOCK_EX;
- eval 'sub LOCK_NB {0x04;}' unless defined &LOCK_NB;
- eval 'sub LOCK_UN {0x08;}' unless defined &LOCK_UN;
- X
- do 'fcntl.ph';
- eval 'sub F_GETFD {1;}' unless defined &F_GETFD;
- eval 'sub F_SETFD {2;}' unless defined &F_SETFD;
- eval 'sub F_GETFL {3;}' unless defined &F_GETFL;
- eval 'sub F_SETFL {4;}' unless defined &F_SETFL;
- eval 'sub O_NONBLOCK {0x0004;}' unless defined &O_NONBLOCK;
- eval 'sub F_SETLK {8;}' unless defined &F_SETLK; # nonblocking
- eval 'sub F_SETLKW {9;}' unless defined &F_SETLKW; # lockwait
- eval 'sub F_RDLCK {1;}' unless defined &F_RDLCK;
- eval 'sub F_UNLCK {2;}' unless defined &F_UNLCK;
- eval 'sub F_WRLCK {3;}' unless defined &F_WRLCK;
- $s_flock = "sslll"; # struct flock {type, whence, start, len, pid}
- X
- # return undef on failure
- sub seize {
- X local ($FH, $lock) = @_;
- X local ($ret);
- X if ($locking eq "flock") {
- X $ret = flock($FH, $lock);
- X return ($ret == 0 ? undef : 1);
- X } else {
- X local ($flock, $type) = 0;
- X if ($lock & &LOCK_SH) { $type = &F_RDLCK; }
- X elsif ($lock & &LOCK_EX) { $type = &F_WRLCK; }
- X elsif ($lock & &LOCK_UN) { $type = &F_UNLCK; }
- X else { $! = &EINVAL; return undef; }
- X $flock = pack($s_flock, $type, &SEEK_SET, 0, 0, 0);
- X $ret = fcntl($FH, ($lock & &LOCK_NB) ? &F_SETLK : &F_SETLKW, $flock);
- X return ($ret == -1 ? undef : 1);
- X }
- }
- SHAR_EOF
- $shar_touch -am 1031100396 'mailprio' &&
- chmod 0755 'mailprio' ||
- echo 'restore of mailprio failed'
- shar_count="`wc -c < 'mailprio'`"
- test 8260 -eq "$shar_count" ||
- echo "mailprio: original size 8260, current size $shar_count"
- fi
- # ============= mailprio.README ==============
- if test -f 'mailprio.README' && test X"$1" != X"-c"; then
- echo 'x - skipping mailprio.README (file already exists)'
- else
- echo 'x - extracting mailprio.README (text)'
- sed 's/^X//' << 'SHAR_EOF' > 'mailprio.README' &&
- mailprio README
- X
- mailprio.README,v 1.2 1996/10/31 17:03:54 sanders Exp
- Version 0.93 -- Thu Oct 31 09:42:25 MST 1996
- X
- Copyright 1994, 1996, Tony Sanders <sanders@earth.com>
- Rights are hereby granted to download, use, modify, sell, copy, and
- redistribute this software so long as the original copyright notice
- and this list of conditions remain intact and modified versions are
- noted as such.
- X
- I would also very much appreciate it if you could send me a copy of
- any changes you make so I can possibly integrate them into my version.
- X
- The current version of this and other related mail tools are available in:
- X ftp://ftp.earth.com/pub/postmaster/
- X
- Even with the new persistent host status in sendmail V8.8.X this
- function can still reduce the lag time distributing mail to a large
- group of people. It also makes it a little more likely that everyone
- will get mailing list mail in the order sent which can help reduce
- duplicate postings. Basically, the goal is to put slow hosts at
- the bottom of the list so that as many fast hosts are delivered
- as quickly as possible.
- X
- CONTENTS
- ========
- X
- X mailprio.README -- simple docs
- X mailprio -- the address sorter
- X mailprio_mkdb -- builds the database for the sorter
- X
- X
- CHANGES
- =======
- X Version 0.92
- X Initial public release.
- X
- X Version 0.93
- X Updated to make use of the (somewhat) new xdelay statistic.
- X Changed -q flag to support new sendmail queue file format (RFD:<addr>).
- X Fixed argument parsing bug.
- X Fixed bug with database getting "garbage" in it.
- X
- X
- CONFIGURATION
- =============
- X
- X You need to edit each script and ensure proper configuration.
- X
- X In mailprio check: #!perl path, $home, $priodb, $locking
- X
- X In mailprio_mkdb check: #!perl path, $home, $priodb, $maillog
- X
- X
- USAGE: mailprio
- ===============
- X
- X Usage: mailprio [-p priodb] [-q] [mailinglists ...]
- X -p priority_database -- Specify database to use if not default
- X -q -- Process sendmail queue format files
- X [USE WITH CAUTION]
- X
- X Sort mailing lists or sendmail V8 queue files by mailprio database.
- X Files listed on the command line are locked and then sorted in place, in
- X the absence of any file arguments it will read STDIN and write STDOUT.
- X
- X Examples:
- X mailprio < mailing-list > sorted_list
- X mailprio mailing-list1 mailing-list2 mailing-list3 ...
- X mailprio -q /var/spool/mqueue/qf* [not recommended]
- X To double check results:
- X sort sorted_list > checkit; sort orig-mailing-list | diff - checkit
- X
- X NOTE:
- X To get the maximum value from a transaction delay based priority
- X function you need to reorder the distribution list (and the mail
- X queue files for that matter) fairly often; you could even have
- X your mailing list software reorder the list before each outgoing
- X message.
- X
- X
- USAGE: mailprio_mkdb
- ====================
- X
- X Usage: mailprio_mkdb [-l maillog] [-p priodb]
- X -l maillog -- Specify maillog to process if not default
- X -p priority_database -- Specify database to use if not default
- X
- X Builds the mail priority database using information from the maillog.
- X
- X Run at least nightly before you rotate the maillog. If you are
- X going to run mailprio more often than that then you will need to
- X load the current maillog information before that will do any good
- X (and to keep from reloading the same information you will need
- X some kind of incremental maillog information to load from).
- SHAR_EOF
- $shar_touch -am 1031100396 'mailprio.README' &&
- chmod 0644 'mailprio.README' ||
- echo 'restore of mailprio.README failed'
- shar_count="`wc -c < 'mailprio.README'`"
- test 3402 -eq "$shar_count" ||
- echo "mailprio.README: original size 3402, current size $shar_count"
- fi
- # ============= mailprio_mkdb ==============
- if test -f 'mailprio_mkdb' && test X"$1" != X"-c"; then
- echo 'x - skipping mailprio_mkdb (file already exists)'
- else
- echo 'x - extracting mailprio_mkdb (text)'
- sed 's/^X//' << 'SHAR_EOF' > 'mailprio_mkdb' &&
- #!/usr/bin/perl
- #
- # mailprio_mkdb,v 1.5 1996/10/31 17:03:53 sanders Exp
- # Version 0.93 -- Thu Oct 31 09:42:25 MST 1996
- #
- # mailprio_mkdb -- make mail priority database based on delay times
- #
- # Copyright 1994, 1996, Tony Sanders <sanders@earth.com>
- # Rights are hereby granted to download, use, modify, sell, copy, and
- # redistribute this software so long as the original copyright notice
- # and this list of conditions remain intact and modified versions are
- # noted as such.
- #
- # I would also very much appreciate it if you could send me a copy of
- # any changes you make so I can possibly integrate them into my version.
- #
- # The average function moves the value around quite rapidly (half-steps)
- # which may or may not be a feature. This version uses the new xdelay
- # statistic (new as of sendmail V8) which is per transaction. We also
- # weight the result based on the overall delay.
- #
- # Something that might be worth doing for systems that don't support
- # xdelay would be to compute an approximation of the transaction delay
- # by sorting by messages-id and delay then computing the difference
- # between adjacent delay values.
- #
- # To get the maximum value from a transaction delay based priority
- # function you need to reorder the distribution list (and the mail
- # queue files for that matter) fairly often; you could even have
- # your mailing list software reorder the list before each outgoing
- # message.
- X
- $usage = "Usage: mailprio_mkdb [-l maillog] [-p priodb]n";
- $home = "/home/sanders/lists";
- $maillog = "/var/log/maillog";
- $priodb = "$home/mailprio";
- X
- while ($ARGV[0] =~ /^-/) {
- X $args = shift;
- X if ($args =~ m/?/) { print $usage; exit 0; }
- X if ($args =~ m/l/) {
- X $maillog = shift || die $usage, "-l requires argumentn"; }
- X if ($args =~ m/p/) {
- X $priodb = shift || die $usage, "-p requires argumentn"; }
- }
- X
- $SIG{'PIPE'} = 'handle_pipe';
- X
- # will merge with existing information
- dbmopen(%prio, $priodb, 0644) || die "$priodb: $!n";
- &getlog_stats($maillog, *prio);
- dbmclose(%prio);
- exit(0);
- X
- sub handle_pipe {
- X dbmclose(%prio);
- }
- X
- sub getlog_stats {
- X local($maillog, *stats) = @_;
- X local($to, $delay);
- X local($h, $m, $s);
- X open(MAILLOG, "< $maillog") || die "$maillog: $!n";
- X while (<MAILLOG>) {
- X next unless / to=/ && / stat=/;
- X next if / stat=queued/;
- X if (/ stat=sent/i) {
- X # read delay and xdelay and convert to seconds
- X ($delay) = (m/ delay=([^,]*),/);
- X next unless $delay;
- X ($h, $m, $s) = split(/:/, $delay);
- X $delay = ($h * 60 * 60) + ($m * 60) + $s;
- X
- X ($xdelay) = (m/ xdelay=([^,]*),/);
- X next unless $xdelay;
- X ($h, $m, $s) = split(/:/, $xdelay);
- X $xdelay = ($h * 60 * 60) + ($m * 60) + $s;
- X
- X # Now weight the delay factor by the transaction delay (xdelay).
- X $xdelay /= 300; # [0 - 1(@5 min)]
- X $xdelay += 0.5; # [0.5 - 1.5]
- X $xdelay = 1.5 if $xdelay > 1.5; # clamp
- X $delay *= $xdelay; # weight delay by xdelay
- X }
- X elsif (/, stat=/) {
- X # delivery failure of some sort (i.e. bad)
- X $delay = 432000; # force 5 days
- X }
- X $delay = 1000000 if $delay > 1000000;
- X
- X # filter the address(es); isn't perfect but is "good enough"
- X $to = $_; $to =~ s/^.* to=//;
- X 1 while $to =~ s/([^()]*)//g; # strip comments
- X 1 while $to =~ s/"[^"]*"//g; # strip comments
- X $to =~ s/, .*//; # remove other stat info
- X foreach $addr (&simplify_address($to)) {
- X next unless $addr;
- X $addr = &canonicalize($addr);
- X $stats{$addr} = $delay unless defined $stats{$addr}; # init
- X # pseudo-average in the new delay (half-steps)
- X # simple, moving average
- X $stats{$addr} = int(($stats{$addr} + $delay) / 2);
- X }
- X }
- X close(MAILLOG);
- }
- X
- # REPL-LIB ---------------------------------------------------------------
- X
- sub canonicalize {
- X local($addr) = @_;
- X # lowercase, strip leading/trailing whitespace
- X $addr =~ y/A-Z/a-z/; $addr =~ s/^s+//; $addr =~ s/s+$//; $addr;
- }
- X
- # @addrs = simplify_address($addr);
- sub simplify_address {
- X local($_) = shift;
- X 1 while s/([^()]*)//g; # strip comments
- X 1 while s/"[^"]*"//g; # strip comments
- X split(/,/); # split into parts
- X foreach (@_) {
- X 1 while s/.*<(.*)>.*/1/;
- X s/^s+//;
- X s/s+$//;
- X }
- X @_;
- }
- SHAR_EOF
- $shar_touch -am 1031100396 'mailprio_mkdb' &&
- chmod 0755 'mailprio_mkdb' ||
- echo 'restore of mailprio_mkdb failed'
- shar_count="`wc -c < 'mailprio_mkdb'`"
- test 4182 -eq "$shar_count" ||
- echo "mailprio_mkdb: original size 4182, current size $shar_count"
- fi
- exit 0