1 WAN, 2 "LAN"



  • In reality, I have 1 WAN, 1 LAN, and 1 OPT.  I have two clients in one building sharing a firewall.  One is on the LAN, the other is on OPT1.

    I create a bunch of standard rules using the shaper wizard, then go back and create a duplicate parent queue for the opt1 interface, a duplicate def queue, and a few other misc queues, then set rules to use those queues. I called the interface CL.

    When, upon attempting a reload I get errors like this:

    
    a:1:{i:1156286665;a:5:{s:2:"id";s:11:"filter_load";s:6:"notice";s:605:"There were error(s) loading the rules: /tmp/rules.debug:25: syntax error
    /tmp/rules.debug:29: queue qCLRoot has no parent
    /tmp/rules.debug:29: errors in queue definition
    /tmp/rules.debug:32: queue qCLdef has no parent
    /tmp/rules.debug:32: errors in queue definition
    /tmp/rules.debug:35: queue qCLacks has no parent
    /tmp/rules.debug:35: errors in queue definition
    /tmp/rules.debug:37: queue qVOIPDown has no parent
    /tmp/rules.debug:37: errors in queue definition
    pfctl: Syntax error in config file: pf rules not loaded The line in question reads [25]: altq on em2 hfsc bandwidth  queue { qlanRoot qCLRoot }";s:3:"url";s:0:"";s:8:"category";s:13:"Filter Reload";s:8:"priority";i:1;}}
    
    

    lines 22-25 looks like this:

    
    scrub all random-id  fragment reassemble
    altq on em0 hfsc bandwidth 8000Kb queue { qwanRoot }
    altq on em1 hfsc bandwidth 8000Kb queue { qlanRoot }
    altq on em2 hfsc bandwidth  queue { qlanRoot qCLRoot }
    
    

    So far as I can tell, this is completely wrong, and not what I entered into the UI.  It should be this on 25:

    altq on em2 hfsc bandwidth 8000Kb queue { qCLRoot }

    Once I manually fix up line 25, and manually do pfctl -f /tmp/rules.debug, it loads up fine.  I'm presuming this is an XML parser error of some sort, but if I'm misconfiguring via the UI, please let me know.  Here's the relevant section of config.xml.  Just in case it comes up, yes there is a qCLdef, and it uses qCLRoot as the parent.:

    
            <shaper><schedulertype>hfsc</schedulertype>
                    <queue><name>qwanRoot</name>
                            <associatedrule>0</associatedrule>
                            <priority>0</priority>
                            <parentqueue>on</parentqueue>
                            <bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype></queue> 
                    <queue><name>qlanRoot</name>
                            <associatedrule>0</associatedrule>
                            <priority>0</priority>
                            <parentqueue>on</parentqueue>
                            <bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype></queue> 
                    <queue><schedulertype><bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype>
                            <priority>0</priority>
                            <name>qCLRoot</name>
                            <borrow><linkshare><linkshare3><linkshare2><linkshare1><realtime><realtime3><realtime2><realtime1><upperlimit><upperlimit3><upperlimit2><upperlimit1><parentqueue>on</parentqueue>
                            <attachtoqueue><associatedrule><rio><red><ecn><defaultqueue></defaultqueue></ecn></red></rio></associatedrule></attachtoqueue></upperlimit1></upperlimit2></upperlimit3></upperlimit></realtime1></realtime2></realtime3></realtime></linkshare1></linkshare2></linkshare3></linkshare></borrow></schedulertype></queue> 
                    <queue><name>qwandef</name>
                            <attachtoqueue>qwanRoot</attachtoqueue>
                            <associatedrule>0</associatedrule>
                            <defaultqueue>true</defaultqueue>
                            <priority>3</priority>
                            <realtime>on</realtime>
                            <realtime3>1%</realtime3>
                            <bandwidth>1</bandwidth>
                            <bandwidthtype>%</bandwidthtype>
                            <qlimit>500</qlimit></queue> 
                    <queue><name>qlandef</name>
                            <priority>3</priority>
                            <attachtoqueue>qlanRoot</attachtoqueue>
                            <associatedrule>0</associatedrule>
                            <defaultqueue>true</defaultqueue>
                            <realtime>on</realtime>
                            <realtime3>1%</realtime3>
                            <bandwidth>1</bandwidth>
                            <bandwidthtype>%</bandwidthtype>
                            <qlimit>500</qlimit></queue> 
                    <queue><schedulertype><bandwidth>1</bandwidth>
                            <bandwidthtype>%</bandwidthtype>
                            <priority>3</priority>
                            <name>qCLdef</name>
                            <borrow><linkshare><linkshare3><linkshare2><linkshare1><realtime>on</realtime>
                            <realtime3>1%</realtime3>
                            <realtime2><realtime1><upperlimit><upperlimit3><upperlimit2><upperlimit1><parentqueue><attachtoqueue>qCLRoot</attachtoqueue>
                            <associatedrule><rio><red><ecn><defaultqueue>on</defaultqueue></ecn></red></rio></associatedrule></parentqueue></upperlimit1></upperlimit2></upperlimit3></upperlimit></realtime1></realtime2></linkshare1></linkshare2></linkshare3></linkshare></borrow></schedulertype></queue> 
    
    --snip--</shaper> 
    


  • ???

    Any chance someone might point me in the right direction here?  I'm looking in config.inc and xmlparse.inc, and I'm just not seeing any references to altq, which is (theoretically) where I would find the rules being written to for load into pf.  The other option is that xmlparse.inc is creating a multi-dimensional array and somehow is mistakenly leaving the LAN queue in the array and simply tagging on my CL queue, thus printing it wrong.  That doesn't explain why bandwidth is being left null though. :\

    If someone can theorize where the breakage is, I am looking into it.  Just be gentle.  Perl coder here, not PHP. :D

    (EDIT: Then Numbski grew a brain and ran grep altq /etc/inc.  Low and behold….shaper.inc.  Digging into it now.  Just from a pseudo-code standpoing, I think my theory above is probably closer to the truth.  It's probably not the print statements that write the file that's at fault, but rather the parser improperly placing data into the multi-dimensional array.  We'll see.  Thoughts on this are most appreciated!)



  • (I feel it is worth noting here that I'm not showing a lack of gratitude.  I want to see pfSense grow, and have even donated to the project.  Every firewall I sell I am donating a portion of that money back to the pfSense project.  I just haven't sold many recently due to being tied up on a single project.)

    Upon further inspection, I think I need to head back to the drawing board.  Comparing the three root queues:

    
     <queue><name>qwanRoot</name>
                            <associatedrule>0</associatedrule>
                            <priority>0</priority>
                            <parentqueue>on</parentqueue>
                            <bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype></queue> 
                    <queue><name>qlanRoot</name>
                            <associatedrule>0</associatedrule>
                            <priority>0</priority>
                            <parentqueue>on</parentqueue>
                            <bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype></queue> 
                    <queue><schedulertype><bandwidth>8000</bandwidth>
                            <bandwidthtype>Kb</bandwidthtype>
                            <priority>0</priority>
                            <name>qCLRoot</name>
                            <borrow><linkshare><linkshare3><linkshare2><linkshare1><realtime><realtime3><realtime2><realtime1><upperlimit><upperlimit3><upperlimit2><upperlimit1><parentqueue>on</parentqueue>
                            <attachtoqueue><associatedrule><rio><red><ecn><defaultqueue></defaultqueue></ecn></red></rio></associatedrule></attachtoqueue></upperlimit1></upperlimit2></upperlimit3></upperlimit></realtime1></realtime2></realtime3></realtime></linkshare1></linkshare2></linkshare3></linkshare></borrow></schedulertype></queue> 
    
    

    The two "default" root queues have the following values set - associatedrule=0, priority=0, parentqueue=on, bandwidth (I have them set identically for now), and bandwidthtype=Kb.

    Looking at my CL root queue, there's something doesn't line up, specifically that  associatedrule is set to null, as opposed to 0 for the other two root queues.  I'm not seeing a place in the UI to manually set associatedrule to zero.  I'm betting that somewhere in the XML parser, there's a test that states that if associatedrule is null, pass it over, but the rule-writing routine has data that it writes out since the array is partially defined.  (Speculating wildly, and out loud at that here…)



  • Just confirmed that associatedrule=0 makes no difference.  The XML parser still mis-writes that line.  I am still having difficulties locating where in the code I would find the routine that interprets the shaper section of the xml, and writes rules.debug.  Anyone at all able to point me in the right direction?



  • Just looking over the part that gets malformed in rules.debug:

    altq on em2 hfsc bandwidth queue { qlanRoot qCLRoot }

    The problem here appears to stem from values in the multi-dimensional array that is created in xmlparse.inc, and gets written back out using altq_class.inc.

    The structure appears to be $data[array1][array2].  So in this case, what I suspect is happening is after the last queue name is read in by xmlparse.inc, it is not initializing the value of $data['queue'][].  When the next one gets written in, it is appending the value of the next read.  Within my config.xml, you have the XML block for qlanRoot, then qCLRoot.  So literally, on my install, if you were to get a list of root queues, it would list them like this:

    qwanRoot
    qlanRoot
    qlanRoot qCLRoot

    I've made this coding error hundreds of times myself in an iterating loop.  The fix is usually something like this:

    [code]
    foreach(@array){
    do something to or with the current value;
    clear the current value going into the next iteration;
    }

    In perl, that "current value" gets referenced by $_ by default.  Where I get into trouble is when I would store that value out into another variable, either because $_ was already in use, or I wanted to be super clear about what value I was messing with.  $_ gets cleared and re-read in on each iteration by default, however any home-grown variables will carry into the next iteration unless I manually init it.  I see areas all over the xmlparse.inc file that isn't just storing a value to a variable, but (ie `$variable = 'newvalue'), but appending ($variable .= 'newvalue').  I'm simply not familiar enough with the codebase or the language to properly debug this.  I've spent several days on it hoping that I would be able to obviate it and either post it here or fix it and post a patch.  Nothing doing.

    Sorry guys.



  • will this help ?
    $variable='none'
    $variable = 'newvalue'
    $variable is now newvalue the old variable is lost

    now the other one
    $variable='first'
    $variable .= 'last'
    $variable is now firstlast
    the dot is making that the old value is glued to the new variable



  • Right I get that.  PHP and perl are syntatically identical in that regard.  I just can't find where it's happening.  That's where I'm looking for help.

    (begs the question how we're getting a space between qlanRoot and qCLRoot though…perhaps it's an array and not a string, but php is being nice and returning it as a space-seperated string?)



  • Just confirming that this bug remains in RC2g.  Just tested, since it appears parser changes were made.



  • Traffic shaping is only supported on 2 interfaces.



  • My apologies for any attempted hijackings.  Since the filter fails to reload with my current config, the upgrade scripts failed until I disabled the pfctl calls.

    Anyhoo.  I understand that it isn't supported, I'm just trying to fix it is all. :)

    Any chance you might point me where in the code this iteration failure might be occuring?  My first guess was /etc/inc/xmlparse.inc, but I'm not seeing at what point rules.debug gets generated. :\



  • If you get it fixed, we'll glady wil commit changes.  We are taking a break from the shaper as it really kicked our ass in the last 1.5 years.



  • Now you're following me. :D

    I was only trying to find as polite a way to ask as I could where this happens:

    Read in config.xml (in particular, the shaper config)
    Iterate through each queue to put it into the multi-dimensional array.
    Write out rules.debug.

    As posted above, if I find where the iterations occur, I'll fix the bug.  I just can't find where in the code this occurs. :)



  • Well, after several hours of screwing around, I have a debug script.  Please try not to laugh at my shoddy php.  It's not my language:

    
    require ("shaper.inc");
    require ("xmlparse.inc");
    /* include all configuration functions */
    require_once("globals.inc");
    require_once("pkg-utils.inc");
    require_once("notices.inc");
    require_once ("config.inc");
    require_once("functions.inc");
    
    global $config;
    $config = parse_config(true);
    
    if (!is_array($config['shaper']['rule'])) {
            print "Huh, not an array.\n";
    }
    else {
            print "An array! w00t!\n";
    }
    
    foreach ($config['shaper']['queue'] as $rule){
                    $name = $rule['name'];
                    $bandwidth = $rule['bandwidth'];
                    $units = $rule['bandwidthtype'];
                    $priority = $rule['priority'];
                    $limit = $rule['qlimit'];
    
                    print "Name - $name\n";
                    print "Bandwidth - $bandwidth$units\n";
                    print "Priority - $priority\n";
                    print "Limit - $limit\n";
                    print "----------------\n\n";
    
    }
    ?>
    
    

    I won't dump the entire output, but here's the three lines of interest:

    Name - qwanRoot
    Bandwidth - 8000Kb
    Priority - 0
    Limit -
    –--------------

    Name - qlanRoot
    Bandwidth - 8000Kb
    Priority - 0
    Limit -

    Name - qCLRoot
    Bandwidth - 8000Kb
    Priority - 0
    Limit -

    This leaves me with some serious head scratching, because that means the XML parser functions are correct, and that it's the rule writing routines that are really at fault.  Grr.....so several hours down the drain.  Now I get to find the routine that writes rules.debug and see if I can hack that up a bit.



  • Welcome to the hell that we spent 1.0+ year on!



  • Thanks for the warm welcome.  ::)

    I think I have narrowed it down to the faulty block of code.  Unfortunately it uses an iteration structure totally different than perl.

    The faulty block starts on line 63 of altq_class.inc.  The iteration structure goes like this:

    for ($i = '1'; $i < count($rules); $i++)

    So we start with 1, count the number of elements in the array $rules (that really screwed me up for a while, btw.  in perl that would be @rules), and on each iteration, increment i we'll keep doing it so long as $i is less than the number of elements in $rules.

    Wordy, but accurate.  I think.

    So we go

    switch ($rules[$i])

    ???  Don't know what this does.  On the first iteration is is $rules[1].  Unless switch means take the $i element, which would be the first element of $rules?

    (I'm still thinking this loop through.  Saving and coming back to it…)



  • We don't use altq_class.inc!

    All code is in /etc/inc/filter.inc and /etc/inc/shaper.inc.



  • @Numbski:

    Thanks for the warm welcome.  ::)

    I think I have narrowed it down to the faulty block of code.  Unfortunately it uses an iteration structure totally different than perl.

    The faulty block starts on line 63 of altq_class.inc.  The iteration structure goes like this:

    for ($i = '1'; $i < count($rules); $i++)

    So we start with 1, count the number of elements in the array $rules (that really screwed me up for a while, btw.  in perl that would be @rules), and on each iteration, increment i we'll keep doing it so long as $i is less than the number of elements in $rules.

    Wordy, but accurate.  I think.

    So we go

    switch ($rules[$i])

    ???  Don't know what this does.  On the first iteration is is $rules[1].  Unless switch means take the $i element, which would be the first element of $rules?

    (I'm still thinking this loop through.  Saving and coming back to it…)

    altq_class.inc is a wild goose chase.  You really want shaper.inc and filter.inc.  FWIW, altq_class.inc isn't in releng_1 - it was removed a while back (although you may still have it in your inc directory if you have an old install that's been upgraded ).

    –Bill



  • Thanks for saving me there. :)

    What send me there is that I did a grep hfsc /etc/inc, and that was the only file that returned a match.

    Presuming that whatever was generating the broken line had to either assign it to a variable or print it outright, seemed like a logical place to look.

    I'll go ahead and delete that file so it doesn't throw me off any further…



  • Hey, he done good!  We have progress!

    New debug script:

    
    /* include all configuration functions */
    require_once("globals.inc");
    require_once("pkg-utils.inc");
    require_once("notices.inc");
    require_once ("xmlparse.inc");
    require_once ("config.inc");
    require_once("functions.inc");
    require_once ("shaper.inc");
    
    global $config;
    $config = parse_config(true);
    $altq_rules  = "";
    $queue_names = "";
    $is_first = "";
    
    if(!is_array($config['shaper']['queue'])) {
            print "Huh, not an array.\n";
    }
    else {
            print "An array! w00t!\n";
    }
    
    $ifdescrs = array('wan', 'lan');
    
    for ($j = 1; isset($config['interfaces']['opt' . $j]); $j++) {
            $ifdescrs[] = "opt" . $j;
    }
    
    foreach ($ifdescrs as $ifdescr => $ifname) {
            $queue_names = "";
            $is_first = "";
    
            print "Evaluating interface $ifname...\n";
    
            $queue_names = find_root_queue($ifname);
    
            print "$queue_names\n";
    
            if($queue_names <> ""){
                    $altq_rules .= "altq on {$config['interfaces'][$ifname]['if']} ";
                    $bandwidth_arr = get_queue_bandwidth($queue_names);
                    $bandwidth = "bandwidth {$bandwidth_arr['bandwidth']}{$bandwidth_arr['bandwidthtype']}";
                    $altq_rules .= "{$config['shaper']['schedulertype']}  {$bandwidth} ";
                    $altq_rules .= "queue { {$queue_names} }";
            }
            $altq_rules .= "\n";
    
    }
    return $altq_rules;
    
    ?>
    
    

    Output is:

    
    # php test.php
    X-Powered-By: PHP/4.4.4
    Content-type: text/html
    
    An array! w00t!
    Evaluating interface wan...
    qwanRoot
    Evaluating interface lan...
    qlanRoot
    Evaluating interface opt1...
    qlanRoot qCLRoot
    Evaluating interface opt2...
    
    

    We have a winner!  So our failure is in the find_root_queue() command.  Back to digging…



  • Oops.  We get into that command, only to find another failure.

    This time, it is in is_subqueue_used_on_interface

    Debug script below:

    
    /* include all configuration functions */
    require_once("globals.inc");
    require_once("pkg-utils.inc");
    require_once("notices.inc");
    require_once ("xmlparse.inc");
    require_once ("config.inc");
    require_once("functions.inc");
    require_once ("shaper.inc");
    
    global $config;
    $config = parse_config(true);
    $altq_rules  = "";
    $queue_names = "";
    $is_first = "";
    
    // Begin pasting find_root_queue function below.
    $queue_names = "";
    
    // Add all interfaces you wish to test to this list.
    $interfaces = array("wan", "lan", "opt1");
            foreach ($interfaces as $ifname){
                    $queue_names = "";
    
            foreach ($config['shaper']['queue'] as $queue) {
                    $rule_interface = "";
                    $q = $queue;
    
                    $name = $q['name'];
                    $parentqueue = $q['parentqueue'];
    
                    /* if we're a parentqueue and aren't attached to another queue we're probably a root */
                    if ((isset($q['parentqueue']) && $q['parentqueue'] <> "") && (!isset($q['attachtoqueue']) || $q['attachtoqueue'] == "")) {
                            /* Confirm that this is a valid queue for this interface */
                            $rule_interface = is_subqueue_used_on_interface($q['name'], $ifname);
    
                            if ($rule_interface == 1) {
                                    // Count the number of characters in $queue_names.
                                    if (strlen($queue_names) > 0) {
                                            /* If it was greater than 0, it means that $queue_names 
                                             had a value set from a previous iteration and we 
                                             need to append $q['name'] this time around with a space
                                            in front of it. This is due to output from 
                                            is_subqueue_used_on_interface. */
                                            $queue_names .= " ";
                                    }
                                    $queue_names .= $q['name'];
                            }
                    }
                    $queue_number++;
            }
            print "$ifname: $queue_names\n\n";
    }
    return $queue_names;         
    
    ?>
    
    

    Outuput is below:

    
    X-Powered-By: PHP/4.4.4
    Content-type: text/html
    
    wan: qwanRoot
    
    lan: qlanRoot
    
    opt1: qlanRoot qCLRoot
    
    

    Getting closer…



  • OMFG I am such a LOSER!!!

    ::)

    Despite what you guys may say, your current code DOES indeed work with multiple interfaces, it's just that I'm such a dork that I couldn't figure it out. :P  My debug scripts finally pointed to the problem, and perhaps this could be a jumping-off point for better error-detection to prevent filter rules that will not load, but I digress.

    Here's the final debug script I wound up with when the error became apparent:

    
    /* include all configuration functions */
    require_once("globals.inc");
    require_once("pkg-utils.inc");
    require_once("notices.inc");
    require_once ("xmlparse.inc");
    require_once ("config.inc");
    require_once("functions.inc");
    require_once ("shaper.inc");
    
    global $config;
    $config = parse_config(true);
    $altq_rules  = "";
    $queue_names = "";
    $is_first = "";
    
    // Begin pasting find_root_queue function below.
    $queue_names = "";
    
    $interfaces = array("wan", "lan", "opt1");
            foreach ($interfaces as $ifname){
                    print "Testing Infterace $ifname:\n";
                    print "=========================\n\n";
                    $queue_names = "";
    
            foreach ($config['shaper']['queue'] as $queue) {
                    $rule_interface = "";
                    $q = $queue;
    
                    $name = $q['name'];
                    $parentqueue = $q['parentqueue'];
    
                    /* if we're a parentqueue and aren't attached to another queue we're probably a root */
                    if ((isset($q['parentqueue']) && $q['parentqueue'] <> "") && (!isset($q['attachtoqueue']) || $q['attachtoqueue'] == "")) {
                            /* Confirm that this is a valid queue for this interface */
                            print "Confirming queue $name\n";
                            $rule_interface = is_subqueue_used_on_interface_fixed($q['name'], $ifname);
    
                            if ($rule_interface == 1) {
                                    // Count the number of characters in $queue_names.
                                    if (strlen($queue_names) > 0) {
                                            /* If it was greater than 0, it means that $queue_names 
                                             had a value set from a previous iteration and we 
                                             need to append $q['name'] this time around with a space
                                            in front of it. This is due to output from 
                                            is_subqueue_used_on_interface. */
    
                                            print "I just found an interfaces with 2 root queues.  I was testing queue $name and interface $ifname.\n";
                                            $queue_names .= " ";
                                    }
                                    print "Adding queue $name and interface $ifname, got a return 1 from is_subqueue_used_on_interface.\n\n";
                                    $queue_names .= $q['name'];
                            }
                    }
                    $queue_number++;
            }
            print "\nRoot Queues: $queue_names\n\n";
            print "****************************\n\n";
    }
    return $queue_names;         
    
    function is_subqueue_used_on_interface_fixed($queuename, $interface) {
    //      print "\t is_subqueue_used_on_interface called.\n";
            global $config;
            $qconfig = $config;
            //Tony adds $attachtotemp
            $attachtotemp = "";
    
            if (!is_array($qconfig['shaper']['queue'])) return 0;
    
            foreach ($qconfig['shaper']['queue'] as $queue) {
                    $attachtotemp = $queue['attachtoqueue'];
                    if($queue['attachtoqueue'] == $queuename) {   
                            print "\t Hmm.  $queuename is attachtoqueue?\n";
                            /* recurse if we're a parent queue */
                            if ($queue['parentqueue'] == "on") {
                                    print "\t ...and parentqueue is set to \"on\".\n";
                                    return is_subqueue_used_on_interface_fixed($queue['name'], $interface);
                            }
    
                            /* If we're not a parent check to see if the queue is used on this interface */
                            $subqueue_interface = filter_is_queue_being_used_on_interface_fixed($queue['name'], $interface);
                            if ($subqueue_interface != ""){
                                    print "\t\t\t Subqueue Interface Found: $subqueue_interface\n\n";
                                    return 1;
                            }
                    }
            }
            return 0;
    }
    
    function filter_is_queue_being_used_on_interface_fixed($queuename, $interface, $direction = 'in') {
            global $config;
            $lconfig = $config;
    
            print "\t\tQueue name passed is $queuename\n";
            print "\t\tInterface passed is $interface\n";
    
            if(!is_array($lconfig['shaper']['rule'])) return null;
            foreach($lconfig['shaper']['rule'] as $rule) {
                    $q  = $direction . 'queue';
                    $if = $direction . '-interface';
                    if(($rule[$q] == $queuename && $rule[$if] == $interface))
                    return $interface;
            }
            return null;
    }
    ?>
    
    

    Here is what I saw happening with this script:

    _Testing Infterace opt1:

    Confirming queue qwanRoot
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qwandef
                    Interface passed is opt1
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qwanacks
                    Interface passed is opt1
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qVOIPUp
                    Interface passed is opt1
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qP2PUp
                    Interface passed is opt1
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qOthersUpH
                    Interface passed is opt1
            Hmm.  qwanRoot is attachtoqueue?
                    Queue name passed is qOthersUpL
                    Interface passed is opt1
    Confirming queue qlanRoot
            Hmm.  qlanRoot is attachtoqueue?
                    Queue name passed is qlandef
                    Interface passed is opt1
            Hmm.  qlanRoot is attachtoqueue?
                    Queue name passed is qlanacks
                    Interface passed is opt1
            Hmm.  qlanRoot is attachtoqueue?
                    Queue name passed is qP2PDown
                    Interface passed is opt1
            Hmm.  qlanRoot is attachtoqueue?
                    Queue name passed is qOthersDownH
                    Interface passed is opt1
            Hmm.  qlanRoot is attachtoqueue?
                    Queue name passed is qOthersDownL
                    Interface passed is opt1
                            Subqueue Interface Found: opt1

    Adding queue qlanRoot and interface opt1, got a return 1 from is_subqueue_used_on_interface.

    Confirming queue qCLRoot
            Hmm.  qCLRoot is attachtoqueue?
                    Queue name passed is qCLdef
                    Interface passed is opt1
                            Subqueue Interface Found: opt1

    I just found an interfaces with 2 root queues.  I was testing queue qCLRoot and interface opt1.
    Adding queue qCLRoot and interface opt1, got a return 1 from is_subqueue_used_on_interface.

    Root Queues: qlanRoot qCLRoot
    *********************************_

    I know my wording could certainly be better here, but what it is saying is that I have opt1 using the qOthersDownL queue, and thus opt1 is trying to use qlanRoot.  opt1 == CL.  So…..

    The fix is simple to do, hard to explain.  Basically every subqueue you create, you can't go crossing over (ie, if I create a qCLRoot, qCLdef) I can't just go and plug traffic from interface CL into qOthersDownL and qOthersDownH, etc.  You have to go in and manually create matching rules to that qCLdef and qCLRoot as appropriate.  THEN the rules will load up as expected and life goes on.

    So you guys can now safely say that pfSense works with at least 3 interfaces.

    It just isn't supported! ;)

    Hopefully my debug script can be tweaked to help others, or, as I said, become a jumping-off point for better error detection.  Whew.  What a day!



  • All that we need now is the wizard for traffic shaping to be updated with your logic then we'll be set.  Are you interested in tackling this?



  • Yep, that code was "supposed" to work (and I've made it work before, it's just painful w/out a wizard to frontend it) :)  But it's backassward.  Rules attached to interfaces and queues and trawling the rules to figure out where a queue is attached is stupid.  It started off that way, I stayed on that path and completed the creation of the mess you have in front of you.  Good detective work btw.

    What I'd like to see is queues assigned to interfaces and rules assigned to queues (or assigned to interfaces, but only display queues that are part of that interface for selection).  At any rate, the shaper has a LOT of room to grow, consider this a first (really…the third) stab at it - the current wizard came out of that stab due to necessity.

    --Bill



  • LOL  My brain is exhausted from the above.  Please note that I don't code php. ;)

    That said, Sullrich, I have to leave to Montreal on Monday, and today is totally booked up.  Perhaps after I return?

    I only have one request.  I have felt a bit in the past that you got the impression that I "used" you guys, and didn't do anything to give back.  Imagine me saying this with a big smile on my face:

    I don't want to hear it anymore. ;)

    Said it before, say it again.  I have a tendency to just think out loud for the benefit of others perhaps seeing something that I've missed.  I don't intend to use anyone.

    Glad you like my detective work.  Made my head hurt at the time…. :P

    (UPDATED THOUGHT)

    ...or, when you create your "down" parts of the rules, just create two.

    ie, you can use qUpH, since that's a WAN rule, but you would need to create qlanDownH, and qopt1DownH, etc.

    Then way when you specify traffic, at a glance a mistake becomes obvious.  Also makes initial rules creation a bit simpler.



  • @billm:

    Yep, that code was "supposed" to work (and I've made it work before, it's just painful w/out a wizard to frontend it) :)  But it's backassward.  Rules attached to interfaces and queues and trawling the rules to figure out where a queue is attached is stupid.  It started off that way, I stayed on that path and completed the creation of the mess you have in front of you.  Good detective work btw.

    What I'd like to see is queues assigned to interfaces and rules assigned to queues (or assigned to interfaces, but only display queues that are part of that interface for selection).  At any rate, the shaper has a LOT of room to grow, consider this a first (really…the third) stab at it - the current wizard came out of that stab due to necessity.

    --Bill

    What about this for wizard logic then….

    Allow the user to choose interfaces to include.  WAN and LAN are default, but allow them to choose opts.

    Go ahead and create q$ifnameRoot and q$ifname.'def'.

    From there in I can see where sanity checking might go a little crazy.  The wizard could go ahead an do default shaping between WAN and the interface of your choice.  Put a notice to the user not to mix apples and oranges - ie, if you create a rule for an interface, the queues you choose must also be for that interface, and end of in the correct root queue.  Perhaps a little error checking during rules creation? (ie, check to see what the root queue is that's in use for the rule, and it the root is associated with an interface that is different from what's selected in the rule, toss at least a warning to the user of what they're doing, and that it could create un-loadable rules?)

    Just a thought or two.  Going to go rest my head now.



  • @Numbski:

    What about this for wizard logic then….

    Allow the user to choose interfaces to include.  WAN and LAN are default, but allow them to choose opts.

    Go ahead and create q$ifnameRoot and q$ifname.'def'.

    From there in I can see where sanity checking might go a little crazy.  The wizard could go ahead an do default shaping between WAN and the interface of your choice.  Put a notice to the user not to mix apples and oranges - ie, if you create a rule for an interface, the queues you choose must also be for that interface, and end of in the correct root queue.  Perhaps a little error checking during rules creation? (ie, check to see what the root queue is that's in use for the rule, and it the root is associated with an interface that is different from what's selected in the rule, toss at least a warning to the user of what they're doing, and that it could create un-loadable rules?)

    Just a thought or two.  Going to go rest my head now.

    Part of the problem (and we have it today) is shaping between LAN and OPT as well as the incorrect queue creation for the three interfaces.  Take this example

    512/512 DSL (I'm making my life simple)….If I assign 512 to lan, 512 to opt and 512 to wan, the upstream is the ONLY direction traffic shaping is correct in.  The 512 to lan + 512 to opt results in 1024 inbound which is larger than your pipe.  I had code to handle this originally, but backed it out because I wasn't thinking about how to solve the "hardware issue" - ie. hardware that doesn't support altq.  Also, lan->opt and opt->lan will be shaped at 512 (but that's an issue anyway)...  The code as it sits needs a LOT more attention (and I expect another complete rewrite) - I just got burnt out on it, I'll probably get back to it at some point, but right now shaping is kinda mind-numbing ;)

    --Bill



  • Oh wow.  I thought I had the motivation to tackle this, re-read the thread, and I've chickened out again.  No wonder this stagnates.  Having messed with it myself, I wouldn't want to touch it either!  My own test scripts barely make sense, and I commented the crap out of them….eek!



  • Yep.  ALTQ shpaing is a black art and one that will bloody your nose at times :(


Log in to reply