[ net-snmp-Bugs-1339477 ] _sess_read re-entered for same session

[SourceForge.net <no...@so...>]

Bugs item #1339477, was opened at 2005-10-27 09:11
Message generated for change (Tracker Item Submitted) made by Item Submitter
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112694&aid=1339477&group_id=12694

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: agent
Group: agentx
Status: Open
Resolution: None
Priority: 5
Submitted By: jmcdonnell (jm-slipstream)
Assigned to: Nobody/Anonymous (nobody)
Summary: _sess_read re-entered for same session

Initial Comment:
When an informsink is configured, the _sess_read 
function can be re-entered during processing of  
notifications from an AgentX sub-agent.  The function is 
not written to be re-entrant for the same session 
though.  In some situations, this can cause the original 
call to the function to close the connection or to simply 
crash.  A stack trace follows showing the path to the 
second call, notice that the sessp value is the same for 
both calls to _sess_read:

#0  _sess_read (sessp=0x81e6380, fdset=0xbfbff340) at 
snmp_api.c:5384
#1  0x80c633a in snmp_sess_read (sessp=0x81e6380, 
fdset=0xbfbff340) at snmp_api.c:5785
#2  0x80c4ad2 in snmp_read (fdset=0xbfbff340) at 
snmp_api.c:5343
#3  0x80a6286 in snmp_synch_response_cb 
(ss=0x817b900, pdu=0x821d500, response=0xbfbff41c, 
pcb=0x80a56d0 <snmp_synch_input>) at 
snmp_client.c:1006
#4  0x80a6321 in snmp_synch_response 
(ss=0x817b900, pdu=0x821d500, response=0xbfbff41c) 
at snmp_client.c:1044
#5  0x8084a6c in send_trap_to_sess (sess=0x817b900, 
template_pdu=0x821d300) at agent_trap.c:834
#6  0x8063471 in send_notifications (major=1, minor=7, 
serverarg=0x821d300, clientarg=0x0) at 
notification/snmpNotifyTable.c:246
#7  0x80dabf9 in snmp_call_callbacks (major=1, 
minor=7, caller_arg=0x821d300) at callback.c:225
#8  0x8084951 in netsnmp_send_traps (trap=-1, 
specific=-1, enterprise=0x8120a70, 
enterprise_length=10, vars=0x821c800, context=0x0, 
flags=0) at agent_trap.c:783
#9  0x8084991 in send_enterprise_trap_vars (trap=-1, 
specific=-1, enterprise=0x8120a70, 
enterprise_length=10, vars=0x821c800) at 
agent_trap.c:797
#10 0x8084afa in send_trap_vars (trap=-1, specific=-1, 
vars=0x821c800) at agent_trap.c:854
#11 0x80977a3 in agentx_notify (session=0x817bd00, 
pdu=0x817be00) at 
mibgroup/agentx/master_admin.c:456
#12 0x8097a82 in handle_master_agentx_packet 
(operation=1, session=0x817bd00, reqid=153, 
pdu=0x817be00, magic=0x0) at 
mibgroup/agentx/master_admin.c:548
#13 0x80c498f in _sess_process_packet 
(sessp=0x81e6380, sp=0x817bd00, isp=0x81f3480, 
transport=0x81f3400, opaque=0x81e8900, olength=106, 
packetptr=0x82221ac "\001\f", length=524) at 
snmp_api.c:5291
#14 0x80c5ca3 in _sess_read (sessp=0x81e6380, 
fdset=0xbfbff820) at snmp_api.c:5663
#15 0x80c633a in snmp_sess_read (sessp=0x81e6380, 
fdset=0xbfbff820) at snmp_api.c:5785
#16 0x80c4ad2 in snmp_read (fdset=0xbfbff820) at 
snmp_api.c:5343
#17 0x804c2be in receive () at snmpd.c:1123
#18 0x804bd7c in main (argc=3, argv=0xbfbff9cc) at 
snmpd.c:974

It looks like the code is trying to do an immediate read 
from the sub-agent.  I don't think it's possible to do this 
given that the original call to the function may contain an 
unprocessed partial message.  Also, trying to do an 
immediate read could cause a deadlock if the sub-agent 
isn't currently capable of responding.

I think this may be the cause of bug 1154117.

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112694&aid=1339477&group_id=12694