[ECOS] SNMP GETBULK leaks 50k per request -- security issue

[ECOS] SNMP GETBULK leaks 50k per request -- security issue

[Tad]

A GETBULK request requiring > 8k bytes in the response 
(snmp_api.c:PACKET_LENGTH or sendto max)
forgets to free the 50k pdu malloc'ed.

Should be able to crash any ecos snmp system with a couple:
bulkget -Cr50 -v 2c -c public 192.168.1.199 system system system system 
system icmp system icmp
which will eat 5 retries x50k at a time

Basically, the snmp_agent.c we're using is POS full of memory leaks if 
snmp_send or other errors occur.

I grabbed the latest v4.2 branch from SF of snmp_agent.c, snmp_agent.h, 
and snmp_api.h which seem to compile for ecos with virtually no changes 
(used the ECOS includes for snmp_agent.c)

The latest snmp_agent.c seems to do a nice job of cleaning up memory and 
has a slightly faster SET operation.

http://net-snmp.cvs.sourceforge.net/net-snmp/net-snmp/agent/snmp_agent.c?view=log&r1=1.100&pathrev=V4-2-patches
et. al.

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: SNMP GETBULK leaks 50k per request -- security issue

[Tad]

...when snmp_send() fails.

Tad wrote:
> A GETBULK request requiring > 8k bytes in the response 
> (snmp_api.c:PACKET_LENGTH or sendto max)
> forgets to free the 50k pdu malloc'ed.
>
> Should be able to crash any ecos snmp system with a couple:
> bulkget -Cr50 -v 2c -c public 192.168.1.199 system system system 
> system system icmp system icmp
> which will eat 5 retries x50k at a time
>
> Basically, the snmp_agent.c we're using is POS full of memory leaks if 
> snmp_send or other errors occur.
>
> I grabbed the latest v4.2 branch from SF of snmp_agent.c, 
> snmp_agent.h, and snmp_api.h which seem to compile for ecos with 
> virtually no changes (used the ECOS includes for snmp_agent.c)
>
> The latest snmp_agent.c seems to do a nice job of cleaning up memory 
> and has a slightly faster SET operation.
>
> http://net-snmp.cvs.sourceforge.net/net-snmp/net-snmp/agent/snmp_agent.c?view=log&r1=1.100&pathrev=V4-2-patches 
>
> et. al.
>

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: SNMP GETBULK leaks 50k per request -- security issue

[Tad]

If patched, snmp_api.c also needs 3 changes shown in SF tree.  2 points 
initialize ->data = 0 and snmp_free_var() free's if data != 0.

Tad wrote:
> ...when snmp_send() fails.
>
> Tad wrote:
>> A GETBULK request requiring > 8k bytes in the response 
>> (snmp_api.c:PACKET_LENGTH or sendto max)
>> forgets to free the 50k pdu malloc'ed.
>>
>> Should be able to crash any ecos snmp system with a couple:
>> bulkget -Cr50 -v 2c -c public 192.168.1.199 system system system 
>> system system icmp system icmp
>> which will eat 5 retries x50k at a time
>>
>> Basically, the snmp_agent.c we're using is POS full of memory leaks 
>> if snmp_send or other errors occur.
>>
>> I grabbed the latest v4.2 branch from SF of snmp_agent.c, 
>> snmp_agent.h, and snmp_api.h which seem to compile for ecos with 
>> virtually no changes (used the ECOS includes for snmp_agent.c)
>>
>> The latest snmp_agent.c seems to do a nice job of cleaning up memory 
>> and has a slightly faster SET operation.
>>
>> http://net-snmp.cvs.sourceforge.net/net-snmp/net-snmp/agent/snmp_agent.c?view=log&r1=1.100&pathrev=V4-2-patches 
>>
>> et. al.
>>
>

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss