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

Re: smux-peer and net-snmp

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

> 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.


> 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.

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.

   This means that a multi-varbind request will only generate one
call to the subagent, and a SET request will not be preceded by a
subagent GET call, so object creation is not a problem.
It also keeps the details of the subagent protocol contentrated
in the one area, and doesn't need any special-casing (other than
recognition of "delegated requests").
   See the routines 'agentx_var' and 'agentx_add_request' for details.


  It should be quite possible to implement SMUX support in a similar
way - probably simpler than the AgentX support has been.   It's one of
a long line of projects I've wanted to have a stab at, but has been
relatively low priority.   If you want to have a go yourself, then
feel free.

Dave

Re: smux-peer and net-snmp

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

Hello, thanks for your reply.

At 30 Nov 2000 12:34:29 -0800,
Wes Hardaker <wjh...@uc...> wrote:
> 
> >>>>> On Thu, 09 Nov 2000 15:03:13 +0900, "N.Tanaka" <nob...@ro...> said:
> 
> N> I have made smux-peer program and be trying to communicate to snmpd
> N> from net-snmp-4.1.2. When snmpd received setRequest from manager, I
> N> could see 2 getRequests from snmpd and 1 setRequest from snmpd like
> N> below.
> 
> N> smux-peer           snmpd             manager
> N> ----+----           --+--             ---+---
> N> |                 |                  |
> N> |                 |<---setRequest--- |
> N> |<---getReq------ |                  |
> N> | --------------->|                  |
> N> |                 |                  |
> N> |<---setReq------ |                  |
> N> | --------------->|                  |
> N> |                 |                  |
> N> |<---getReq------ |                  |
> N> | --------------->|                  |
> N> |                 |                  |
> N> |<---SOut-------- |                  |
> N> | --------------->|                  |
> N> |                 | ---getResponse-->|
> N> |                 |                  |
> 
> N> (1) Is this normal ? (I have thought only one setRequest from
> N> snmpd would apper in this situation ).
> 
> 1 is normal, just because of the way we implement set support in our
> agent.  I'm confused as to why you'd be getting the second one
> though.

Hmm, do you think 1st getRequest is __really__ needed for setRequest ?
Assume the response with noSuchObject error against 1st getRequest.
Smux peer may want to create new table entry when it receives
setRequest for unknown oid, but snmpd won't send setRequest because
previous getRequest fails. ( This is our case )

Actually, when snmpd receives response with error, such as noSuchObject
, against 1st getRequest, it will crash. (4.2.pre1). I've reported
another mail (below).

	Message: 4664457
	FROM: N.Tanaka
	DATE: 11/15/2000 04:04:20
	SUBJECT: RE:  SMUX - snmpset

Attached patch ( against 4.2.pre2 ) suppresses sending previous
getRequest before setRequest to smux-peer. This works.

> Unfortunately, no one is actively maintaining the smux code at the
> moment which is one reason that your question has not gotten answered
> sooner...

I could not see 2nd getRequest with 4.2.pre2. Were there some changes
about smux ?

----------------------------------------------------------------------
*** agent/snmp_vars.c	Thu Nov 16 00:49:19 2000
--- snmp_vars.c	Fri Dec  1 15:06:30 2000
***************
*** 141,146 ****
--- 141,153 ----
   */
  #define MIB_CLIENTS_ARE_EVIL 1
   
+ 
+ extern u_char 	*var_smux (struct variable *, oid *, size_t *, int, size_t *,
+                            WriteMethod **write_method);
+ extern u_char 	*var_smux_for_set (struct variable *, oid *, size_t *, int,
+ 					 size_t *, WriteMethod **write_method);
+ 
+ 
  extern struct subtree *subtrees;
  int subtree_size;
  int subtree_malloc_size;
***************
*** 330,339 ****
  		    DEBUGMSG(("snmp_vars"," ...\n"));
  
  		gaga:
! 		    access = (*(vp->findVar))(cvp, name, namelen, exact,
! 						  len, write_method);
! 	    	    DEBUGMSGTL(("snmp_vars", "Returned %s\n",
! 				(access==NULL) ? "(null)" : "something" ));
  
  			/*
  			 * Check that the answer is acceptable.
--- 337,354 ----
  		    DEBUGMSG(("snmp_vars"," ...\n"));
  
  		gaga:
!  		    if ((vp->findVar == var_smux) && (pdu->command == SNMP_MSG_SET)) {
! 		    	access = var_smux_for_set(cvp, name, namelen, exact,
! 							  len, write_method);
! 	    	 	   DEBUGMSGTL(("snmp_vars", "Returned %s\n",
! 					(access==NULL) ? "(null)" : "something" ));
! 				}
! 				else {
! 		    	access = (*(vp->findVar))(cvp, name, namelen, exact,
! 							  len, write_method);
! 	    	 	   DEBUGMSGTL(("snmp_vars", "Returned %s\n",
! 					(access==NULL) ? "(null)" : "something" ));
! 				}
  
  			/*
  			 * Check that the answer is acceptable.
*** agent/mibgroup/smux/smux.c	Fri Nov  3 06:23:43 2000
--- smux.c	Fri Dec  1 15:07:22 2000
***************
*** 124,130 ****
--- 124,134 ----
  };
  
  
+ int	val_smux;
+ u_char 	*var_smux_for_set (struct variable *, oid *, size_t *,
+ 				int, size_t *, WriteMethod **write_method);
  
+ 
  void
  smux_parse_peer_auth(const char *token, char *cptr)
  {
***************
*** 295,300 ****
--- 299,351 ----
  	}
  }
  
+ 
+ u_char *
+ var_smux_for_set(struct variable *vp,
+ 	oid *name,
+ 	size_t *length,
+ 	int exact,
+ 	size_t *var_len,
+ 	WriteMethod **write_method)
+ {
+ 	u_char *valptr, val_type;
+ 	smux_reg *rptr;
+ 
+ 	*write_method = NULL;
+ 
+ 	/* search the active registration list */
+ 	for (rptr = ActiveRegs; rptr; rptr = rptr->sr_next) {
+ 		if (!compare_tree(vp->name, vp->namelen, rptr->sr_name,
+ 		    rptr->sr_name_len))
+ 			break;
+ 	}
+ 	if (rptr == NULL)
+ 		return NULL;
+ 	else if (exact && (*length < rptr->sr_name_len))
+ 		return NULL;
+ 
+ 	*write_method = var_smux_write; 
+ 
+ 
+ 	*var_len = sizeof(val_smux);
+ 	val_type = ASN_INTEGER;
+ 	valptr = (u_char*)&val_smux;
+ 
+ 
+ 	if ((compare_tree(name, *length, rptr->sr_name,
+ 	    rptr->sr_name_len)) != 0) {
+ 		/* the peer has returned a value outside
+ 		 * of the registered tree
+ 		 */
+ 		return NULL;
+ 	} else {
+ 		/* set the type and return the value */
+ 		vp->type = val_type;
+ 		return valptr;
+ 	}
+ }
+ 
+ 
  int 
  var_smux_write(
  	int action,

RE: snmpconf needs testing

[RADMAN S. <Ste...@CT...>]

VERY useful stuff. This really could take the crypt out of snmp*.conf :)

tested with "./snmpconf -c snmpconf.dir -g basic_setup" right off the CVS
smooth operation

Minor change proposal:

local/snmpconf.dir/snmd.conf/monitor:
change
"multiple file Do you want to configure the agents ability to file sizes?"
to
"multiple file Do you want to configure the agents ability to monitor file
sizes?"
and add info about the units of "maxsize".

more to follow after make install

Stefan


-----Original Message-----
From: Wes Hardaker [mailto:wjh...@uc...]
Sent: Thursday, 30 November, 2000 18:40
To: net...@li...
Subject: snmpconf needs testing



If anyone has a chance, build of the cvs snapshot and run "snmpconf -g
basic_setup" and see if it looks ok.  (or just snmpconf to test other
aspects of it).

If you don't want to do a make install, then:

  cd local
  ./snmpconf -c snmpconf.dir -g basic_setup

-- 
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

Re: URGENT: [Fwd: Maximum size of a trap packet]

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

> I originally sent this message about 10 days but it got drowned in the flood
> of emails. Could somebody please help on this issue?

I'm sorry, but I've got absolutely no idea what's failing.
I'd probably need to try running things under the debugger.

You say that the subagent never gets to COMMIT.
Does it receive the AgentX "Cleanup Set" request, or not? 
What's the error-status field in the response to the earlier Agent
"Commit Set" ? 
   (Note that these do not correspond to the "obvious" internal passes,
so don't be misled by the names).

  If the subagent receives the Cleanup Set request, then the problems
in the subagent - if it doesn't then the problem's in the master.

Other than that, I'm not sure what to look at next.
Without access to the actual code I'm a bit stumped.


  I presume you've tried with the most recent CVS code?

>                       I got an email from one Ning Zhu
> asking me about the exact same problem he's having.

I doubt that's the same problem, actually.
80 characters isn't really that long, and snmptrapd has a
significantly different code base.   I suspect Ning Zhu's
problem is somewhat different.   It's not clear from the message
you quote whether this is a problem with AgentX, snmpd or snmptrapd.


Dave

P.S:  Can you please post messages to *one* of the mailing lists, not both.
      Increasing our workload won't help get things solved faster!

Re: Some observations with SilverCreek agent tester

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

"Michael J. Slifcak" wrote:
> 
> If you actually use that tool, do use File->Save Results before
> you change agent profile (context).  It gives a better summary
> than what I have cobbled here.
> 
> Hi.  This kind of testing should have been done several weeks ago.
> 
> Blame me for falling behind.

And more blame for not including the report... see attachment,
unpack and read the "summary" file.

Highlights:
 SNMPv1 SET for a read-only variable expects noSuchName,
                                but response times out
 SNMPv2 SET for a read-only variable expects notWritable,
                                but response times out
 icmpInMsgs count doesn't match expected ICMP echo reply count
 laLoadFloat is format FLOAT, but seen as OPAQUE
 ipDefaultTTL (OpenBSD) is 0 , s/b 1
 ifType instance error (OpenBSD), maybe nbr of interfaces is wrong ?
 ifSpecific is wierd (OpenBSD)


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

Re: Beef about cleanup messages

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

>>>>> On Thu, 30 Nov 2000 18:14:28 -0500, "Michael J. Slifcak" <msl...@is...> said:

Michael> When exiting the agent started with -f -C -L,
Michael> these NEW messages display :
 
Michael> Tree node iso has no parent
Michael> Tree node ccitt has no parent
Michael> Tree node joint-iso-ccitt has no parent

I've been meaning to go nuke these.  Niels?

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

=?ISO-2022-JP?B?GyRCNS82SDJINWEkYBsoQg==?=

[IER <tom...@f8...>]

突然のお手紙をお許しください。ご不要の方にはお詫び申し上げます。

私はIERジャパンの村尾広治と申します。オランダの某上場企業の1部門と契約して働いている独立事業主です。

貴重なお時間を尊重するためにダイレクトな表現をさせていただきます。貴方が現在の職業と収入に完全に満足していらっしゃるのでしたら、この情報はご不要です。どうぞ破棄してください。

まだ読みつづけていらっしゃいますか？よかったです！

私達も数年前は貴方と全く同じでした。ハードワークに値する収入を得る方法を探していました。すばらしい情報を提供させていただきます。方法はあります。私達は生きた証人です。私のパートナーである米国人のケースでは、この方法でほんの数年の間に年商20億円の事業を築き上げています。これは米国でベストセラーとなったリチャード¥ポー著「WAVE4」　に紹介されている実話です。私自身昨年、初年度において年商7400万円の事業を築く事ができました。今後さらに急成長する見通しです。

私達は当初、本業を離れるリスクを冒すことなく週に数時間程度のパートタイムでスタートしました。そして…

・資金なしに、
・従業員なしに、
・売掛金なしに、
・買掛金なしに、
・一般管理費なしに、

自宅ベースのSOHOビジネスとして築き上げました。

どのようにしてこれを実現できるかをご紹介させていただくことができます。できすぎていると思われるでしょうか？できすぎていたら調べないで通り過ごすべきでしょうか？私達はそうは思いませんでした。

私達にとってパートナーである会社はアムステルダム株式市場に株式を公開している100年以上の歴史を持つ上場企業で、王室よりロイヤルの称号を戴いているオランダを代表する企業数社の1つです。また私が所属する部門はナスダック100銘柄指数企業で、フォーチュン誌により2年連続米国最急成長企業100社に選ばれています。株式公開企業ですから、事実を検証するのに必要な全ての情報は公開されています。

私達は米国で急成長しているこの会社のドットコムビジネスを日本に導入するプロジェクトに取り組んでいます。ウォールストリートのアナリスト、ソロモン・スミス・バーニーはインターネット業界の動向を予測するために全てのドット・コム企業を調査した結果、このプロジェクトを業界の次世代リーダーと予測しています。

多忙な中からフレックスタイムで時間、才能、能力を投資することによって、もう１つの選択肢を切り開くことを望む人を少数求めています。なぜなら多忙な人こそ優秀で生産性が高いからです。私達が米国最急成長企業であることからわかるように、このユニークな新しいビジネスモデルはすでに確立され、実証されています。このプロジェクトに参加することで、どのような利益を享受することができるのかを、さらにお知りになりたい思っていただけましたら、末尾の欄にご記入の上、Eメールでご連絡をください。共通の関心を持つ事ができるかどうかを判断するための資料をご送付させていただきます。

この招待状は貴方が今までに受け取った中で最も貴重な情報です。貴方の意欲と才能を求める手紙を、米国の最急成長企業から最後に受け取ったのはいつのことでしょうか？これほどのポテンシャルを持つビジネスベンチャーにリスクなしに参加するチャンスが最後にあったのはいつのことでしょうか？

ご返事をお待ちしております。ありがとうございました。

敬具

村尾広治

Independent Executive Recruiters Japan
Koji Murao
Email: tom...@f8...

＊ クロススポンサリングを意図したものではありません＊

-------------------------------------------------------------------------------------------------------------

資料請求申しこみ

氏名		[		                                                            ]
年齢		[	　　　                                                            　　 ]
郵便番号　　　　  [〒		                                                            ]
住所		[							]
電話番号       	[	　                                                            　　　　 ]
携帯電話番号	[	                                                            	]

Some observations with SilverCreek agent tester

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

If you actually use that tool, do use File->Save Results before
you change agent profile (context).  It gives a better summary
than what I have cobbled here.


Hi.  This kind of testing should have been done several weeks ago.

Blame me for falling behind.

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

Beef about cleanup messages

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

When exiting the agent started with -f -C -L,
these NEW messages display :
 
    Tree node iso has no parent
    Tree node ccitt has no parent
    Tree node joint-iso-ccitt has no parent

These messages are not logged when snmpd forks
the foreground process.

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

Beef about some error messages from agent

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

kvm_read error messages do not identify the kernel
symbol in the failure report.  They should!


The associated auto_nlist report should either incorporate
the kvm_read cruft, or be dropped, because it doesn't
add any useful information.
 

The next two lines are repeated as a set


  kvm_read(*,0,0x40076710, 140) = -1: kvm_read: Bad address
  auto_nlist failed on cnt at location 0

Recent agent change fails test on linux

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

All tests passed (when agentx and examples/example are added to the
module list).

This afternoon, I dimly recall seeing a change made to one of
the agentx files.  One test fails.  Details are attached.


27:testing AgentX illegal SET handling support...FAIL
Summary: 30 / 31 succeeded.


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

Re: URGENT: [Fwd: Maximum size of a trap packet]

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

>>>>> On Thu, 30 Nov 2000 13:28:55 -0800, Sudheer Tumuluru <stu...@re...> said:

Sudheer> I originally sent this message about 10 days but it got
Sudheer> drowned in the flood of emails. Could somebody please help on
Sudheer> this issue? I got an email from one Ning Zhu asking me about
Sudheer> the exact same problem he's having. His messages follows mine
Sudheer> below...

AgentX packets are significantly larger than SNMP packets, so you
might have to increase it beyond what you have the snmp packet limit
set to.

However, 64000 would be enough I think...

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

Re: snmpconf needs testing

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

>>>>> On 30 Nov 2000 22:00:42 -0000, ni...@ba... (Niels Baggesen) said:

Niels> That looks quite reasonable. How about a man page?

Man page?  Whats that?

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

Re: snmpconf needs testing

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

On 30 Nov 2000 09:40:26 -0800 Wes Hardaker <wjh...@uc...> wrote:
>
>If anyone has a chance, build of the cvs snapshot and run "snmpconf -g
>basic_setup" and see if it looks ok.  (or just snmpconf to test other
>aspects of it).

That looks quite reasonable. How about a man page?

/Niels

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

URGENT: [Fwd: Maximum size of a trap packet]

[Sudheer T. <stu...@re...>]

Hi,
        I originally sent this message about 10 days but it got drowned in the flood
of emails. Could somebody please help on this issue? I got an email from one Ning Zhu
asking me about the exact same problem he's having. His messages follows mine below...

Thanks,
Sudheer

Sudheer Tumuluru wrote:

> Hi,
>         Dave, thanks for the help. I changed the SNMP_MAX_MSG_SIZE #define from
> 1472 to 64000. It works perfectly when doing an snmp_send(). But when the snmpd
> gets a big packet and needs to forward it to a subagent, it gives a segmentation
> fault and quits in some instances. At other times, the subagent gets
> RESERVE1, RESERVE2 and some ACTION updates but stops after that without COMMIT. I
> suspect this is something to do with AgentX packet size. I checked snmp_agent.h but
> found the SNMP_MAX_PDU_SIZE already set to 64000. Is there any other place where
> there is a size limit and if it is, how can I extend it?
>
> TIA,
> Sudheer
>
> Dave Shield wrote:
>
> > > Looks like the PDU I'm trying to send chokes when it reaches a little over 300
> > > OIDs (20 + (13 entries per row * 22 rows)). Some of the OIDs have octect
> > > string values and I guess the exact number varies depending on the length of
> > > these strings.
> >
> > That sounds plausible.  Checking the code, the size of the encoded PDU is
> > restricted by the size of the (static) buffer used to build it. This is set
> > at SNMP_MAX_MSG_SIZE  - defined to be 1472 bytes.
> >
> >   One option would be to increase the value of this definition, so that
> > your normal requirements do not hit this limit.
> >
> > >              I am just looking for a method to tell me when the PDU is full
> > > so I can stop stuffing it further with OIDs.
> >
> > Sorry - there isn't any such indication.  And the current design of the
> > packet encoding is such that it wouldn't be simple to add one.  There's
> > no concept of "partially building" a packet - the expectation is that
> > the packet is complete before it's encoded.
> >
> >   It might be possible to hack something together - perhaps by calling
> > 'snmp_build' and testing the return code (and discarding the encoded
> > packet), but this would probably be fairly inefficient.
> >
> >   The alternative would be to build the full packet as you'd like to
> > send it, and strip off excess OIDs until the send call succeeds.
> >
> > Sorry if this isn't of much help to you.
> >
> > Dave
> >
> > _______________________________________________
> > Net-snmp-users mailing list
> > Net...@li...
> > http://lists.sourceforge.net/mailman/listinfo/net-snmp-users

> Subject: RE: Maximum size of a trap packet
> Date: Wed, 29 Nov 2000 01:39:20 -0500
> From: "Ning Zhu" <nz...@op...>
> To: "'Sudheer Tumuluru'" <stu...@re...>
>
> Hello!
>
> I saw your post from a snmp mailing list.  I am having almost exactly
> problems as you did - when the msg length exceeds 80 characters, the
> snmptrapd can't seem to receive it correctly.
>
> Would you be so kind to tell me how you solve it? - I lost track of all
> follow-ups on this question.
>
> Many thanks!
> N.
>

Re: lm_sensors hack

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

>>>>> On Wed,  8 Nov 2000 09:04:21 +0000 (GMT), Martin Oldfield <m...@ma...> said:

Martin> I've hacked some support for Linux's lm_sensors into snmpd. I
Martin> hesitate to call it a patch because my knowledge of MIBs and
Martin> whatnot is essentially non-existent.

Sorry for the delay.

Martin> 1. To get some feedback and improvements from people who actually
Martin> understand SNMP and have experience in dealing with it.

You can always post it to the list, and hopefully someone can take a
look at it.

Martin> 2. To get this code folded into the main distribution.

Then you definitely need to give it to us!  Anyway, if you send it to
the list and it looks reasonable, then I'll (or someone'll) add it to
the CVS tree for the next release.  If you need a particular copyright
on it that is different than the standard one that covers the package,
it may be placed in the contrib directory of if the copyright's don't
interoperate well.

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

Re: smux-peer and net-snmp

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

>>>>> On Thu, 09 Nov 2000 15:03:13 +0900, "N.Tanaka" <nob...@ro...> said:

N> I have made smux-peer program and be trying to communicate to snmpd
N> from net-snmp-4.1.2. When snmpd received setRequest from manager, I
N> could see 2 getRequests from snmpd and 1 setRequest from snmpd like
N> below.

N> smux-peer           snmpd             manager
N> ----+----           --+--             ---+---
N> |                 |                  |
N> |                 |<---setRequest--- |
N> |<---getReq------ |                  |
N> | --------------->|                  |
N> |                 |                  |
N> |<---setReq------ |                  |
N> | --------------->|                  |
N> |                 |                  |
N> |<---getReq------ |                  |
N> | --------------->|                  |
N> |                 |                  |
N> |<---SOut-------- |                  |
N> | --------------->|                  |
N> |                 | ---getResponse-->|
N> |                 |                  |

N> (1) Is this normal ? (I have thought only one setRequest from
N> snmpd would apper in this situation ).

1 is normal, just because of the way we implement set support in our
agent.  I'm confused as to why you'd be getting the second one though.

Unfortunately, no one is actively maintaining the smux code at the
moment which is one reason that your question has not gotten answered
sooner...

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

Re: Minimal size of SNMP daemon

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

>>>>> On Thu, 23 Nov 2000 11:56:12 +0100, Eri...@RS... said:

Eric> When I run the configure-file with the option
Eric> "-with-out-mib-modules "mibII snmpv3 ucd_snmp"" to exclude the
Eric> optional MIB modules, the options seem to have no effect, the
Eric> size of the agent binaries/libraries stays the same as without
Eric> the option. Is there an error in configure ?

Um, now it should definitely decrease in size.  Maybe the rebuild
didn't go as planned?  Start with a "make distclean" before running
configure and see if that helps.

Eric> The variable name "$new_without_mib_modules" in the file
Eric> configure.in appears only once in the file. I suppose it´s
Eric> missspelled and should read "$new_with_out_mib_modules" ?? (like
Eric> line 715,720,748....)

Whoops.  Thanks, I'll fix that.

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

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

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

>>>>> On Mon, 27 Nov 2000 18:53:39 +0530, Shshank_Sharma <Shs...@ht...> said:

Shshank> How does an SNMP manager form valid object identifiers(OIDs)
Shshank> for tabular variables that have multiple variables indexing
Shshank> them ?  Does it just do an snmpwalk from a higher level and
Shshank> see what are the valid indices by inpecting the returned OIDs
Shshank> ? or does it have some way of forming them itself ?

oid indexes are concatinated together to the end of the OID.  You
probably need to read a good book on MIBs to truly appreciate how it
works though...  In short, managers can figure out ahead of time what
the oid to use should be if they know the index values.

Shshank> Question # 2

Shshank> Do I need to give *valid* values for the indexing variables
Shshank> in a table instance when I am creating that instance in the
Shshank> sub-agent implementation and registering it with the master
Shshank> agent ?  (By *valid*- I mean the actual values of those
Shshank> parameters as present at the NE) Or can I just initialize
Shshank> those variables to some default values ?

Your subagent can do whatever it wants if it registers the entire
table instance with the master agent.  It itself is then responsible
for whatever lies beneath the registered OID and you're responsible
for returning the right results.  The master agent doesn't need to
know about yoru indexes at all.

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

Re: finger in snmp

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

>>>>> On Mon, 27 Nov 2000 10:55:18 -0400 (GMT-0400), "Barrios G. Carlos I." <ba...@in...> said:

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

There is probably a mib to support a list of currently logged in
users, but I don't know what it is off the top of my head and the
net-snmp agent doesn't support it.

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

NOTICE: sourceforge.net problems

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

The sourceforge download server is currently being upgraded and our
files (and most other projects) have disappeared at the moment.  The
good news is that in the long run, the download server is supposed
to be an even better machine, but in the mean time we need to wait for
the synchronization between the sourceforge machines to finish.
Apparently its taking a bit longer than sourceforge expected.

Thanks for the patience,
-- 
Wes Hardaker
Please mail all replies to net...@li...

AgentX client crashes master agent

[Timo K. <tk...@in...>]

Hi,

I think this was discussed earlier (at least in some part), but problem is following:
When subagent using AgentX is killed (for example, with ctrl+c), the next request to the subtree
which subagent provided will crash the master agent. I tried this with 4.2pre2 package as well 
with the sources from the cvs.

So I spent this day with Insure++ figuring out what goes wrong and finally got into following solution
In the snmplib/snmp_api.c, in the function snmp_sess_select_info(), lines 4344 and 4348, I added a call
unregister_mibs_by_session(slp->session) and unregister_mibs_by_session(slp) just before the call to
snmp_close(). 

As I have no idea how the master agent acutally works internally, this was merely trial-and-error style
fix.. is there a "real" fix for this? (It really seems that subagent won't get unregistered when it dies..)

And also, adding those unregister...() calls, is not enough, as if I break the subagent while a snmp request
is running, master agent will still crash, this time SIGPIPE in snmp_sess_async_send...

Timo Kujala

snmpconf needs testing

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

If anyone has a chance, build of the cvs snapshot and run "snmpconf -g
basic_setup" and see if it looks ok.  (or just snmpconf to test other
aspects of it).

If you don't want to do a make install, then:

  cd local
  ./snmpconf -c snmpconf.dir -g basic_setup

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

Re: last call for 4.2 problems/patches

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

>>>>> On 30 Nov 2000 10:49:47 +0100, Frank Strauss <st...@ib...> said:

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

Did both.  Thanks!

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

Use of SNMP package with 64-bit variables?

[Sorrell, A. <Al_...@tr...>]

Using V4.1.2/SNMP module 3.1.0, Perl 5.005_02 on Sparc Solaris 2.6

I'm trying to retrieve some of the HC (High Capacity) 64-bit counters from
the ifMIB.ifMIBObjects.ifXTable.ifXEntry table. If I print the results as
%20s, values are correct,
however if printed as %20d, they sometimes are not (presumably because of
being > 2**32-1 ??).
First example is using %20d, 2nd using %20s, 3rd using %20g:

$ HC_test.pl      
       Variable name            New Value            Old Value Status
        ifHCInOctets            248688099            248164185     
     ifHCInUcastPkts                 2025                 2026  BAD
 ifHCInMulticastPkts              2314562              2309761     
 ifHCInBroadcastPkts                    0                    0     
       ifHCOutOctets          -1429376126          -1436955206     
    ifHCOutUcastPkts                    0                    0     
ifHCOutMulticastPkts              5475261              5463230     
ifHCOutBroadcastPkts                    0                    0     

$ HC_test.pl      
       Variable name            New Value            Old Value Status
        ifHCInOctets            257924972            257384839     
     ifHCInUcastPkts                 1388                 1416  BAD
 ifHCInMulticastPkts              2389617              2384850     
 ifHCInBroadcastPkts                    0                    0     
       ifHCOutOctets           2991183468           2982079925     
    ifHCOutUcastPkts                    0                    0     
ifHCOutMulticastPkts              5667865              5654670     
ifHCOutBroadcastPkts                    0                    0 

$ HC_test.pl
      Variable name            New Value            Old Value Status
        ifHCInOctets          2.72713e+08          2.71929e+08     
     ifHCInUcastPkts                  530                  528     
 ifHCInMulticastPkts           2.5131e+06          2.50523e+06     
 ifHCInBroadcastPkts                    0                    0     
       ifHCOutOctets          3.19811e+09          3.18229e+09     
    ifHCOutUcastPkts                    0                    0     
ifHCOutMulticastPkts          5.96761e+06           5.9495e+06     
ifHCOutBroadcastPkts                    0                    0     

Since Perl normally handles numbers internally as floating point variables, 
am I safe in assuming that normal arithmetic operations are valid?

Al