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

ucd-snmp-4.2.pre2.tar.gz not found on SourceForge

[Jeffrey D. W. <jdw...@mu...>]

Greetings Great & Mighty ucd-snmp-coders !

I've been trying to educate myself on SNMP.  I tried to download the latest
version from
sourceforge.net, but there doesn't seem to be a file with the link for
'ucd-snmp-4.2.pre2.tar.gz'.
Is there a problem with SourceForge ?  I downloaded the 4.1.2 version from
Switzerland.

Thanks.
-- 
Regards,
   Jeffrey D. Watson
   jdw...@mu...
   http://jdwatson.home.mindspring.com
   Motto: Too much of a good thing is not Enough!


*tag* A bartender is just a pharmacist with a limited inventory.

Re: last call for 4.2 problems/patches

[Frank S. <st...@ib...>]

Wes> We hope to release version 4.2 at the end of next week, so if anyone
Wes> has any issues with the 4.2.pre2 release that we haven't addressed
Wes> yet, please let us know as soon as possible!

Could you please add a `configure --help' line for the TUNNEL-MIB and
apply the following small patch to agent/mibgroup/tunnel/tunnel.c?

--- tunnel.c-old	Thu Oct 12 22:47:32 2000
+++ tunnel.c	Thu Nov 30 10:45:40 2000
@@ -85,7 +85,7 @@
 
 
 
-/* This is used, this the TUNNEL-MIB aguments ifTable. */
+/* This is used, because the TUNNEL-MIB augments ifTable. */
 extern unsigned char *var_ifEntry(struct variable *,
 				  oid *, size_t *,
 				  int, size_t *,
@@ -156,8 +156,6 @@
 
 
 
-char *configfilename;
-
 extern int register_sysORTable(oid *, size_t, const char *);
 extern int unregister_sysORTable(oid *, size_t);
 
@@ -711,17 +709,6 @@
 
 
 
-#if 0
-static int writeStatus(int action, unsigned char *var_val,
-		       unsigned char var_val_type, size_t var_val_len,
-		       unsigned char *statP, oid *name, size_t name_len)
-{
-    return SNMP_ERR_NOTWRITABLE;
-}
-#endif
-
-
-
 unsigned char *var_tunnelIfEntry(struct variable *vp,
 				 oid *name, size_t *length,
 				 int exact, size_t *var_len,
@@ -895,9 +882,6 @@
 	ret_int = 1; /* active */
 	*var_len = sizeof(ret_int);
 	vp->type = ASN_INTEGER;
-#if 0
-	*write_method = writeStatus;
-#endif
 	return (u_char *) &ret_int;
     default:
 	return 0;

Re: snmptrap api

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

On Wed, 29 Nov 2000 11:07:08 -0500 (EST) Jas...@ms... wrote:
>	thanks for the info...just a random question then...how come the
>snmptrap.c file uses synch_response?  (that's where i got the idea
>from)...but the app works great now, thanks!

It does when it sends an inform which is an acknowledged trap. Otherwise is uses
snmp_send

/Niels

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

Re: Agentx vs SIGILL

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

> 2) SIGILL sent to the subagent
>    The subagent crashed, ok.
>    But the master agent also crashed !
>    Is it a normal behavior ?

No - it's neither normal, desired or expected.
The master agent is meant to have code to detect the abrupt
disappearance of a subagent, and recover smoothly.
But this hasn't been tested particularly thoroughly, so it
doesn't surprise me that it's not working properly.


>    Is it possible to change this ?

Given that the long-awaited 4.2 release is imminent, it's probably
not sensible to try and rush through a fix at this late stage.
The AgentX support is still experimental, so it's not unreasonable
that it doesn't cope with pathological cases just yet.
  (You should find that a more graceful shutdown by the subagent
doesn't have the same disaterous effect).

  But of course, we do need to get this beefed up before this
feature is released as "stable code".   I'd certainly hope that
this would be fixed before the first net-snmp release (and we
can always release a patch in the meantime).

  If you happen to find the cause in the next day or so, and it
turns out to be a fairly trivial fix, then it might as well go in
anyway.  But I'd expect things to take a little longer, and it
doesn't seem fair to delay the 4.2 release yet again.

Dave

[Fwd: Re: New port OS/2]

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

Thank you, Sergey.  Sorry for copying, but your explanation
is very valuable, and it should go to the -coders list.  -Slif

-------- Original Message --------
From: "Sergey I. Yevtushenko" <ev...@na...>
Subject: Re: New port OS/2
To: "Michael J. Slifcak" <msl...@is...>

On Wed, 29 Nov 2000 08:02:32 -0500, Michael J. Slifcak wrote:

Hi Michael,

>I'm sorry to be so late in reply; Paid work has not afforded
>time to reply to you sooner.  I've reviewed your patch to
>ucd-snmp-4.2.pre2.  Very nicely written.

Thanks. 

>I have some concerns about the approach you used in redefining
>some of the function signatures.  
>Please describe what the define below accomplishes in OS/2.

>+#define CCONV   _System         /* Use _System as a calling convention.
>+                                   This is most suitable for everyone
>*/

Resulting function prototype will look like below:

void _System foo(...)

This tells compiler to use special format for call: defines placement
and order of operands
and tells who (caller or callee) is responsible for clearing stack.
This particular convention is similar to standard calling convention on
PC architecture:
parameters are passed in stack, first put in stack last passed parameter
(i.e. parameters
passed from right to left) and caller clears the stack. Just for
information: used Visual Age C++ 
compiler for other calls (including its own LIBC) uses so called
_Optlink calling convention,
where parameters passed in registers and stack (if needed).

By default most compilers under OS/2 using different calling
conventions. This is not 
a problem when entire application (including underlying libraries) is
compiled with 
one compiler and identical command line parameters. But this causes
problems when 
there is a mix of options and/or compilers. 
Another (even more important and complex) issue is placing library in
shared module (DLL). 
In this case application that uses this library may be even written on
other language.
For such a cases using mentioned above _System calling convention is a
good choice
because any language/compiler must know this convention: without
knowledge of 
it compiler will be unable to call system API's (they are exported with
this calling convention).

This issue is closely linked with structure packing but I haven't
addressed it at this time.
Plan to complete it together with the port of agent application.

Thanks.

Regards,
	Sergey.

*--------------------------------------
ES@Home

Re: snmptrap api

[<Jas...@ms...>]

	thanks for the info...just a random question then...how come the
snmptrap.c file uses synch_response?  (that's where i got the idea
from)...but the app works great now, thanks!

jason

Agentx vs SIGILL

[Jargot J. <Jer...@fe...>]

 version: UCD-SNMP version 4.2.pre2 + bug #326 patch
  System: SunOS 5.7 Generic_106541-08 sun4u sparc SUNW,Ultra-250
Protocol: AgentX



  Hi all !


I compiled the master agent to handle the MibII and a subagent to handle the Mib HOST.

I ran both of them:
  For master agent:
  > ./snmpd -f -c test.conf
  For subagent:
  > ./snmpd -X -f -c test.conf

I ran a snmpwalk on Mib HOST (handled by the subagent) in an another xterm:
  > snmpwalk localhost private .1.3.6.1.2.1.25


--

Then I used SIGILL onto the agents:


1) SIGILL sent to the master agent

   The master agent crashed. The subagent did not, and it was unreachabled.
   
   OK


2) SIGILL sent to the subagent

   The subagent crashed, ok.

   But the master agent also crashed !

   Is it a normal behavior ?

   If there is several subagents, and if one of the subagent crashes, then the master agent crashes, and so, all other subagents become unreachabled.

   Should it be different ?

   Is it possible to change this ?


   Could someone help me ?


 thanks in advance,

  bye,

       Jerome

Re: agentx question?

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

> I have a question about allocating an index. If a subagent allocates an
> index and then is reset or the process is killed and restarted how does the
> subagent know or aquire the same index it had registered for previously?

If it needs to re-use the same index value, it's up to the subagent to
remember this value across invocations.   It can then register it with
the master agent using one of the 'register_xxx_index' utility routines
in 'agent_index.c'

  See the testing code at the end of this file for examples.

Dave

Re: snmptrap api

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

On Tue, 28 Nov 2000 21:00:13 -0500 (EST) Jas...@ms... wrote:
>	hey again~  so i'm trying to now get my external program to send
>an snmptrap...teh good news is that it does successfully send a trap, logs
>it accordingly in /var/log/messages, etc...the problem is that my program
>never returns from the call to snmp_synch_response...

No, because TRAP has no response. Hmmm, maybe snmp_synch_response should learn 
something about traps?. Anyway, when issuing a trap just call snmp_send, and 
thats it.

/Niels

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

snmptrap api

[<Jas...@ms...>]

	hey again~  so i'm trying to now get my external program to send
an snmptrap...teh good news is that it does successfully send a trap, logs
it accordingly in /var/log/messages, etc...the problem is that my program
never returns from the call to snmp_synch_response...
	i wrapped the code into an object, class SNMP (duh).  don't ask
about the separate instantiation of the session called foo...
	am i missing something in here that's considered "standard?"  

	i've had no trouble whatsoever sending snmpget and snmpset
messages, with successful returns from snmp_synch_response.

{
 SNMP tempSession("localhost", DEFAULT_COMM_NAME,DEFAULT_COMM_LENGTH,
		    DEFAULT_TRAP_PORT);
  snmp_session foo(tempSession.getTrapSession());
  tempSession._connectedSession = snmp_open(&foo);
  tempSession._storedPdu = snmp_pdu_create(SNMP_MSG_TRAP);
  read_objid(oid_,tempSession._oid, &tempSession._oid_len);

snmp_add_var(tempSession._storedPdu,tempSession._oid,tempSession._oid_len, 
	       type_, value_);
  cout << "got here0" << endl;
  tempSession.status = snmp_synch_response(tempSession._connectedSession,

tempSession._storedPdu,&tempSession._responsePdu);
  //send_trap_vars(-1,-1,_storedPdu->variables);
  cout << "got here00" << endl;
  if (tempSession.status == STAT_SUCCESS && _responsePdu->errstat ==
SNMP_ERR_NOERROR)
    ;
  else 
    snmp_sess_perror("snmptrap",_connectedSession);
  if (tempSession._responsePdu)
    snmp_free_pdu(tempSession._responsePdu);
  snmp_close(tempSession._connectedSession);
}

thanks as always,

jason

register_mib_range() question

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

Hi,

I am having trouble understanding how to use register_mib_range. I have an
agentx subagent that aquires an index to a table. I would then like the
subagent to register a single row in the table. So I want to use the new
index value to register a fully qualified mib region. For example, I have a
module oid of:
	1.3.6.1.4.1.1751.2.79.1.1.1

and a variable table that looks like:

struct variable2 configTable_variables[] = {
#define CONFIGDEVICECOUNT 3
{CONFIGDEVICECOUNT, ASN_INTEGER, RONLY, var_configTable, 2, {1, 1} },
       ...

NOTE there are 10 variables in this table.

So if we assume the index the subagent aquires is 2, the subagent has to
register for the following oid's

      1.3.6.1.4.1.1751.2.79.1.1.1.1.[1-10].2 

the first new oid would be:
 
       1.3.6.1.4.1.1751.2.79.1.1.1.1.1.2   (we will call it newoid)
 
So the subagent registers with:
       
      register_mib_range("configTable", 
                         (struct variable *)configTable_variables, 
                         sizeof(struct variable2),
                         sizeof(configTable_variables)/sizeof(struct
variable2),
                         newoid, 
	                   15,
                         DEFAULT_MIB_PRIORITY,
                         14,
                         10);

 This gives registers a set of oids with the length of newoid + the suffix
oid in the variable table. If I use the module id as follows:

      register_mib_range("configTable", 
                         (struct variable *)configTable_variables, 
                         sizeof(struct variable2),
                         sizeof(configTable_variables)/sizeof(struct
variable2),
                         oid, 
	                   12,
                         DEFAULT_MIB_PRIORITY,
                         14,
                         10);

It does not register the correct oids. Does anyone have any ideas? Is there
another interface function? 

George

agentx question?

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

Hi,

I have a question about allocating an index. If a subagent allocates an
index and then is reset or the process is killed and restarted how does the
subagent know or aquire the same index it had registered for previously?

George

[Fwd: about ucd-snmp]

[Shiow J. J. <sj...@cl...>]

Raw IO values for the UCD MIB ?

[Jochen K. <joc...@tu...>]

Hello!

I implented ssIORawSent and ssIORawReceive (or whatever you want to call them) on
Solaris, like the raw CPU Counters, these are the raw blocks written/read counts for
a CPU.  Can we add these to the MIB, like the raw CPU values?  I don't know what to
call them, following the ssCpuRawUser etc. style it would be the above.  Then it
would be

ssIORawSent OBJECT-TYPE
    SYNTAX      Counter32
    MAX-ACCESS  read-only
    STATUS      current
    DESCRIPTION
        "Blocks sent (written) to a block device."
    ::= { systemStats 54 }

and

ssIORawReceived OBJECT-TYPE
    SYNTAX      Counter32
    MAX-ACCESS  read-only
    STATUS      current
    DESCRIPTION
        "Blocks received (read) from a block device."
    ::= { systemStats 55 }

using the next free OIDs after the CpuRaw section.  

Please tell me what you think, whether this makes sense at all etc.

Jochen
-- 
Technology is dominated by two types of people:
        those who understand what they do not manage
        and those who manage what they do not understand!

Re: read_objid() capability to interpret strings

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

>>>>> On 23 Nov 2000 14:31:05 -0000, ni...@ba... (Niels Baggesen) said:

Niels> It is quite easy to do, so I might as well change it. I really
Niels> also would like to have the possiblity of no-quotes => mib
Niels> decides, but how do you then know what to do with the index 0
Niels> (is it the string "0", or does it mean a zero-length string
Niels> specified by oid?)

This would be my vote (and in fact I was also going to mention it) for
how to do things.  I'd like to see it translated without the quote
requirements if the mib is present.  Only if it isn't present would
the user be required to enter quote marks to specify it as a string
and not a node.  As far as null length strings go, how about .. being
equal to a null length string or oid segment?  That makes the most
sense to me.

Now, if I can only get to a network node to send this today...
-- 
Wes Hardaker
Please mail all replies to net...@li...

Re: C++ style comment in parse.c

[Jochen K. <joc...@tu...>]

Hello!

On 27-Nov-2000 ni...@ba... wrote concerning Re: C++ style comment in parse.c:

>>A C++ style comment in C code.  Sun's compiler is that picky.  Somebody care to fix

> Fixed.

Thank you.

> How did the compile go after you fixed it?

Ran right thru.  I can confirm the latest CVS agent working on Solaris 2.5.1 and 2.7
on Sun hardware.  Some warnings during compile but no errors.

Jochen
-- 
Every nonzero finite dimensional inner product space has an
orthonormal basis. It makes sense, when you don't think about it.

Re: C++ style comment in parse.c

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

On Mon, 27 Nov 2000 13:03:42 +0100 (CET) Jochen Kmietsch
<joc...@tu...> wrote:
>Hello!
>
>I am trying to compile the latest net-snmp version, using
>
>cc: WorkShop Compilers 5.0 98/12/15 C 5.0
>
>and here is what I get:
>
>/bin/sh ../libtool --silent --mode=compile cc -g -Dsolaris2 -I. -I.. -I. -I./..
>-I/usr/local/share/openssl/include -c -o parse.lo parse.c
>"parse.c", line 4637: syntax error before or at: /
>cc: acomp failed for parse.c
>
>The code is
>
>    struct node *nnp;
>   //printf("merge defval --> %s\n",np->defaultValue);
>    nnp = parse_objectid(fp, name);                   
>
>A C++ style comment in C code.  Sun's compiler is that picky.  Somebody care to
fix

Fixed.

How did the compile go after you fixed it?

/Niels

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

Some issues

[Sergey I. Y. <ev...@na...>]

Hi,

I'd like to discuss two important issues which I believe interesting not only for me.

While doing OS/2 port I dig through entire source tree and found that there is no 
single formatting style in the sources. This makes reading of the sources rather 
hard not because of particular style but rather because of changes of the style 
between source files and even inside one file.
I propose to adjust one style and then reformat sources with indent. 
Resulting command line for indent together with description of the style will
be a valuable part of the project documentation. As a starting point I propose
to review formatting style of FreeBSD and Linux. Probably we can accomodate 
one of these with minimal changes. 

Other issue also an organizational one.  I have found that there are no instructions
for newbies how to behave in the list and how to issue a patch for review by coders.
Thanks to Mike Slifcak, I highly appreciate his help but short file with
instructions included in package could save valuable time for both of us.


Regards,
	Sergey.

*--------------------------------------
ES@Home

Re: Release candidate

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

ni...@ba... wrote:
> 
> I have now done and installation, run make test, and done a walk on the
> following systems (all x86 based):
> 
>    RedHat 5.0 (kernel 2.0.32)
>    Mandrake 7.1 (kernel 2.2.15)
>    RedHat 7.0 (kernel 2.2.16)
>    Solaris 7
>    Solaris 8
>    NetBSD 1.4.2
>    NetBSD 1.5a
>    OpenBSD 2.6
>    FreeBSD 3.5
>    FreeBSD 4.1
> 
> It would be nice with a test for - at least - FreeBSD 2.2.8 (and maybe 5) and
> OpenBSD 2.7. And if anybody has a HP, AIX, IRIX, sparc (64-bit), ...
> 
> Should we upload the config.h and walk output to some common repository?
> 
> /Niels
> 

Attached please find result of build/walk on OpenBSD 2.8.
config.guess, nohup.out (build + test output),
of.out (snmpwalk -OS public .1 output).
Regards,
-Mike Slifcak

Re: New port OS/2

[Sergey I. Y. <ev...@na...>]

Hi,

Below attached my patches to make net-snmp compile under OS/2 with Visual Age C++ 3.6.5.
Some comments (without any particular order, just how they come to mind). 

1. To make visible my changes in manually edited config.h I copied original
config.h.in into config.h in original sources.

2. I have used C++ compilation mode to force stronger type checks.
Some changes (mostly explicit conversions between char* and u_char* and vice versa)
are direct consequence of that.

3. Patches for agent are not included, they are not completed but I'm working on them.

4. Significant part of these patches are addition of the calling convention constant in declarations
(and implementation where this is relevant).

Comments and suggestions are welcome.

Regards,
	Sergey.

*--------------------------------------
ES@Home

help required : *SNMP managers* and *indexing*

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

Hi Everyone, 

The first one is a general query about an SNMP manager.
The SNMP manager wants to retrieve information from an SNMP agent.
***************************
Question # 1
How does an SNMP manager form valid object identifiers(OIDs) for tabular
variables that have multiple variables indexing them ?
Does it just do an snmpwalk from a higher level and see what are the valid
indices by inpecting the returned OIDs ? or does it have some way of forming
them itself  ?
***************************


The second one is about table indexing. I am implementing an SNMP agent for
a network element (NE). They talk to each other using UDP messages.
***************************
Question # 2
Do I need to give *valid*  values for the indexing variables in a table
instance when I am creating that instance in the sub-agent implementation
and registering it with the master agent ?
(By *valid*- I mean the actual values of those parameters as present at the
NE)
Or can I just initialize those variables to some default values ?
***************************

Any and all pointers will be acknowledged with thanks.
The time and energy you spent on this is appreciated.
You guys are just great !
Thanks and warm regards,

Shshank Sharma
embedded systems
+91- 40- 3306767 x 7691
E-mail : shs...@ht...
Internet : http://profiles.yahoo.com/shashank_s

Life is a labyrinth, always go straight and turn right

finger in snmp

[Barrios G. C. I. <ba...@in...>]

hi, 


I have heard that you could use snmp in order
to know who is on console. just like a finger.


%finger @ravel.uc.co

[ravel] 
Login       Name              TTY Idle    When    Where
jeffy    Jeffy Pilar Jimenez   co      Mon 10:48 
jeffy    Jeffy Pilar Jimenez   p2    3 Mon 10:48  :0.0                
alexanbm Alexander  Barrios M  p3   1d Thu 16:45  150.135.110.147     
araujol  Luis R. Araujo R.     p4    9 Mon 10:11  boole.uc.co    
wilmerm  Wilmer Rolando Mendo  p5 2:04 Mon 07:45  160.135.126.42  


something like this.


I hope if it does exist, please tell me how could
I use it.


thanks 

#################################################
#                                               #
######         Carlos I. Barrios G.        ######
#                                               #
#                 Ing. Sistemas                 #
#                 Ing. Mecanica                 #
#                                               #
###         Universidad de los Andes.(ULA)      #
###  		Merida-Venezuela 	      ###
#                                               #
#                                               #
#################################################

C++ style comment in parse.c

[Jochen K. <joc...@tu...>]

Hello!

I am trying to compile the latest net-snmp version, using

cc: WorkShop Compilers 5.0 98/12/15 C 5.0

and here is what I get:

/bin/sh ../libtool --silent --mode=compile cc -g -Dsolaris2 -I. -I.. -I. -I./..
-I/usr/local/share/openssl/include -c -o parse.lo parse.c
"parse.c", line 4637: syntax error before or at: /
cc: acomp failed for parse.c

The code is

    struct node *nnp;
   //printf("merge defval --> %s\n",np->defaultValue);
    nnp = parse_objectid(fp, name);                   

A C++ style comment in C code.  Sun's compiler is that picky.  Somebody care to fix
to

/* printf("merge defval --> %s\n",np->defaultValue); */

?

Jochen
-- 

1st NET-SNMP roundtable at San Diego IETF conference?

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

The last ucd-snmp roundtable was a lot of fun, so I thought we'd have
another one at the upcoming IETF conference, since I know a fair
amount of people are planning on attending that one.  At the last one
we had dinner and beer followed by a trip to a pool hall, where
Michael Slifcak was clearly noted to be the better pool player.

Anyone who is planning on being in San Diego, CA, USA sometime between
the 10th and 15th of December and wants to attend, send me a note
(including when you'll be around) so I can coordinate a date and
meeting times.  Anyone who wants to go to the conference but has yet
to sign up, the pre-registration closes December 1st (see
http://www.ietf.org/).

(Since we're now under the net-snmp umbrella, this will be the 1st
net-snmp roundtable following the 1st ucd-snmp roundtable).

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

last call for 4.2 problems/patches

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

We hope to release version 4.2 at the end of next week, so if anyone
has any issues with the 4.2.pre2 release that we haven't addressed
yet, please let us know as soon as possible!

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