Menu
▾
▴
#56 Modified snmp_agent.c
There are two problem fixes in snmp_agent.c
1. In snmpset, if the value passed is wrong, it was
not returning errorindex.Problem was due to, in case
of SET request failure, snmp_agent.c was not
preserving error_index value.
2. In case of SET request's ACTION case fails, then
UNDO was called on all varbinds. Modified code calls
UNDO for only successful varbinds and FREE for all the
varbinds.
Ex:
Consider there are two varbinds to be SET & second
varbind fails in ACTION phase.
Existing code calls UNDO for BOTH the varbinds.
Code I have changed, calls UNDO only for first varbind
and returns. Then goes to FREE phase, which will clean
up for both varbinds.
snmp_agent.c
Logged In: YES
user_id=322303
As the second problem fix is change in fundamental
architecture and therefore never be accepted. Thats
why *now* I have updated snmp_agent.c with only *1st*
problem fix.
IE,
In snmpset, if the value passed is wrong, it was
not returning errorindex.Problem was due to, in case
of SET request failure, snmp_agent.c was not
preserving error_index value.
Attached is the patch file.
Difference to 4.2.3
Returning correct error index
Error in UNDO phase is ignored
Logged In: YES
user_id=322303
Updated according to Mr.Srini Kode's comment:
Hi Latha,
I looked at the diff file you had sent and still see the
problem I pointed out.
I believe we must not return the error index got from the
UNDO operation,because the network manager will like to
know the error index of the object that caused the error in
ACTION state because of which the UNDO state is reached.
So in your file, if you do not do "error_index = asp-
>index" in UNDO
state
then it should be ok.
The same thing goes for error status also.
Thanks
Srini
Attached is the latest patch file
Logged In: YES
user_id=88893
The main error index patch has now been applied to the
4.x branch of the CVS tree. Investigations showed that
the new approach in the main development branch handled
this correctly anyway.
Thanks for your work on this
Dave