Control pppd behaviour

Hello,

I understand that pppd uses parameters from different files such as those located in /etc/ppp directory to control its behaviours on failures etc.

I wanted to know if its possible to say for example implement a different retry algorithm(user defined). Is this possible without modifying code ?

Quoted 8 lines · Hashmat Khan

Hello,

I understand that pppd uses parameters from different files such as those located in /etc/ppp directory to control its behaviours on failures etc.

I wanted to know if its possible to say for example implement a different retry algorithm(user defined). Is this possible without modifying code ?

What do you try "to retry" ? when one provider is down you can easy switch to others. like this (not working) example:

#!/bin/sh
for PROV in prov1 prov2 prov2
do
       pppd call $PROV
       if [ everythink_works ]
               exit 0
done

exit 1

re,
 wh

I am not looking for multiple options with which pppd needs to try. Let me give an example, normally if auth fails...it will retry after 'a' seconds for 'n' number of times.
What I would need is say it should try after 'a' seconds for first retry, after '2a' seconds for second try after first retry fails and so on..etc. Basically I want to change the retry mechanism. Is there a way to change without changing the code ?

Thanks James for long explanation.

Let me explain. I am actually looking for a different retry mechanism for failures. My retry interval would vary depending on the number of retries. To given an example:
nth retry is to be attempted at (n - 1)3 x 10 secs after previous attempt

So the retry interval is not constant but derived from a formula.

Are such things possible without modifying the code ?

Quoted 7 lines · Ashmath Khan

I am not looking for multiple options with which pppd needs to try. Let me give an example, normally if auth fails...it will retry after 'a' seconds for 'n' number of times.
What I would need is say it should try after 'a' seconds for first retry, after '2a' seconds for second try after first retry fails and so on..etc. Basically I want to change the retry mechanism. Is there a way to change without changing the code ?

It's still unclear what you're after.

Are you asking about the details of how the protocol itself works?  In other words, are you asking about how you can modify the algorithm that decides when to send PPP negotiation packets?

If so, then you'll need to change the source code and recompile.  There are documented parameters that allow you to change the maximum number of retries and the interval between retries, but none that change the algorithm used to determine when retries occur.  The algorithm is simple and fixed: when the timer expires and there are retries left, the TO+ event is delivered, and that causes the timer to be restarted and a new message to be generated.

Perhaps the real question here is: "what are you trying to do?"  If you could explain what higher-level problem you're trying to solve, it might be possible for someone to give you a better answer.

Are you trying to work around a problem?  If so, what problem, with what symptoms and debug log messages?  Are you trying to interoperate with a peer that has bugs?  Are you trying to experiment with PPP itself?

On the off chance that you're trying to "work around" an error message such as "LCP: timeout sending Config-Requests", you shouldn't.  That message doesn't mean there's anything wrong with the timer algorithm. It means that either the physical link isn't working properly or that the peer just isn't running PPP at all.  With a dial-up link, it usually means that the chat script has failed.

--
James Carlson         42.703N 71.076W         <[email protected]>

Quoted 6 lines · Ashmath Khan

Thanks James for long explanation.

Let me explain. I am actually looking for a different retry mechanism for failures. My retry interval would vary depending on the number of retries. To given an example:
nth retry is to be attempted at (n - 1)3 x 10 secs after previous attempt

Why?  You still haven't explained what you're doing.

What problem are you attempting to solve?  Can you give us any hints at all?

Quoted 3 lines

So the retry interval is not constant but derived from a formula.

Are such things possible without modifying the code ?

No.  See pppd/fsm.c, f->timeouttime, and the TIMEOUT() macro.

--
James Carlson         42.703N 71.076W         <[email protected]>

Ok.
Its a requirement from my customer which I need to implement ! My customer doesn't want normal retries but want this to be tried. As simple as that.
In addtion:and after m retries, I would want to switch to some default credentials and continue the retries.
Does this explain ?

James or others,

Is there a documentation explaining what each ppp option stands for options file, I do understand some but others I want to know in detail eg:
plugin pppoe.so eth1
nodetach
nodefaultroute
persist
maxfail 0
noauth
user ashmat

nopcomp
lcp-max-failure 5
lcp-max-configure 10
lcp-max-terminate 2
lcp-restart 3
lcp-echo-failure 3
lcp-echo-interval 10
mru 1492

Ok.
Its a requirement from my customer which I need to implement ! My customer doesn't want normal retries but want this to be tried. As simple as that.
In addtion:and after m retries, I would want to switch to some default credentials and continue the retries.
Does this explain ?

Quoted 21 lines · James Carlson

Quoted 6 lines · Ashmath Khan

Thanks James for long explanation.

Let me explain. I am actually looking for a different retry mechanism for failures. My retry interval would vary depending on the number of retries. To given an example:
nth retry is to be attempted at (n - 1)3 x 10 secs after previous attempt

Why?  You still haven't explained what you're doing.

What problem are you attempting to solve?  Can you give us any hints at all?

Quoted 3 lines

So the retry interval is not constant but derived from a formula.

Are such things possible without modifying the code ?

No.  See pppd/fsm.c, f->timeouttime, and the TIMEOUT() macro.

--
James Carlson         42.703N 71.076W         <[email protected]>

Quoted 12 lines · Ashmath Khan

James or others,

Is there a documentation explaining what each ppp option stands for options file, I do understand some but others I want to know in detail eg:
plugin pppoe.so eth1
nodetach
nodefaultroute
persist
maxfail 0
noauth
user ashmat

Quoted 8 lines

nopcomp
lcp-max-failure 5
lcp-max-configure 10
lcp-max-terminate 2
lcp-restart 3
lcp-echo-failure 3
lcp-echo-interval 10
mru 1492

Have you read the man page?  How about RFC 1661?

What more details would you like to have?  Which particular options are confusing?  (Or are they all troublesome?)

Perhaps we could offer some help here or fix the man page where it's deficient, but please give us some idea what you need.

--
James Carlson         42.703N 71.076W         <[email protected]>

What would I need to do so that pppd doesn't retry for any failures? thanks.

Quoted 7 lines · Ashmath Khan

Ok.
Its a requirement from my customer which I need to implement ! My customer doesn't want normal retries but want this to be tried. As simple as that.
In addtion:and after m retries, I would want to switch to some default credentials and continue the retries.
Does this explain ?

Unfortunately, no, but we're getting closer.

Are you talking about PPP negotiation or about separate invocations of pppd?

If it's separate invocations of pppd, then you can certainly wrap pppd with a script that checks the exit code (make sure you use the "updetach" option so you get something sensible), and use the exit code to determine what to do next.  See the man page for details.

If you're talking about the individual PPP negotiation messages that a single pppd invocation sends to its peer (i.e., you want to send PAP Authenticate-Request using user "joe" the first two times and then "bob" the next two times), then you'll need to modify the source to do as you're asking.

Fortunately, it's pretty easy to modify.

--
James Carlson         42.703N 71.076W         <[email protected]>

Quoted 1 line · Ashmath Khan

What would I need to do so that pppd doesn't retry for any failures? thanks.

Can you define precisely what you mean by "retry" and "failure" and what particular cases you're concerned about?

Sorry to be obtuse here, but it's just plain unclear to me what you're after, and I'd very much like to be helpful.

The Point-to-Point Protocol includes a negotiation mechanism that resends messages when there's no response and sometimes when there are disagreements between the peers.  I *suspect* you're not really interested in that, though you seem to have expressed interest in changing it.

The pppd implementation also includes ways to restart the entire connection after failure.  That behavior is not the default.  You must use options such as "demand" or "persist" in order to get the daemon to retry after fatal problems, such as authentication rejection or parameter negotiation failure.

If you don't want it to do that, then just don't use those options.

Please see the pppd man page for details on these options, and post a *precise* description of what you want it to do and under what conditions.

--
James Carlson         42.703N 71.076W         <[email protected]>

Its neither about PPP negotiation nor about separate invocation of pppd.

As I mentioned its about retry mechanism or the algorithm or you can call it the strategy employed to retry if things fail.

For the normal retries if the retry interval is 30 secs and max retries is 5 ... then the 1st retry is at 30 secs
2nd at 60 secs
3rd at 90 secs

But I need it something like this:
1st retry at 30 secs
2nd retry at 120 secs
3rd retry at 300 secs etc..

Hope its clear. thanks.

I want to consider all failure cases(at lcp, at chap etc), any negotiation failure etc, for which pppd would retry with the server from scratch.
Basically I don't pppd to retry after a failure but I want pppd to exit, and then the retry will be done by my program by relaunching pppd. is this possible ? maybe i need to keep the retries as zero ? thanks.

Quoted 16 lines · Ashmath Khan

Its neither about PPP negotiation nor about separate invocation of pppd.

As I mentioned its about retry mechanism or the algorithm or you can call it the strategy employed to retry if things fail.

For the normal retries if the retry interval is 30 secs and max retries is 5 ... then the 1st retry is at 30 secs
2nd at 60 secs
3rd at 90 secs

But I need it something like this:
1st retry at 30 secs
2nd retry at 120 secs
3rd retry at 300 secs etc..

Hope its clear. thanks.

Quoted 7 lines · Ashmath Khan

I want to consider all failure cases(at lcp, at chap etc), any negotiation failure etc, for which pppd would retry with the server from scratch.
Basically I don't pppd to retry after a failure but I want pppd to exit, and then the retry will be done by my program by relaunching pppd. is this possible ? maybe i need to keep the retries as zero ? thanks.

How about using the pppd "updetach" option, and using a script to invoke and retry pppd as desired?

--
James Carlson         42.703N 71.076W         <[email protected]>

Thanks James.

Quoted 1 line

From man page:

updetach
With this option, pppd will detach from its controlling terminal
once it has successfully established the ppp connection (to the point where the first network control protocol, usually the IP control protocol, has come up).

How about if chap authentication fails ? then updetach won't work.

what is controlling terminal here ?

Thanks James.

I do understand the daemon part. We do run the pppd as daemon that is in the background, we don't run from terminal but an application runs this. And able to catch exit codes as well !

But pppd doesn't exit if say authetication fails, instead it will retry.

I did have a look at maxfail:
maxfail n
Terminate after n consecutive failed connection attempts. A value
of 0 means no limit. The default value is 10.

I wanted to have 0 retries. Is there a way ?

Quoted 9 lines · Ashmath Khan

Thanks James.

Quoted 1 line

From man page:

updetach
    With this option, pppd will detach from its controlling terminal once it has successfully established the ppp connection (to the point where the first network control protocol, usually the IP control protocol, has come up).

How about if chap authentication fails ? then updetach won't work.

If authentication fails, then pppd exits with an error code as documented.  The man page entry for updetach documents the _latest_ point at which the daemon will exit, not the earliest point.

Basically, by using updetach, you can rely on the exit code from pppd to tell you whether it was successful in starting the session (exit code 0) and is now running in the background, or whether it failed to start (exit code != 0).

I suspect this is the behavior you're looking for.

Quoted 1 line

what is controlling terminal here ?

It's a UNIX term.  Processes may or may not have what's called a "controlling terminal."

When you start a process from a shell, the controlling terminal is usually your tty.  This allows you to (for instance) press ^C and interrupt that process.  It's what also terminates any lingering processes with SIGHUP when you log out.

Processes that run as daemons generally disconnect from the controlling terminal, so that they are no longer affected by the user who originally started the process (unless explicitly sent a signal with "kill").

For more information, you may want to try googling the term.

--
James Carlson         42.703N 71.076W         <[email protected]>

But pppd doesn't exit if say authetication fails, instead it will retry.

I did have a look at maxfail:
maxfail n
Terminate after n consecutive failed connection attempts. A value
of 0 means no limit. The default value is 10.

I wanted to have 0 retries. Is there a way ?

Quoted 5 lines · Ashmath Khan

Thanks James.

I do understand the daemon part. We do run the pppd as daemon that is in the background, we don't run from terminal but an application runs this. And able to catch exit codes as well !

The controlling terminal in that case is the one (if any) that's inherited from your application.

Quoted 1 line

But pppd doesn't exit if say authetication fails, instead it will retry.

Then I think you've got it misconfigured.  pppd doesn't do that by default.

If you see otherwise, then please post complete debug logs along with all of your configuration settings.

Quoted 6 lines

I did have a look at maxfail:
maxfail n
    Terminate after n consecutive failed connection attempts. A value of 0 means no limit. The default value is 10.

I wanted to have 0 retries. Is there a way ?

Using 1 would do what you want.

But that parameter has no meaning if you've configured pppd correctly for your situation.  "maxfail" does something if you're using "persist" or "demand."  It doesn't do anything otherwise.

You shouldn't have to touch that.

--
James Carlson         42.703N 71.076W         <[email protected]>

Thanks James.

Here is the current options file for our pppd.
plugin pppoe.so eth1
nodetach
nodefaultroute
persist
maxfail 0
noauth
user ashmat
nopcomp
lcp-max-failure 5
lcp-max-configure 10
lcp-max-terminate 2
lcp-restart 3
lcp-echo-failure 3
lcp-echo-interval 10
mru 1492
usepeerdns
linkname 2
unit 2

I do see the nodetach option. I have to understand this from my colleagues why its not running as daemon.

So I think setting maxfail to 1 should do the trick i.e, pppd should exit immediately after a failed connection attempt. Will pppd exit if lcp echo timesout as well ?

Quoted 33 lines · James Carlson

Quoted 5 lines · Ashmath Khan

Thanks James.

I do understand the daemon part. We do run the pppd as daemon that is in the background, we don't run from terminal but an application runs this. And able to catch exit codes as well !

The controlling terminal in that case is the one (if any) that's inherited from your application.

Quoted 1 line

But pppd doesn't exit if say authetication fails, instead it will retry.

Then I think you've got it misconfigured. pppd doesn't do that by default.

If you see otherwise, then please post complete debug logs along with all of your configuration settings.

Quoted 6 lines

I did have a look at maxfail:
maxfail n
Terminate after n consecutive failed connection attempts. A value
of 0 means no limit. The default value is 10.

I wanted to have 0 retries. Is there a way ?

Using 1 would do what you want.

But that parameter has no meaning if you've configured pppd correctly for your situation. "maxfail" does something if you're using "persist" or "demand." It doesn't do anything otherwise.

You shouldn't have to touch that.

--
James Carlson 42.703N 71.076W <[email protected]>

Thanks James.

Here is the current options file for our pppd.
plugin pppoe.so eth1
nodetach
nodefaultroute
persist

maxfail 0

lcp-max-failure 5
lcp-max-configure 10
lcp-max-terminate 2
lcp-restart 3

So I think setting maxfail to 1 should do the trick i.e, pppd should exit immediately after a failed connection attempt. Will pppd exit if lcp echo timesout as well ?

Quoted 22 lines · Ashmath Khan

Thanks James.

Here is the current options file for our pppd.
plugin pppoe.so eth1
nodetach
nodefaultroute
persist
maxfail 0
noauth
user ashmat
nopcomp
lcp-max-failure 5
lcp-max-configure 10
lcp-max-terminate 2
lcp-restart 3
lcp-echo-failure 3
lcp-echo-interval 10
mru 1492
usepeerdns
linkname 2
unit 2

Quoted 2 lines

I do see the nodetach option. I have to understand this from my colleagues why its not running as daemon.

You probably do a fork and exec to run pppd so that your app can catch pppd exiting.

Quoted 3 lines

So I think setting maxfail to 1 should do the trick i.e, pppd should exit immediately after a failed connection attempt. Will pppd exit if lcp echo timesout as well ?

As you're using persist, your app wants pppd to do the retries. Have you considered changing your app to respawn pppd when it fails? i.e. remove the persist and maxfail options. That way, you could implement the exponential backoff retry algorithm in your app rather than in pppd.

Maybe pppd needs a new option to enable exponential backoff of retries when persist is enabled?

Quoted 30 lines

Quoted 27 lines · James Carlson

Quoted 5 lines · Ashmath Khan

Thanks James.

I do understand the daemon part. We do run the pppd as daemon that is in the background, we don't run from terminal but an application runs this. And able to catch exit codes as well !

The controlling terminal in that case is the one (if any) that's inherited from your application.

Quoted 1 line

But pppd doesn't exit if say authetication fails, instead it will retry.

Then I think you've got it misconfigured.  pppd doesn't do that by default.

If you see otherwise, then please post complete debug logs along with all of your configuration settings.

Quoted 6 lines

I did have a look at maxfail:
maxfail n
    Terminate after n consecutive failed connection attempts. A value of 0 means no limit. The default value is 10.

I wanted to have 0 retries. Is there a way ?

Using 1 would do what you want.

But that parameter has no meaning if you've configured pppd correctly for your situation.  "maxfail" does something if you're using "persist" or "demand."  It doesn't do anything otherwise.

Quoted 6 lines

Quoted 6 lines

You shouldn't have to touch that.

--
James Carlson         42.703N 71.076W         <[email protected]>

--
James Chapman
Katalix Systems Ltd
http://www.katalix.com
Catalysts for your Embedded Linux software development

Thanks James.

I am thinking of this:
maxfail 1
persist # or demad

Keeping these two options will make pppd exit and my app can later control starting pppd later depending on the exit code and the algorithm.
Sometimes we also use demand instead of persist depending on the user configuration from UI.

What do you think ?