ProCurve Series 2300 and 2500 Switches Release Notes

248
Software Fixes
■ IGMP — Checking whether an IP DA and/or an IGMP Group Address is a valid IP multicast
address before taking any IGMP action on it.
■ IGMP — Fixed Group-Specific Query (GSQ) timing in Normal Leave case to be a minimum
of 1 second (as IGMP standard specifies and as the GSQs advertise). This occurs when the
Querier forces an interval between GSQs. The internal GSQ timer value was increased to
force this change; otherwise, the time between when the timer is set and when it is serviced
can result in GSQs that are usually .2-sec to .6-sed apart, instead of the minimum 1-sec. Prior
to the fix, the switch sometimes pruned a group only ~3/4 of a second after sending out a
GSQ. Some end nodes take slightly longer than this to reply to the GSQ.
■ IGMP — Fixed the case where IGMPv3 Join contains an invalid IP Mcast address or a
reserved IP Mcast address in the IGMP Group Address field. Previously, the switch would
attempt to stop processing the Join and mistakenly double-free (one symptom can be a
"software exception at alloc_free.c ... buf_free: corrupted buffer") or double-forward the Join
packet. Now, the switch will simply bypass processing the "offending" sub-record and
continue with the rest of the Join, freeing or forwarding the packet only once.
■ IGMP — Not currently checking whether an IP DA and/or an IGMP Group Address is a valid
IP multicast address before taking any IGMP action on it.
■ IGMP — A Group-Specific Query (GSQ) timeout is currently .2 to .6 seconds, rather than
the specified default of 1 second.
■ IGMP — When an IGMP v3 Join contains an invalid IP Multicast address or a reserved IP
Multicast address in the IGMP Group Address field, the switch will attempt to stop processing
the Join, and mistakenly double-free, or double-forward the Join packet. One possible
symptom is a switch crash similar to:
software exception at alloc_free.c ... buf_free: corrupted buffer
■ Memory Leak — Related to the "WhatsUp Gold" network management application. Trig-
gered when configuring the enable/login password on the switch.
■ Security/Vulnerability — "Cross-site scripting" issue. One of the Nessus (see
www.nessus.org) tests fails, reporting: “Vulnerability found on port http (80/tcp).
■ SNMP — The OID ifAlias is defaulted to "not assigned", causing Network Node Manager to
log error messages. (The fix is to default ifAlias to a zero-length string, as stated in the MIB,
or make each port have a unique value.)
■ SNMP — The switch does not support community names other than PUBLIC in traps.
■ SNMP/Crash — A walkmib of the cdpCacheDeviceId OID using an ifIndex value of the
Default VLAN causes the switch to crash with a bus error similar to:
-> Bus error: HW Addr=0x5265766d IP=0x002592e8 Task='mSnmpCtrl'
Task ID=0x12c2158 fp: 0x00000005 sp:0x012c1e28 lr:0x00259430