Hello everyone. I my name is Wojtek and I am new to the list. I would like to get your opinion on the problem that has been troubling me lately.
I put together my own table closely following this example: http://net-snmp.sourceforge.net/dev/agent/data__set_8c-example.html The main difference between the tables in the example and mine is that mine has around 23 columns. The "large" number of columns makes my agent crash on set requests. I suspect that this is due to a bug that reveals itself when operating on larger tables (with larger number of columns that is, and not necessarily many rows). Initially the agent would crash randomly, but I was able to get a crash on demand. I would issue set requests to two rows in a table in one snmpset, twice. That is, I would do snmpset table.row1 = v1 table.row1 = v2 ... table.row1 = v23 \ table.row2 = w1 table.row2 = w2 ... table.row2 = w23 snmpset table.row1 = v1 table.row1 = v2 ... table.row1 = v23 \ table.row2 = w1 table.row2 = w2 ... table.row2 = w23 The second call would crash the agent. (I tried to reproduce this on the table from the tutorial, but it seems to be too small; I was able to reproduce this on slightly larger tables with a more than 2 calls to snmpset) I started looking at the code of the helper handler and determined that the agent crashes at the very end of the set request, particularly on the STATE_COMMIT, when it executes (table_dataset.c:588) data = netsnmp_table_data_set_find_column(data, table_info->colnum); The "data" structure is corrupted, and since this is obtained from the "row" structure that would mean that "row" is corrupted. The interesting part is that whatever is computed in table_dataset.c:588 is never used on STATE_COMMIT, so skipping that line on this particular state seems to resolve the issue. However, I would like to get a deeper understanding of what really is going on, to make sure the change really fixes a bug, instead of hiding it deeper. So the question is: in the context of STATE_COMMIT what is the meaning of "row" variable? Why would we need to execute netsnmp_table_data_set_find_column on COMMIT? Wouldn't we want to operate on newrow, instead of row, just like in modes SET_RESERVE1 and SET_RESERVE2 (see table_dataset.c:557) (I am just guessing here)? Why is the "row" structure corrupted? Thanks for any clues. Hopefully this discussion will let me gain understanding of the code and help you improve the software. Wojtek ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier. Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Net-snmp-coders mailing list Net-snmp-coders@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
