SNMP_Session.pm 1.14: open_trap_session now takes an
optional ipv4only argument. It defaults to 1, but by
passing 0, you can create a session that accepts traps over IPv6 in
addition to IPv4. The receive_trap method has been
enhanced by returning an additional value which is the address
family. There is a new receive_trap_1 method, which
differs from receive_trap in that it simply returns the
trap and a sockaddr structure, rather than the trap, the host
address, the port, and the address family.
BER.pm 1.14: Added POD documentation. Started adding automated
unit tests.
SNMP_Session.pm 1.14: Added POD documentation.
SNMP_util.pm 1.13: Parse OIDs in NOTIFICATION-TYPE names. Change
by Mike Mitchell/jaccobs.
SNMP_Session.pm 1.13: Fixed the optional socket-reuse code to
handle IPv4 and IPv6 sockets separately.
SNMP_Session.pm 1.12: Upgraded to Artistic License 2.0 upon a
suggestion from Tom Callaway.
SNMP_util.pm 1.12: Rewritten MIB parsing code from Mike
Mitchell.
SNMP_util.pm 1.11: Improvement to loop detection in snmpwalk.
Should cope with more broken agents. Change by Mike Mitchell.
SNMP_Session.pm 1.10: Mike Fischer reported that on some systems,
notably Linux, recv() may block even though a
preceding select() for readability has returned
successfully. Once condition where this can happen is when a UDP
checksum on an incoming datagram doesn't verify. To avoid our
library from blocking in this case, we have to
pass MSG_DONTWAIT to recv(). This behaviour is
selected by a new optional argument to receive_response_3,
and used by request_response_5.
SNMP_util.pm 1.09: Fix from Mike Mitchell for parsing qualified
symbolic OIDs.
SNMP_Session.pm 1.08: Fixed a bug in the SNMPv2c version of
map_table_start_end that would cause errors when the
$end argument is actually being used. Thanks to Jan van
Keulen for submitting the patch.
SNMP_Session.pm 1.07: Fixed strict subs error with
newer versions of Perl. Thanks to Gerry Dalton for spotting
this.
SNMP_Session.pm 1.06: decode_trap_request now
understands SNMPv2 inform requests in addition to SNMPv1 traps and
SNMPv2 traps. Contributed by Andrew Cornford-Matheson.
SNMP_util.pm 1.06: snmpwalkhash fix by Laurent
Girod.
index.html: The code is no longer available from
ftp.switch.ch, but via HTTP from
www.switch.ch.
SNMP_Session.pm 1.05: No change; version incremented for
SNMP_util.pm and BER.pm.
BER.pm 1.05: Internal restructuring: The new (BER.pm 1.02)
pretty-printer registration method is now used for most standard SNMP
types, too.
SNMP_util.pm 1.04: The subroutines snmpget,
snmpgetnext and snmpset will now detect when they
are called in scalar context, and return only the first (and typically
the only) of the retrieved values in this case.
SNMP_util.pm 1.03: Added a missing part of Mike Mitchell
pretty-printer registration patch.
BER.pm 1.02: Additional decoders (``pretty printers'') can now be
registered and unregistered for type codes using the
register_pretty_printer and
unregister_pretty_printer subroutines. Implemented by Mike
Mitchell.
BER.pm 1.02: Corrected encoding of large integer-like
values.
SNMP_util.pm 1.02: snmpwalkhash now takes an additional
optional argument that should be a reference to a hash. If used,
snmpwalkhash will insert the retrieved values into that
hash.
BER.pm 1.01: Properly parse the variant length format in
integers. Thanks to Milen
Pavlov for pointing out this bug. Also, decoding should be
slightly more efficient in general, because one substr()
operation per (sub-) object has been eliminated.
SNMP_Session 1.00: Added Luc Pauwels' implementation of the
use_16bit_request_ids option. Some SMC devices seem to
require this.
SNMP_Session 0.99: Improved request ID generation so that
avoid_negative_request_ids is always obeyed.
SNMP_util 0.99: Added Gauge32 support to
snmpset. Thanks to Christopher J. Tengi for the
idea. Mike Mitchell provided new code that supports all known
types.
SNMP_Session 0.98: Portability fix in IPv6 support.
SNMP_util 0.98: Support encoding of OIDs containing multiple
quoted strings.
SNMP_util 0.97: Added support for TimeTicks values in
snmpset. Patch from Joerg Kummer
SNMP_Session 0.97: Exported
$SNMP_Session::default_avoid_negative_request_ids, by
Philippe Simonet.
SNMP_Session 0.97: IPv6 support added by Valerio Bontempi and
Lorenzo Colitti.
SNMP_Session 0.96: Intermediate version with IPv6 support; not
published.
BER.pm 0.95: Fixed operator precedence bug in
decode_sequence().
SNMP_Session.pm 0.94: A new slot capture_buffer has been
added to the SNMP_Session classes, courtesy Jakob Ilves. See the sample
script test/capturetest.pl for an example of how to use
this.
BER.pm 0.94: New subroutines pretty_generic_sequence,
decode_generic_tlv, courtesy Jakob Ilves.
BER.pm 0.94: decode_by_template, decode_oid,
decode_sequence, decode_string: Explicitly signal
error when the PDU is undefined or too short. This improves error
handling for some malformed PDUs, as generated by certain test
suites.
SNMP_Session.pm 0.93: There's a new variable
$default_avoid_negative_request_ids. If it is set to a
non-zero value, newly created SNMP_Session objects will be
configured so that only request IDs in the range 0..231-1
will be used. This is needed to work around a bug in several SNMP
agents. If you sometimes see requests fail, and the error message
always shows a negative request ID when that happens, please notify
the vendor of your agent of the bug. While the vendor fixes the
problem, you can set the variable mentioned above to work around the
bug. The problem has been described in a few mails on the
mrtg-developers mailing list that can hopefully be found here
in the archive.
SNMP_util.pm 0.93: snmpwalk now will walk multiple OID
trees simultaneously. If more than one OID is passed to the function,
the returned OID:value pairs will contain the entire OID
instead of just the leaf from the starting point.
SNMP_util.pm 0.93: snmpwalkhash now passes the textual
OID of the starting point as the seventh argument to the passed-in
hash function.
SNMP_util.pm 0.92: New subroutine snmpmaptable4, which
allows specifying the max-repeaters.
SNMP_util.pm 0.92: New MIB parsing code. Most MIBs are processed
in one pass, but if an OID cannot be fully resolved, an extra pass is
done. Also, OBJECT-GROUP entries are now
recognized.
SNMP_util.pm 0.92: The string-to-OID conversion routines now
transparently convert quoted strings to instance indexes. For
example, "/usr" (including the quotes) would be converted
to 4.47.118.97.114.
SNMP_Session.pm 0.91: Fixed a bug in map_table_start_end
for SNMPv2_Session (which uses get-bulk) which had
caused a superfluous query after the entire table has already been
provably traversed. Thanks to Michael Deegan for pointing this
out.
SNMP_util.pm 0.90: New version from Mike Mitchell.
snmpwalkhash and snmpwalk now share most
code.
SNMP_util.pm 0.90: Corrected handling of the optional port number in
snmpopen.
test/if-counters.pl: Support 64-bit counters, and use
ifAlias to portably get interface descriptions.
SNMP_util.pm 0.89: New version from Mike Mitchell, with new
snmpwalkhash subroutine by Laurent Girod.
SNMP_Session.pm 0.89: Made lenient_source_port_matching
the default.
SNMP_util.pm 0.89: Fixed a long-standing bug where the code would
generate PDUs with request-ids with large positive values,
violating the SNMP spec which mandates that request-ids be
Integer32s. Thanks to Sergio Macedo for
finding this bug.
SNMP_util.pm 0.88: Added missing use Carp;
statements in packages SNMPv1_Session and
SNMPv2_Session. Thanks to Michael Deegan.
SNMP_util.pm 0.87: No change from 0.86, but increased version
number to reflect change in BER.pm (SNMPv2 exception codes).
BER.pm 0.87: pretty_print now silently returns undef
when decoding SNMPv2 exception codes (noSuchObject,
noSuchInstance, or endOfMibView, see RFC 1905).
Original patch by Bert
Driehuis.
BER.pm 0.86: pretty_print now silently returns undef
when given an undefined value, rather than issuing incomprehensible
warnings.
SNMP_util.pm 0.86: New snmpmaptable subroutine.
This is a more user-friendly version of map_table and is
described in the README.SNMP_util
file. From Mike Mitchell.
SNMP_util.pm 0.86: Support for the awesome get-bulk
operator, both directly through the new snmpgetbulk
subroutine, and transparently via snmpwalk when the
session's SNMP version is >= 2 and the use_getbulk
slot is set (as it is by default). From Mike Mitchell.
SNMP_Session.pm 0.85: If a local address is specified in
snmpopen, don't convert it using inet_aton,
because this is handled by the Socket library. Fix from Mike McCauley.
SNMP_Session.pm 0.85: Added
lenient_source_port_matching slot to the session object.
Set this to communicate with weird SNMP agents that send the response
from a port other than 161. Suggestion from Henri Gomez.
SNMP_util.pm 0.84: New version from Mike Mitchell.
snmpopen now parses an optional hash argument for
options. Also reintroduced defaulting of the UDP port to 161 for
``normal'' sessions.
SNMP_Session.pm 0.84: Clarified documentation concerning the
%pretty_oids hash, upon a suggestion from Alistair Mills.
SNMP_Session.pm 0.83: The source address in response packets is
now ignored when matching responses against outstanding queries. In a
couple of previous revisions, the source address in response packets
had to match the destination address in the corresponding query.
Unfortunately some agents may use a different source address in
responses. If you want the strict behavior back, you can send the
lenient_source_address_matching slot of the session object to
zero.
SNMP_Session.pm 0.83: The source address for outgoing packets can
now be specified as an additional optional argument to
open. If you don't specify it, the system will choose the
source address by itself, usually corresponding to the interface on
which packets are sent.
SNMP_Session.pm 0.83: Fixed a bug which had caused requests to be
resent upon receipt of packets which don't match the outstanding
query.
SNMP_Session.pm 0.82: Fixed retry logic to avoid sending a last
retry without waiting for the response anymore. Thanks to Brett T Warden for the
fix.
BER.pm 0.82: OIDs with only two subids can now be encoded. The
most common case is the "null" OID (0.0). Thanks to Brett T Warden for pointing out
that this didn't work.
BER.pm 0.82: pretty_print() now handles UInteger32 objects.
Patch by Brett T
Warden.
BER.pm 0.81: Subids in the range 2^31 - 2^32-1 are now encoded
correctly. Thanks to Rik
Hoorelbeke for noticing the problem.
SNMP_Session.pm 0.81: A cosmetic bug in the SNMPv2 version of
map_table (which uses get-bulk) was corrected. The
user-supplied function is now always called on the same number of
arguments. Before this, missing values at the end of a table row
would lead to the function being called with fewer arguments. Now
there will be undef values for those, too. Thanks to Ulrich Schmid for pointing this
out.
SNMP_Session.pm 0.80: A portability bug was fixed in the code
that matches incoming responses to outstanding requests. The bug had
manifested itself, notably on some FreeBSD versions, by timeouts
waiting for responses, because the library thought the responses came
from another address than the corresponding requests had been sent
to.
SNMP_Session.pm 0.79: A new variable
$SNMP_Session::recycle_socket has been introduced. When this
variable is set to a non-zero value (the default is zero), all newly
created SNMP_Session objects will share the same UDP socket. This
saves file descriptors and system calls, but will cause problems with
multiple outstanding SNMP requests on different session objects.
Applications which don't perform parallel/asynchronous SNMP requests
can safely set this variable to reduce OS overhead somewhat.
Suggestion from Ulrich
Schmid.
SNMP_Session.pm 0.79: The handling of incoming packets from
unexpected addresses has been cleaned up. If a packet is received
from an IP address other than the one to which the request has been
sent, this packet is silently ignored, as mandated by the SNMP
standard.
SNMP_Session.pm 0.79: Receive-only session objects (such as the
ones created by open_trap_session now have undef
as the remote_addr value, rather than IP address
0.0.0.0.
BER.pm 0.79: There are new exported subroutines for encoding
different types of values: encode_uinteger32,
encode_counter32, encode_counter64,
encode_gauge32.
SNMP_Session.pm 0.78: The map_table implementation for
SNMPv2 sessions has been completely rewritten to reliably support
tables with holes in them. Note that this hasn't been completely
validated or tested yet, but at least it has been found to handle
common cases reasonably.
BER.pm 0.77: Added support for long integers, so that
Counter64 values can be handled without risk of losing
precision (due to automatic coercion to floating-point representation
by Perl). When a BER-encoded integer is so long that it might not fit
in a 32-bit unsigned integer, then we use Math::BigInt
arithmetics to convert it. Warning: code which calls the
decoding functions should be prepared to handle Math::BigInt
values if Counter64 values can be accessed. In most respect,
those long integers behave just like ordinary integers. A notable
exception is that they print with a leading ``+'' sign.
SNMP_Session 0.77: The SNMPv2 implementation of
map_table_start_end has been enhanced to cope with the case
where a response PDU ends with a truncated table row, courtesy Paul E. Erkkila. Note that tables with
missing entries still aren't handled correctly when SNMPv2 (and thus
the get-bulk operator is used).
SNMP_util 0.73: Added a fix from Mike Mitchell (originally Johannes Demel) to treat
OBJECT-IDENTITY like OBJECT IDENTIFIER.
SNMP_Session 0.76: Added some debugging support for
map_table, which still doesn't work reliably when the SNMPv2
get-bulk operator is used.
test/if-counters.pl: Added -c option to enable
Cisco-specific variables (which are no longer retrieved by default as
in previous versions).
SNMP_Session 0.75: Fixed a bug in versions 0.73-0.74 where
creation of trap listener sockets would fail. Rather than using
bind to bind to the trap port, we now pass a
LocalPort argument to INET->new.
SNMP_Session 0.75: Parse SNMPv2-Trap-Requests in addition to
SNMPv1 ones. A caller can tell whether an SNMPv1 or an SNMPv2 request
has been received by testing whether the SNMPv1-specific fields are
defined. The sample script test/trap-listener has been
updated to understand SNMPv2 Traps.
SNMP_Session 0.75: New subroutine v2_trap_request_send
which sends SNMPv2 Trap PDUs. The sample script
test/trap-send has been extended to generate either type of
trap on request.
SNMP_Session 0.74: Put under copyright and Artistic
License.
SNMP_util 0.72: Changed snmpgetnext so it will return
the next lexicographical larger OID number, even if it is not in the
same OID tree. Before the change snmpget would discard the return
value if the OID wasn't in the same tree. In snmpMIB_to_OID,
an "unitialized variable" warning was removed. (Changes by Mike Mitchell, the author)
SNMP_Session 0.73: The UDP socket associated with each SNMP
session object is now created using IO::Socket::INET->new.
Before this change, a Perl file descriptor name was generated for each
new socket. Apparently this caused a file descriptor leak. The new
code is much cleaner and doesn't have that problem anymore. Hopefully
the newly introduced dependency on the Socket::IO module is not a
problem. Thanks to Andrew
W. Elble for suggesting this change.
BER.pm 0.72: A new variable,
$BER::pretty_print_timeticks, has been introduced to give
users control over the degree of pretty-printing of TimeTicks
values. If left at the defaults, TimeTicks values will be
converted to strings such as 14 days, 6:56:07. If you
set it to zero, the same value will simply pretty-print as
123456789, which should be interpreted in units (ticks)
of 10ms.
SNMP_util.pm 0.71: New subroutines snmpLoad_OID_Cache and
snmpQueue_MIB_File for loading MIBs in compact format.
SNMP_util.pm 0.71: snmpset now accepts ipaddr
as a type specifier.
SNMP_util.pm 0.70: Mike
Mitchell added code for parsing MIB files. See the description of
snmpMIB_to_OID in README.SNMP_util.
SNMP_util.pm 0.69: Enable parsing of community strings which
contain @ characters.
SNMP_util.pm 0.58: Check for errors from encode_oid, so
that illegal OIDs generate error messages. Allow for suppression of
warnings by setting $SNMP_Session::suppress_warnings to a
value greater than one.
SNMP_Session.pm 0.68: Added methods receive_request and
decode_request to support SNMP agents. Contributed by Mike McCauley.
SNMP_Session.pm 0.67: Implement a map_table_start_end
method for SNMPv2c_Session which uses get-bulk. See
Walking Tables With get-bulk
for more information.
SNMP_Session.pm 0.66: Fix from Alan Nichols to the
handling of error replies. In the last few revisions of SNMP_Session,
requests to which the agent responded with a non-zero
errorStatus were erroneously retried.
SNMP_Session.pm 0.66: When the errorStatus is zero, we
don't care about the errorIndex. This makes us liberal
enough to cope with very old versions of the CMU agent code which
sometimes put a non-zero errorIndex in normal response
packets.
BER.pm 0.66: Changed the $VERSION number to be in line
with SNMP_Session.pm's.
BER.pm 0.66: Changed encode_oid so that it signals an
error when passed an illegal Object ID, such as one whose first subid
isn't 0, 1 or 2. This should help people who try to use output from
CMU/UCD SNMP directly in MRTG.
SNMP_Session.pm 0.65: Fix error message when binding to UDP port
fails (Jonathan
Zhukovsky).
SNMP_util.pm 0.57: Small change to avoid warnings with "-w" when
session parameters are defaulted.
SNMP_util.pm 0.56: New subroutine snmpmapOID, Mike Mitchell.
SNMP_Session.pm 0.64: Fix of a bug in the detection of missing
responses by Mike Mitchell.
SNMP_Session.pm 0.63: Fixed response matching logic to ignore
out-of-sequence responses. This has been a long-standing bug which
made it very hard to use a single session object for multiple queries
to devices that are (sometimes) slow in responding.
SNMP_Session.pm 0.62: Scoping problem with pretty_address()
fixed.
Makefile.PL: New file which allows for easy installation
according to standard Perl convention. Kudos to Clinton Wong.
SNMP_Session.pm 0.61: The sender address of the last response
received for a session is stored in
$session->{'last_sender_addr'}. In connection with
broadcast or multicast addresses, this can be used to discover SNMP
agents listening to specific communities, as illustrated in
test/discover.
SNMP_session.pm 0.59: Added methods set_timeout,
set_retries, and set_backoff that can be used to
tune the retransmission algorithm. Fixed a bug in map_table
that would cause an index of "0" to terminate the table walk.
SNMP_util.pm 0.54: First version to be distributed with the
package, courtesy Mike
Mitchell.
BER.pm 0.58: Added encode_timeticks() subroutine. This
was used in the sample script for sending
traps, but was only defined in test/trap-test.pl.
Thanks to Gary Berger for
noticing this.
BER.pm 0.57: Added encode_ip_address() subroutine on a
suggestion by Mike
Diehn.
test/ber-test.pl: Added more test cases contributed by
Mike Mitchell.
SNMP_Session 0.57: table walking support See ``Walking Tables'' below for how this is
used.
BER 0.56: New encode_int() subroutine contributed by Mike Mitchell. Fixes incorrect
encoding in the range +-2^15-2^23 and generalized to integers of any
size.
BER 0.55: Fix an arithmetic bug in the uptime
pretty-printer. Kudos to Niels
Bakker for noticing this.
SNMP_Session 0.56: Fix a bug which occurs when an error should be
signaled while an SNMPv1_Session is being opened. Noticed by Dan Cox and Iouri Pakhomenko.
BER 0.52: Ignore a leading dot when encoding an OID. This is to
avoid trouble when people cut&paste OIDs from CMU/UCD SNMP, where
a leading dot is used to mark a "fully qualified" OID.
SNMP_Session 0.55: The SNMP_Session module no longer calls
warn when the variable
$SNMP_Session::suppress_warnings is set to non-zero (it is
zero by default). The error message from SNMP_Session can be
retrieved as $SNMP_Session::errmsg.
SNMP_Session 0.54, BER.pm 0.51: Errors in the BER module are now
passed upwards by the SNMP_Session module. Before this change, one
could not distinguish malformed SNMP responses from no response at
all. The BER module now no longer calls die(), but returns undefined
values.
Added test/arp which prints the NetToMedia table from a
remote host.
SNMP_Session 0.53: Avoid passing numeric IP addresses to
inet_ntoa(), by Daniel
L. Needles
SNMP_Session 0.52: setRequest support based on code contributed
by Matthew Trunnell. See
``Set Requests'' below for how this is
used.
SNMP_Session 0.51: Improved error messages by printing the
session in error messages if possible, and the OID and error message
whenever the agent sends back an error.
SNMP_Session 0.50, BER 0.50: The BER and
SNMP_Session modules both have version numbers according to the
convention in Exporter.pm. That is, you can now insist on a
minimal version of the modules by saying e.g.
use BER "0.50";
use SNMP_Session "0.52";
The initial version numbers are 0.50 for both modules.
The pretty printer should now print unsigned 32-bit values (such
as Counters and Gauges) correctly, i.e. values larger than
231 are printed as large positive
numbers rather than negative numbers. Note that this can cause
problems depending on how you handle the output of the pretty
printer, since those string representations of large numbers may not
be convertible to integers using atoi() or similar
functions.
The subroutines in SNMP_Session never call die() anymore
if it encouters error situations. Instead, they issue a warning and
return undef. Brad
Ritchie managed to convince me that library code should never
die. Unfortunately I haven't revised BER.pm yet, so
subroutines related with BER transfer syntax encoding and decoding
may still die.
The code has been cleaned up to use more of the standard
functionality of the Perl 5 Socket.pm module. That should
have eliminated some potential portability problems (and delegated
responsability for potential bugs :-). Note that this means that the
code requires Perl 5.002 or later.
Both source files now make extensive use of strict for
better compile-time error checking. Please notify me in case you
have any problems because of this.
The library now attempts to retransmit queries for which no
reponse has been received during a given time. The default
parameters for the retransmission logic have been discussed at length
in the mrtg mailing list, and seem to work quite well, both
against overloaded routers that simply drop some SNMP requests and
against routers that are behind slow or lossy links. If you have
feedback on the default parameters, please drop me an e-mail.
When I implemented the retransmission logic, I also fixed
handling of request IDs. In older versions, the request ID was never
changed between reqeuest, which could lead to (late) resposes being
associated with the wrong request. Now the request ID is incremented
for each request, and mismatching responses are ignored. For
retransmissions, the request ID isn't changed. If we did change it,
we could estimate response time and implement an adaptive
retransmission algorithm. This has been left for further
study.
