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

Re: Compilation error with debugging off

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

>>>>> On Wed, 18 Oct 2000 19:38:34 -0700, Sudheer Tumuluru <stu...@re...> said:

Sudheer> I am getting the following compile error on the latest
Sudheer> checkout of cvs code. I ran the configure script with the
Sudheer> "--disable-debugging" option. Looks like "_DBG_IF_" does not
Sudheer> get defined when this option is on. I defined it as 0 in
Sudheer> snmp_debug.h in the #else part of "#ifndef SNMP_NO_DEBUGGING"
Sudheer> and it worked. I don't know why it is defined inside the
Sudheer> "#ifndef" as "snmp_get_do_debugging()" in the first place
Sudheer> when it could be set to 1.

I've changed this on the net-snmp cvs server so it should work in the 
future...

-- 
Wes Hardaker
Please mail all replies to ucd...@uc....

Duplicating snmp values at another location in the MIB

[MEDBERRY,DAVID (HP-Loveland,ex1) <dav...@hp...>]

All,

What is the simplest way to use/duplicate SNMP information at another place
in the MIB?

Example:

I want to report these values:

interfaces.ifTable.ifEntry.ifSpeed.1 = Gauge: 10000000
interfaces.ifTable.ifEntry.ifSpeed.2 = Gauge: 10000000


at:

myPrivateMib:ifSpeed.1 = Gauge: 10000000
myPrivateMib:ifSpeed.2 = Gauge: 10000000

Is there an easy way to get access to these values within a mib2c module:
myPrivateMib.c?

If anyone has sample code (or if this same thing is already done within the
ucd-snmp , er uh, net-snmp code somewhere) I would appreciate knowing about
it.

Thanks,

-dave medberry

Re: TCP_Count_Connections() and FreeBSD 3+

[Bert D. <dri...@pl...>]

On Tue, 24 Oct 2000, Michael J. Slifcak wrote:

>  I see these errors (in no particular order) when I run
> the agent as a non-privileged user on FreeBSD 3.4-STABLE :
  [snip]
> auto_nlist failed on boottime at location 0

I think Romain was not referring to auto_nlist failures, but rather to
an issue where kvm_read would try to read from unmapped kernel memory
(which I think the Bad Address message refers to). Probably a pointer
into never-never land...

> > Anyone out there working with FreeBSD 3.x or later?  I was getting
> > the following messages in snmpd.log everytime I snmpwalk'd a
> > FreeBSD 3.4 machine:
> >         TCP_Count_Connections - inpcb: Bad address
> >         kvm_read(*, 11bb, 0xbfbfaa84, 96) = 0: kvm_read: Bad address
> > 
> > I borrowed sysctl()-based code from netstat to get around this, but
> > I'm not convinced that I've done the Right Thing.

That code looks okay to me (I must admit I haven't delved real deep into
this particular bit of code, as I don't run 3.4 anywhere and didn't run
into the issue on 4.0).

In general, kvm access should be replaced with sysctl where possible (to
avoid requiring root access is one good reason, in addition, sysctl is a
much cleaner interface, with fewer opportunities to go awry, because
sysctl guarantees the data won't go foom while you're reading it).

The mibJJ code will have less kvm and more sysctl, at least for the
BSD's (did anyone got a chance to look at the diffs for mibJJ and *BSD
I sent, b.t.w.?)

Of course, in practice the sysctl interface isn't as well documented as
it might have been, but hey, it's a start...

Cheers,

				-- Bert

Bert Driehuis -- dri...@pl... -- +31-20-3116119
If the only tool you've got is an axe, every problem looks like fun!

Re: TCP_Count_Connections() and FreeBSD 3+

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

FYI -  From the current CVS snapshot,

bash-2.03$ agent/snmpd -p 8788 -f -C -L
init_kmem: kvm_openfiles failed: /dev/mem: Permission denied
bash-2.03$ agent/snmpd -p 8788 -f -C -L -r
Creating directory: /var/ucd-snmp
Creating directory: /var/ucd-snmp
UCD-SNMP version 4.2.pre0

 I see these errors (in no particular order) when I run
the agent as a non-privileged user on FreeBSD 3.4-STABLE :


auto_nlist failed on ipstat at location 0
auto_nlist failed on icmpstat at location 0
open: No such file or directory
auto_nlist failed on tcpstat at location 0
auto_nlist failed on tcb at location 0
TCP_Count_Connections - inpcb: No such file or directory
auto_nlist failed on udpstat at location 0
auto_nlist failed on cnt at location 7373654d
auto_nlist failed on bufspace at location 0
auto_nlist failed on cp_time at location 0
auto_nlist failed on boottime at location 0

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

Romain Kang wrote:
> 
> Anyone out there working with FreeBSD 3.x or later?  I was getting
> the following messages in snmpd.log everytime I snmpwalk'd a
> FreeBSD 3.4 machine:
>         TCP_Count_Connections - inpcb: Bad address
>         kvm_read(*, 11bb, 0xbfbfaa84, 96) = 0: kvm_read: Bad address
> 
> I borrowed sysctl()-based code from netstat to get around this, but
> I'm not convinced that I've done the Right Thing.
> 
> Any feedback?
> 
> Thanks,
> Romain Kang
> ro...@kz...
> 
> *** tcpTable.c.BAK      Mon Sep  4 15:09:08 2000
> --- tcpTable.c  Tue Oct 10 13:25:48 2000
> ***************
> *** 359,366 ****
> --- 359,368 ----
>   #ifndef solaris2
>   #ifndef linux
> + #ifndef freebsd3
> +
>   /*
>    *    Print INTERNET connections
>    */
> 
>   int TCP_Count_Connections (void)
>   {
> ***************
> *** 425,428 ****
> --- 427,490 ----
>         return(Established);
>   }
> + #else /* freebsd3 */
> +
> + int
> + TCP_Count_Connections()
> + {
> +       char *buf;
> +       const char *mibvar;
> +       struct tcpcb *tp;
> +       struct inpcb *inp;
> +       struct xinpgen *xig, *oxig;
> +       struct xsocket *so;
> +       size_t len;
> +       u_long proto;           /* for sysctl version we pass proto # */
> +       int     Established = 0;
> +
> +       proto = IPPROTO_TCP;
> +       mibvar = "net.inet.tcp.pcblist";
> +
> +       len = 0;
> +       if (sysctlbyname(mibvar, 0, &len, 0, 0) < 0) {
> +               snmp_log_perror("TCP_Count_Connections/sysctl");
> +               return 0;
> +       }
> +       if ((buf = malloc(len)) == 0) {
> +               snmp_log(LOG_ERR, "TCP_Count_Connections/malloc %d", len);
> +               return 0;
> +       }
> +       if (sysctlbyname(mibvar, buf, &len, 0, 0) < 0) {
> +               snmp_log_perror("TCP_Count_Connections/sysctl");
> +               free(buf);
> +               return 0;
> +       }
> +
> +       oxig = xig = (struct xinpgen *)buf;
> +       for (xig = (struct xinpgen *)((char *)xig + xig->xig_len);
> +            xig->xig_len > sizeof(struct xinpgen);
> +            xig = (struct xinpgen *)((char *)xig + xig->xig_len)) {
> +               tp = &((struct xtcpcb *)xig)->xt_tp;
> +               inp = &((struct xtcpcb *)xig)->xt_inp;
> +               so = &((struct xtcpcb *)xig)->xt_socket;
> +
> +               /* Ignore sockets for protocols other than the desired one. */
> +               if (so->xso_protocol != proto)
> +                       continue;
> +
> +               /* Ignore PCBs which were freed during copyout. */
> +               if (inp->inp_gencnt > oxig->xig_gen)
> +                       continue;
> +
> +               if ((tp->t_state == TCPS_ESTABLISHED) ||
> +                   (tp->t_state == TCPS_CLOSE_WAIT))
> +               {
> +                       Established++;
> +               }
> +       }
> +
> +       free(buf);
> +       return Established;
> + }
> + #endif        /* freebsd3 */
>   #endif
> 
> _______________________________________________
> Net-snmp-coders mailing list
> Net...@li...
> http://lists.sourceforge.net/mailman/listinfo/net-snmp-coders

TCP_Count_Connections() and FreeBSD 3+

[Romain K. <ro...@kz...>]

Anyone out there working with FreeBSD 3.x or later?  I was getting
the following messages in snmpd.log everytime I snmpwalk'd a
FreeBSD 3.4 machine:
	TCP_Count_Connections - inpcb: Bad address
	kvm_read(*, 11bb, 0xbfbfaa84, 96) = 0: kvm_read: Bad address

I borrowed sysctl()-based code from netstat to get around this, but
I'm not convinced that I've done the Right Thing.

Any feedback?

Thanks,
Romain Kang
ro...@kz...

*** tcpTable.c.BAK	Mon Sep  4 15:09:08 2000
--- tcpTable.c	Tue Oct 10 13:25:48 2000
***************
*** 359,366 ****
--- 359,368 ----
  #ifndef solaris2
  #ifndef linux
+ #ifndef freebsd3
+ 
  /*
   *	Print INTERNET connections
   */
  
  int TCP_Count_Connections (void)
  {
***************
*** 425,428 ****
--- 427,490 ----
  	return(Established);
  }
+ #else	/* freebsd3 */
+ 
+ int
+ TCP_Count_Connections()
+ {
+ 	char *buf;
+ 	const char *mibvar;
+ 	struct tcpcb *tp;
+ 	struct inpcb *inp;
+ 	struct xinpgen *xig, *oxig;
+ 	struct xsocket *so;
+ 	size_t len;
+ 	u_long proto;		/* for sysctl version we pass proto # */
+ 	int	Established = 0;
+ 
+ 	proto = IPPROTO_TCP;
+ 	mibvar = "net.inet.tcp.pcblist";
+ 
+ 	len = 0;
+ 	if (sysctlbyname(mibvar, 0, &len, 0, 0) < 0) {
+ 		snmp_log_perror("TCP_Count_Connections/sysctl");
+ 		return 0;
+ 	}
+ 	if ((buf = malloc(len)) == 0) {
+ 		snmp_log(LOG_ERR, "TCP_Count_Connections/malloc %d", len);
+ 		return 0;
+ 	}
+ 	if (sysctlbyname(mibvar, buf, &len, 0, 0) < 0) {
+ 		snmp_log_perror("TCP_Count_Connections/sysctl");
+ 		free(buf);
+ 		return 0;
+ 	}
+ 
+ 	oxig = xig = (struct xinpgen *)buf;
+ 	for (xig = (struct xinpgen *)((char *)xig + xig->xig_len);
+ 	     xig->xig_len > sizeof(struct xinpgen);
+ 	     xig = (struct xinpgen *)((char *)xig + xig->xig_len)) {
+ 		tp = &((struct xtcpcb *)xig)->xt_tp;
+ 		inp = &((struct xtcpcb *)xig)->xt_inp;
+ 		so = &((struct xtcpcb *)xig)->xt_socket;
+ 
+ 		/* Ignore sockets for protocols other than the desired one. */
+ 		if (so->xso_protocol != proto)
+ 			continue;
+ 
+ 		/* Ignore PCBs which were freed during copyout. */
+ 		if (inp->inp_gencnt > oxig->xig_gen)
+ 			continue;
+ 
+ 		if ((tp->t_state == TCPS_ESTABLISHED) ||
+ 		    (tp->t_state == TCPS_CLOSE_WAIT))
+ 		{
+ 			Established++;
+ 		}
+ 	}
+ 
+ 	free(buf);
+ 	return Established;
+ }
+ #endif	/* freebsd3 */
  #endif
  

subscriber

[jl <li...@bo...>]

DQo=

Compilation error with debugging off

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

              I am getting the following compile error on the latest
checkout of cvs code. I ran the configure script with the
"--disable-debugging" option. Looks like "_DBG_IF_" does not get defined
when this option is on. I defined it as 0 in snmp_debug.h in the #else
part of "#ifndef SNMP_NO_DEBUGGING" and it worked. I don't know why it
is defined inside the "#ifndef" as "snmp_get_do_debugging()" in the
first place when it could be set to 1.

gcc -g -O2 -Wall -Wstrict-prototypes -Wwrite-strings -Wcast-qual
-Wno-unused -Wno-char-subscripts -Dlinux -I/usr/include/rpm -I. -I.. -I.
-I./.. -c read_config.c -o read_config.o
read_config.c: In function `read_config_print_usage':
read_config.c:600: `_DBG_IF_' undeclared (first use in this function)
read_config.c:600: (Each undeclared identifier is reported only once
read_config.c:600: for each function it appears in.)
make: *** [read_config.lo] Error 1

-Sudheer
--
  ------------------------------------------------------------------------
Sudheer Tumuluru
stu...@re...
Software Development Engineer, Real Broadcast Network
RealNetworks Inc. http://www.realnetworks.com
  ------------------------------------------------------------------------

subscriber

[jl <li...@bo...>]

DQo=

(no subject)

[<ic...@wy...>]

Re: init_kmem()

[Bert D. <dri...@pl...>]

On Wed, 18 Oct 2000, David Wilusz wrote:

> I've used your agent successfully on Sol 2.6.  I assumed that a 2.6 build
> would work on Sol 2.7, but when I try to run it on the 2.7 box, I get this
> error: 
> init_kmem: kvm_open failed: Error 0

In general, any tool that uses kmem compiled on a given OS will only run
on that OS revision, and UCD-SNMP is no exception. I am not knowledgable
about the Solaris version, but given Suns track record I'd guess it to
be as true there. Solaris, actually, raises the bar by requiring a
rebuild per CPU type.

Background: kmem (Kernel memory, aka kvm, Kernel Virtual Memory) is a
method of getting at kernel internals by directly reading kernel memory
based on debugging symbols embedded in the kernel. Kernel internal
structures are by their very nature undocumented and do change between
versions. It is an Evil (if a necessary Evil) way of getting at
variables that are not exposed through a dedicated system call.
Operating systems have gotten better in this respect, with more and more
internals being exposed through system calls like sysctl(), but UCD-SNMP
is still playing catch-up.

Anyway, the long and short of it is: compiling UCD-SNMP on any version
of any OS results in a binary that is quite likely to be specific to
that version.

Cheers,

					-- Bert

Bert Driehuis -- dri...@pl... -- +31-20-3116119
If the only tool you've got is an axe, every problem looks like fun!

Anyone Interested In libsnmp.dll built by Cygwin tools?

[John C. W. <joh...@c-...>]

UCD-SNMP-Coders,

Does anyone have an interest in a libsnmp.dll built by the Cygnus (Cygwin)
tools?

Regards,
John

********************************************
John C. Westmoreland
Broadband Network Division
C-Cube Microsystems, Inc.
1778 McCarthy Boulevard
Milpitas, California 95035
Phone:  408-490-5111
Fax:    408-490-5508
E-Mail:  joh...@c-...
Pager: (800) 562-0687
Text Page: 800...@mo...
********************************************

"Most of the important things in the world have been
accomplished by people who have kept on trying
when there seemed to be no hope at all."
--Dale Carnegie