git: cfb1a82d49c5 - releng/14.4 - bsnmp: do not act on an SNMPv3 discovery message that carries bindings
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Tue, 17 Nov 2026 17:04:21 UTC
The branch releng/14.4 has been updated by markj:
URL: https://cgit.FreeBSD.org/src/commit/?id=cfb1a82d49c5fa7ef32c1d9a79b160b60138b9b7
commit cfb1a82d49c5fa7ef32c1d9a79b160b60138b9b7
Author: Maxim Konovalov <maxim@FreeBSD.org>
AuthorDate: 2026-11-13 03:00:02 +0000
Commit: Mark Johnston <markj@FreeBSD.org>
CommitDate: 2026-11-16 21:16:22 +0000
bsnmp: do not act on an SNMPv3 discovery message that carries bindings
An SNMPv3 message with an empty engine id and empty user name is handled
as snmpEngineID discovery: snmp_pdu_auth_user() sets SNMP_MSG_AUTODISCOVER
and returns without a USM user, and snmp_pdu_auth_access() then grants the
message access. A genuine discovery request carries no variable bindings
and only elicits a Report, but a discovery PDU that did carry bindings was
executed anyway -- so a GET or SET ran with no authentication and without
the VACM access check.
This affects two cases. With an empty user name the message is wholly
unauthenticated, so anyone able to reach the agent could issue a SET. It
also affects a configured noAuthNoPriv user: such a user has no digest to
verify, so it too can set the AUTODISCOVER flag (an authenticated request
before time synchronisation) and have a bindings-carrying PDU bypass VACM
and operate outside its views.
Refuse any AUTODISCOVER message that arrives with variable bindings
instead of acting on it. Empty-binding discovery, the authenticated
time-window exchange, and ordinary authenticated SNMPv3 are unaffected.
While here, free the PDU on the SNMPD_INPUT_FAILED path of snmpd_input().
Every other path releases it, but the failed path consumed the datagram
and returned without freeing the decoded bindings -- so a rejected PDU
leaked any heap they held, such as the buffer for an OCTET STRING in a
SET. Rejecting a bindings-carrying discovery message now reaches that
path, so an unauthenticated client could otherwise grow the agent's
memory without bound.
For that free to be safe, snmp_input_start() must leave the PDU in a
defined state on every return. It zeroed the PDU only after the initial
snmp_pdu_snoop() check, so a malformed datagram that failed the snoop
returned SNMPD_INPUT_FAILED with the caller's PDU uninitialised; freeing
it would then walk a garbage binding count. Zero the PDU up front,
before the snoop, so all of these paths leave an empty PDU that frees
cleanly.
Approved by: so
Security: FreeBSD-SA-26:89.bsnmp
Security: CVE-2026-127742
Reviewed by: harti
Fixes: 135f7de5ddb6
---
contrib/bsnmp/snmpd/main.c | 17 ++++++++++++++++-
1 file changed, 16 insertions(+), 1 deletion(-)
diff --git a/contrib/bsnmp/snmpd/main.c b/contrib/bsnmp/snmpd/main.c
index 928b84121f82..d1377756b64e 100644
--- a/contrib/bsnmp/snmpd/main.c
+++ b/contrib/bsnmp/snmpd/main.c
@@ -496,6 +496,10 @@ snmp_input_start(const u_char *buf, size_t len, const char *source,
ret = SNMPD_INPUT_OK;
+ /* Zero the PDU before anything can return, so a failure path that
+ * runs before the header is decoded still leaves it safe to free. */
+ memset(pdu, 0, sizeof(*pdu));
+
/* look whether we have enough bytes for the entire PDU. */
switch (sret = snmp_pdu_snoop(&b)) {
@@ -508,7 +512,6 @@ snmp_input_start(const u_char *buf, size_t len, const char *source,
}
b.asn_len = *pdulen = (size_t)sret;
- memset(pdu, 0, sizeof(*pdu));
if ((code = snmp_pdu_decode_header(&b, pdu)) != SNMP_CODE_OK)
goto decoded;
@@ -645,6 +648,17 @@ decoded:
(pdu->engine.engine_boots == 0 || pdu->engine.engine_time == 0)) {
snmpd_usmstats.not_in_time_windows++;
ret = SNMPD_INPUT_FAILED;
+ } else if ((pdu->flags & SNMP_MSG_AUTODISCOVER) != 0) {
+ /*
+ * A discovery message (RFC 3414) may only elicit a Report and
+ * carries no variable bindings of its own. Reaching here with
+ * bindings means the message was not authenticated as a user
+ * (empty security name, or a noAuthNoPriv user): acting on it
+ * would run a GET or SET that bypasses USM authentication and
+ * the VACM access check. Refuse it.
+ */
+ snmpd_usmstats.unknown_users++;
+ ret = SNMPD_INPUT_FAILED;
}
if ((code = snmp_pdu_auth_access(pdu, ip)) != SNMP_CODE_OK)
@@ -1088,6 +1102,7 @@ snmpd_input(struct port_input *pi, struct tport *tport)
* check the access first. We don't want to return an error response
* to a proxy PDU with a wrong community */
if (ierr == SNMPD_INPUT_FAILED) {
+ snmp_pdu_free(&pdu);
/* for streaming transports this is fatal */
if (pi->stream)
return (-1);