net-snmp-coders — Discussion of internal project coding (Please use this OR -users, NOT both!)

SNMP GET v. SET implementation

[Sean M. <Se...@pr...>]

Having looked through the documentation, source code and archives, I am
still at a loss when it comes to SNMP SET operations.  

Actually, I can set a variable, but what I am not clear on is the
implementation of GETs against a variable that is read/write.  As there are
not separate routines for a get or set, and I don't see any parameter that
indicates the operation as a GET/SET/GETNEXT, it seems that a GET operation
against a read/write variable actually rewrites the variable with the
current value and then returns that variable.  

I am hoping this is not the case and there is an elegant solution or method
of operation distinction as I am writing a MIB to monitor and control a
device that communicates over a 9600 serial line using VT100 so time becomes
an issue and superfluous communication has an impact.

Any help at all is appreciated.

Sean Mathias
Prosolve

Re: Some observations with SilverCreek agent tester

[Michael J. S. <msl...@is...>]

Try this on something other than Linux RH 6.1 or OpenBSD 2.8 

<opinion>
The SilverCreek SNMP tester is interesting. If you are building
an embedded agent, you might consider purchasing and using the
full version.
In no way do I personally nor professionally benefit from mentioning
SilverCreek nor InterWorking Labs.  I'm just passing on my opinion.

This is the last mention I'll make on this topic.

</opinion>

I used the SilverCreek SNMP tester (demo) to test the recent release
of ucd-snmp agent, see http://www.netsnmp.org/ , follow Downloads link.

I understand that the SilverCreek "context" is really a name
for a device description that for SNMPv3 includes user identity
and auth string, and SNMPv3 "Context".  Get the .PDF file from
them when you get the demo, it has a better explanation of this.

I did finally succeed to run the SNMPv3 test suite.  Results are
attached.

Repeat-by :
 1. Install SilverCreek demo from http://www.iwl.com/
 2. Install ucd-snmp agent and applications, (see attached "mkit"). 
    Try the attached snmpd.conf file as
/usr/local/share/snmp/snmpd.conf,
    which contains unusual line :
	createUser initial MD5 testauth

 3. Start ucd-snmp agent.
 4. Start SilverCreek demo.
 5. Click Context->New Context.  A new "SNMP Context" dialog appears.
 5.1. Set IP address; Select SNMPv3.
 5.2. Set user=initial, auth=testauth [Match values in step 2]
 5.3. Press Select.  This creates a "context" to use the SNMPv3 tests.
 6. Click "Chapters->3.0 SNMPv3 Tests"
 7. Click "Run Tests->All Tests".  Wait for completion.
 8. Click "Chapters->10.0 MIB-II Tests"
 9. Click "Run Tests->All Tests".  Wait for completion.
10. Click "Save Results->All Chapters".  Write to file.

  Except for the 10.1.2 failures, all the pertinent results from
  both test suites will be in the file you specify.

  You will see something like the results in the attached file.

  The "issue with GETNEXT" That I posted touches on the
  lexicographical errors reported here.  Run the test script (attached)
  to 

 NOTE: If you change the "context" before saving, previous results are
lost.

Regards,
-Mike Slifcak, Internet Security Systems, Inc.


Matthew Palmer wrote:
> 
> Hello Michael,
> 
> Thank you for recently downloading the limited free trial version of SilverCreek, the official SNMP Test Suite, from our website, http://www.iwl.com. The free trial is limited, but with a few sample tests, including some Tcl source code to see how easily the tests could be modified, and the demo User's guide, the limited free trial will give you an idea of the "look and feel" of SilverCreek and its general operation.
> 
> SilverCreek offers a comprehensive, automated SNMP testing solution and is an essential component of reliable SNMP implementations. SilverCreek will also smoothly integrate into your existing Quality Assurance test harness.
> 
> InterWorking Labs offers several different versions of SilverCreek and I will be happy to prepare a quote for you based on your departmental requirements.
> 
> If you need me to provide supporting documentation, or to answer any questions you may have regarding the capabilities of SilverCreek, please feel free to contact me at ma...@iw... or at (831) 430-3610 ext. 30.
> 
> In the meantime, we will be sending you a package of information, making sure to include the most useful materials. Please contact me if the address listed below is not correct.
> 
> Michael J. Slifcak
> Internet Security Systems, Inc.
> 6600 Peachtree-Dunwoody Road NE
> 300 Embassy Row Suite 550
> Atlanta, GA 30328
> 
> I look forward to discussing your SNMP testing needs and how SilverCreek can meet and exceed your expectations.
> 
> Sincerely,
> Matthew Palmer
> Sales Representative
> InterWorking Labs

Issues with "make test"

[Michael J. S. <msl...@is...>]

Why does running 'make test' for the newer AgentX tests
create a size=0 /tmp/s.AANNNN file [A=alphabetic , N=numeric] ?

Also, tests 25-28 (AgentX related) fail on OpenBSD 2.8.
Test 27 lies (the old "if testing failure, and failure to start, false
positive).

Re: recommended site or books on Tnm package

[Wes H. <har...@us...>]

>>>>> On Tue, 05 Dec 2000 15:14:05 +0800, "wk lee" <le...@ho...> said:

wk> i'm currently working on a project that requires sending SNMP
wk> messages in Tcl. i understand that i need to install a Tnm
wk> extension in Tcl to send SNMP messages to retrieve information
wk> from the devices on the network. i've managed to load in the Tnm
wk> package 3.0.0 that comes with Scotty 2.1.11. is there any
wk> recommended websites or books on the Tnm command in the package ??

You should contact them rather than talk to this list.  We're
unrelated to that package.

-- 
Wes Hardaker
NAI Labs
Network Associates

Issue with GETNEXT

[Michael J. S. <msl...@is...>]

I'm having some difficulty understanding the results I see
when I try GETNEXT requests over tables that have multiple
variable length indices.  I'm not getting the results that
I expect, and I'd like to understand if this is a protocol
issue, a MIB issue, or an agent implementation issue.


Walk the entire tree.  On my agent the first object is a scalar :
 
system.sysDescr.0 = Linux X 2.2.17 #2 Mon Nov 13 17:26:24 EST 2000 i686
 
And the last two objects are multiply-indexed table entries :
 
SNMP-VIEW-BASED-ACM-MIB::vacmViewTreeFamilyStatus."anonymousView002".1
SNMP-VIEW-BASED-ACM-MIB::vacmViewTreeFamilyStatus."anonymousView003".1

When I trim the last subidentifier of the second-last object,
and issue a GETNEXT request, I expect to see the second-last object.
[Yes, I escape the double quotes with the backslash character !]
Instead, I see the last object.   Please explain.

When I trim the last subidentifier of the last object,
and issue a GETNEXT request, I expect to see the last object.
Instead, I see an error.   Please explain.

The attached shell script will demonstrate what I've observed.

Regards,
-Mike Slifcak, Internet Security Systems, Inc.

Re: recommended site or books on Tnm package

[Peter P. <Pet...@br...>]

> hi,
> 
> i'm currently working on a project that requires sending SNMP messages in 
> Tcl. i understand that i need to install a Tnm extension in Tcl to send SNMP 
> messages to retrieve information from the devices on the network. i've 
> managed to load in the Tnm package 3.0.0 that comes with Scotty 2.1.11. is 
> there any recommended websites or books on the Tnm command in the package ??
> 
> regards
> /wk

There is no book that I am aware of, but there are manual pages and 
a website:

http://wwwsnmp.cs.utwente.nl/~schoenw/scotty/

(not really relevant directly to net-snmp)

-- 
-----------------------------------------------------------------------------
| Peter Polkinghorne, Computer Centre, Brunel University, Uxbridge, UB8 3PH,|
| Pet...@br...   +44 1895 274000 x2561       UK          |
-----------------------------------------------------------------------------

Re: 4.2 is released

[Wes H. <har...@us...>]

>>>>> On Tue, 5 Dec 2000 16:15:39 +1100 , "HOOD, Andy" <ah...@we...> said:

HOOD,> That URL doesn't work for me. This one does

Whoops.  Thanks.

-- 
Wes Hardaker
NAI Labs
Network Associates

recommended site or books on Tnm package

[wk l. <le...@ho...>]

hi,

i'm currently working on a project that requires sending SNMP messages in 
Tcl. i understand that i need to install a Tnm extension in Tcl to send SNMP 
messages to retrieve information from the devices on the network. i've 
managed to load in the Tnm package 3.0.0 that comes with Scotty 2.1.11. is 
there any recommended websites or books on the Tnm command in the package ??

regards
/wk

_____________________________________________________________________________________
Get more from the Web.  FREE MSN Explorer download : http://explorer.msn.com

RE: 4.2 is released

[HOOD, A. <ah...@we...>]

That URL doesn't work for me. This one does

http://sourceforge.net/project/showfiles.php?group_id=12694&release_id=17583
<http://sourceforge.net/project/showfiles.php?group_id=12694&release_id=1758
3> 


-----Original Message-----
From:	Wes Hardaker [mailto:wjh...@uc...]
Sent:	Tuesday, December 05, 2000 12:25 PM
To:	net...@li...
Subject:	4.2 is released


I've just placed the 4.2 tar ball on the sourceforge files site.

A huge thank you to all who helped contribute to this release.  The
number of changes in it are numerous and impressive, thanks to all of
you.

A formal announcement will be mailed out tomorrow. Anyone who wants to
grab it and do last minute checks before then is welcome to.

http://sourceforge.net/files/?group_id=12694

-- 
Wes Hardaker
Please mail all replies to net...@li...
_______________________________________________
Net-snmp-coders mailing list
Net...@li...
http://lists.sourceforge.net/mailman/listinfo/net-snmp-coders

4.2 is released

[Wes H. <wjh...@uc...>]

I've just placed the 4.2 tar ball on the sourceforge files site.

A huge thank you to all who helped contribute to this release.  The
number of changes in it are numerous and impressive, thanks to all of
you.

A formal announcement will be mailed out tomorrow. Anyone who wants to
grab it and do last minute checks before then is welcome to.

http://sourceforge.net/files/?group_id=12694

-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: Hints about MRTG

[Wes H. <har...@us...>]

>>>>> On 4 Dec 2000 21:19:28 -0000, ni...@ba... (Niels Baggesen) said:

Niels> I have just added a section to the tutorial about using MRTG to
Niels> monitor CPU and disk usage with the net-snmp demon. I hope Wes
Niels> is going to publish it when he is through with creating the 4,2
Niels> release tonight.

It should be on the web page now.

(the 4.2 changelog is being constructed as I type)

-- 
Wes Hardaker
NAI Labs
Network Associates

RE: help required : *SNMP managers* and *indexing*

[Szaboky, G. (George) <sz...@lu...>]

Wes> Yes.  You must use the real index when registering an index level
Wes> registration with a master agent.

I am having trouble doing an index level registration with a master agent. I
can register a table just fine. I use the REGISTER_MIB() macro. I have tried
using several functions in agent_registry.c without much luck. Can someone
explain how to accomplish this or point me to an example?

Thanks in advance,

George

About the pre-compiled MIBs support

[Joey S. <jo...@cy...>]

Hi guys,

	I work in a company that is experimenting with your agent and I saw
that the support for precompiled MIBs is without an author.  I was wondering
what that means and what sort of work that it needs.  I have worked on SNMP
Research's code for a while ( since v.8 ) and I would like to see if I can
do something.  Hopefully, it's something that I can do.

Please respond to this email address:

jo...@cy... or jsa...@ho...

Regards,
Joey Salanga
CyberIQ Systems

Hints about MRTG

[<ni...@ba...>]

I have just added a section to the tutorial about using MRTG to monitor CPU and
disk usage with the net-snmp demon. I hope Wes is going to publish it when he is
through with creating the 4,2 release tonight.

Any comments or further ideas are most welcome.

/Niels

-- 
Get your firstname@lastname email for FREE at http://Nameplanet.com/?su

Re: Perl module tests 100% OK

[<gm...@no...>]

%% Regarding Re: Perl module tests 100% OK; you wrote:

  >>>>>> On Sat, 02 Dec 2000 21:32:02 -0500, "G. S. Marzot" <ma...@ti...> said:
  G> Well at least on my linux box. I think it is good to go...I may get
  G> a chance to test it on Solaris and win32 but that may wait until
  G> Monday.

  wh> Excellent news.  Are you ok for a release now then?

Alot more than I was before...I just discovered another little bug with
the test scripts where they do not use the correct MIBDIR to load mibs
with -NET-SNMP-PATH targetted builds.

  wh> Have you checked that the problem in the bug database has been fixed?
  wh> It's still marked as open?

It is open as far as I know - I have not been able to recreate it -
still trying.

-GSM

  wh> -- 
  wh> Wes Hardaker
  wh> Please mail all replies to net...@li...
  wh> _______________________________________________
  wh> Net-snmp-coders mailing list
  wh> Net...@li...
  wh> http://lists.sourceforge.net/mailman/listinfo/net-snmp-coders



-- 
G.S. Marzot                        email: gm...@no...
Nortel Networks                    voice: (978)288-3990
600 Tech Park  M/S E65-60-405      
Billerica, MA  01821                 fax: (978)670-8145

RE: help required : *SNMP managers* and *indexing*

[Shshank_Sharma <Shs...@sa...>]

So well then, what you are really saying is that for tables that are
basically relevant to the NE, and whose indexing values are *meaningful* (
as in only the NE can say what those values should be), the information
should be filled out after asking the NE itself.
Sounds pretty enlightening to me. Thanks for the valuable help, Dave and
Wes. 
You guys are just storehouses of *ways-out* information. (applause)
With thanks and warm regards,
Shshank
PS: An afterthought. 
Dave/Wes, would you really be interested in having a look at the MIBs I am
implementing.
I mean, ofcourse, if you can spare the time. It'll be kind of you. :-)

> -----Original Message-----
> From:	Dave Shield [SMTP:D.T...@cs...]
> Sent:	Monday, December 04, 2000 4:27 PM
> To:	Shshank_Sharma
> Cc:	ucdavissnmpcoders; Varaprasad_Ajjampudi; PradeepKumar_Keloth
> Subject:	help required : *SNMP managers* and *indexing*
> 
> > What I have is this :
> > I am writing an SNMP sub-agent (implementing three MIBs) for an H.323
> > Gatekeeper( the NE).
> 
> > Would you really say that I retrieve actual values from the NE before I
> > register the row with the master agent ?
> 
> It partly depends on the structure of the MIBs.
> 
> If a particular table is fully implemented within a single subagent,
> then registering it doesn't need the index values at all.
>    Alternatively, if you're registering individual rows indexed by
> "an arbitrary integer", then this won't rely on data from the NE
> in order to register - you just have to keep track within the agent
> of which rows you've registered as what.
> 
>    However, if the MIB table uses a "meaningful" index (such as process
> IDs, or interface names), then your agent will need to know that
> information before it can register the rows - most likely by querying
> the NE.
> 
>   It's impossible to tell without knowing more about the structure
> of the MIB that you're implementing.
> 
> Dave

Re: Perl module tests 100% OK

[Wes H. <wjh...@uc...>]

>>>>> On Sat, 02 Dec 2000 21:32:02 -0500, "G. S. Marzot" <ma...@ti...> said:

G> Well at least on my linux box. I think it is good to go...I may get
G> a chance to test it on Solaris and win32 but that may wait until
G> Monday.

Excellent news.  Are you ok for a release now then?

Have you checked that the problem in the bug database has been fixed?
It's still marked as open?

-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: release is mostly ready... waiting till monday.

[Wes H. <wjh...@uc...>]

>>>>> On Mon, 04 Dec 2000 11:04:47 +0000, Dave Shield <D.T...@cs...> said:

Dave> I'm not sure whether this would be better as a "Feature" or a
Dave> "Fix".  Your call, Wes.

I've moved it to features.

-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: smux-peer and net-snmp

[N.Tanaka <nob...@ro...>]

Hi.

At Fri, 01 Dec 2000 13:24:19 +0000,
Dave Shield <D.T...@cs...> wrote:
> 
> > Hmm, do you think 1st getRequest is __really__ needed for setRequest ?
> 
> No - it's not fundamentally required.  It's just a consequence of
> the request-processing model of the agent, and how the SMUX support
> has been implemented.

I see.

> > Attached patch ( against 4.2.pre2 ) suppresses sending previous
> > getRequest before setRequest to smux-peer. This works.
> 
> I have to say that your proposed change bring an uncomfortably
> large amount of protocol-specific code into the guts of the agent.

Exactly. That was pretty patchy.

> Can I suggest you might like to consider an alternative approach
> that I used for the AgentX support.   In this model, the 'get'
> routine is essentially a dummy handler, that simply adds the
> requested object to an outgoing request, and trivially returns success.
> Nothing is actually sent to the subagent until all the variables
> have been processed.
[...]

Ok. I will also investigate 'agentX solution'. Thanks for the
suggestion.

Bye.

Re: smux-peer and net-snmp

[Jochen F. <jo...@sc...>]

Hi Dave,

> As far as AgentX is concerned, the request contains an indication of
> what range it's concerned with.  So the subagent shouldn't return a
> valid response the first time anyway.
> 
>   Is this a problem you're seeing with AgentX or SMUX - what version of
> the code?

It's SMUX and the latest CVS (versions prior to my patch from October,
GETNEXT didn't work correctly). An SMUX agent is required to behave as if
it servers the full object tree itself. There is no range at the
registration.

> The UCD agent doesn't currently try to cache results (I *think*).
> It would be possible, but trying to keep track of the responses
> (particularly for interleaved requests to subagents) might be
> "non-trivial".

That's true.

Cheers,
Jochen

Re: smux-peer and net-snmp

[Dave S. <D.T...@cs...>]

> I guess this would also fix the second case of duplicate requests:
> 
> If a subagent has more than one registered tree and a GETNEXT request to
> the subagent rolls over to the next registered subtree, the master agent
> will ignore the response and repeat it later.

As far as AgentX is concerned, the request contains an indication of
what range it's concerned with.  So the subagent shouldn't return a
valid response the first time anyway.

  Is this a problem you're seeing with AgentX or SMUX - what version of
the code?


> A cached response would avoid the second subagent call.

The UCD agent doesn't currently try to cache results (I *think*).
It would be possible, but trying to keep track of the responses
(particularly for interleaved requests to subagents) might be
"non-trivial".

Dave

Re: (no subject)

[Dave S. <D.T...@cs...>]

> i'm curently working on a project which requires semding SNMP commands in 
> Tcl. i was wondering if you could help me on this. i've loaded in the Tnm 
> package and that comes with Scotty.


Hmmm...  You'd probably get more help by looking on the Scotty web site
<http://wwwhome.cs.utwente.nl/~schoenw/scotty/> or asking on the mailing
list <tk...@ib...>.


>                                        when i issuing the "snmp session" 
> command, it returns me with the following.
> 
> % snmp session
> bad option "session": must be alias, delta, expand, find, generator, info, 
> listener, notifier, oid, responder, type, value, wait, or watch

Odd - when I try this, I get

	$ scwish
	% snmp session
	snmp0
	% 

as expected.   That's with scotty version 2.0.2   (yes, I know that's
creakingly old!)

  Are you sure that you're issuing this command from within 'scwish' ?

> i understand that there're snmp agents around such as CMU SNMP, WIlMA and 
> MIB compilers such as mosy and snacc. do i need to load them in as well ??

No - unless scotty has changed dramatically in the last few years,
you shouldn't need anything else for this command to work.
Obviously, you'll need to have agents somewhere, in order to send them
queries, but you shouldn't need any additional packages for the basic
functionality.

But your best hope of assistance will be with the scotty people themselves.

Dave

Re: release is mostly ready... waiting till monday.

[Dave S. <D.T...@cs...>]

> Here's what I have in the way of NEWS updates so far.
> What have I missed?

> *4.2*
>
>  Fixes:
>     - Yet more AgentX improvements and bug fixes.


Hmmm... 
this is factually accurate, I suppose, but not very informative.

The AgentX support is now a lot more robust, and complete than
the 4.1.[12] release.  This doesn't really come across under the
bare heading "more improvements".

  How about something like

	- Significant improvements to the AgentX support.
	  Still beta-status code, but probably safe to use for
	    non-critical systems.

I'm not sure whether this would be better as a "Feature" or a "Fix".
Your call, Wes.



Dave

Re: help required : *SNMP managers* and *indexing*

[Dave S. <D.T...@cs...>]

> What I have is this :
> I am writing an SNMP sub-agent (implementing three MIBs) for an H.323
> Gatekeeper( the NE).

> Would you really say that I retrieve actual values from the NE before I
> register the row with the master agent ?

It partly depends on the structure of the MIBs.

If a particular table is fully implemented within a single subagent,
then registering it doesn't need the index values at all.
   Alternatively, if you're registering individual rows indexed by
"an arbitrary integer", then this won't rely on data from the NE
in order to register - you just have to keep track within the agent
of which rows you've registered as what.

   However, if the MIB table uses a "meaningful" index (such as process
IDs, or interface names), then your agent will need to know that
information before it can register the rows - most likely by querying
the NE.

  It's impossible to tell without knowing more about the structure
of the MIB that you're implementing.

Dave

Re: simple agent

[Dave S. <D.T...@cs...>]

   The simplest form of cut-down agent would be obtained by running

	configure --enable-mini-agent --with-mib-modules=mibII

But I wouldn't expect that to be particularly slim.  That only
really omits some of the other MIB modules, which won't save you
a great deal.  To slim this down significantly, you'll need to
omit everything that's not required.

Part of the problem you'll face is that there isn't really
a very clean breakdown of "this feature is implemented in these
files" as far as the library is concerned.   For example, SNMPv3
support is integrated within the main snmp_api.c file, as well as
the v3-specific files (e.g. snmpv3.c, snmpusm.c, etc).

 If you're trying to strip this down to the bare minimum,
there's probably quite a bit you can omit.   If you're happy to
work internally without MIB files (i.e. hardwire everything,
you probably won't need the MIB parsing routines (mib.c & parse.c).
If you don't use config files, then you won't need the routines
to handle those (default_store.c, read_config.c).  It's quite
likely that you won't need alarms and the like (snmp_alarm.c)

   As a first approximation, I'd start with the following library
files:
	asn1.c  snmp_api.c   snmp_client.c    snmp.c

possibly plus some of the utility files (e.g. strtol.c, strtoul.c,
snprintf.c, tools.c)    I'd be surprised if that compiled, and you
may well need to add a few other files, and/or strip out unused code
from these.   But that's the core of the functionality you'll need.


   As far as the agent is concerned, the core files are  snmpd.c,
snmp_agent.c, snmp_vars.c, agent_registry and mib_modules.
You can probably omit agent_index, agent_read_config, agent_trap,
auto_nlist and kernel - though again you'd probably need to strip
out the references to them from elsewhere.
  Particularly snmpd.c - there's a lot of stuff there that you
probably won't need.   


   Whether it's worth the trouble of going to this level depends on
how much space you've got to squeeze this into.   Just removing
SNMPv3 support might be sufficient.
   It'd be useful to know how you get on.

Dave