|
Bugzilla – Full Text Bug Listing |
| Summary: | connmgr.nlm notifies nsure audit of a user logout after the logout is finished | ||
|---|---|---|---|
| Product: | [Novell Products] Nsure Audit | Reporter: | Brian Hieb <bhieb> |
| Component: | Instrumentation - NetWare FS | Assignee: | Doug Eddy <deddy> |
| Status: | VERIFIED FIXED | QA Contact: | Will Hoffman <whoffman> |
| Severity: | Normal | ||
| Priority: | P2 - High | CC: | rich.brunt |
| Version: | 1.0.3 | Keywords: | English |
| Target Milestone: | 2.0 (Huron) FCS | ||
| Hardware: | Other | ||
| OS: | Other | ||
| Whiteboard: | |||
| Found By: | Customer | Services Priority: | |
| Business Priority: | Blocker: | --- | |
| Marketing QA Status: | --- | IT Deployment: | --- |
|
Description
Jane Kestner
2005-04-20 17:08:09 UTC
This has been recently duplicated. The suggestion being made on this bug, basically says that auditnw should keep a list of all of the connection numbers and the data that goes with them, and then when the logout is triggered, reference that information to log the data. This would probably make auditnw unstable, and make auditnw hog memory. This suggestion also assumes that auditnw would not ever be unloaded, as if we were unloaded, we would then lose the little connection table we keep. Something to consider would be, what happens when someone logs out after auditnw has been loaded. What happens if someone logs in before auditnw gets loaded? We will not have matching information, or worse mismatched information when the log out event occurs. We should push the OS team to work on creating a new API that we can register for. We discussed this in a defect review meeting a month or two ago. Those present were Tom Larsen, Rich Brunt, Brian Hieb, Doug Eddy, and Rob Beauchamp. We discussed the pros and cons of an API and we came to a conclusion that a new API for Novell Audit would not be an appropriate thing to do. Rich suggested that the Audit team do what Ross Doxey suggests, and that is to keep track of connections similar to what connection manager does. Yes, it is a duplication of effort, but it does ensure that we get the data we are looking for. As far as memory is concerned, it does not appear to be very memory intensive. The downside to this is time. It would take time to write the code and then to test it. Since they had Novell Audit 2.0 (Huron) in beta, it was decided that they would not be able to add it in until a future time. I think this is more of a defect than an enhancement. We sell our product as an auditing solution and we tell our customers that we audit logins and logouts. Yet when you look closely at the product, the logout information provided is incomplete. And that doesn't look good to our customers because in their eyes, it simply does not work. Talked to Troy Rovig on CPR team. We need to understand the changes that were made to Login.c back on March 2000. I called and left a message for Ross Doxey to see if we can track down the original defect from March 2000 and find out who fixed it then, etc. Maybe if we can get the original defect number from March 2000, we can look it up and see what happened. Added 2 fields in what gets passed to the callback for anyone registered for the event EVENT_LOGOUT_CONNECTION. This was requested by the NSure audit team. Now, in addition to the connection number, the login name and the network type, length and address are passed to the callback. This won't affect anyone not looking for these fields, but for those that are, they will be available. (LOGIN.C) Fixed in v65.sp5. The additional info that is now returned besides the connection number is the username and the network address info. This is composed of three parameters: UINT32 connection number /*existing parameter previously returned */ unsigned char * loginName /* this is the new 2nd parameter returned consisting of a byte length preceeded login name */ unsigned char * addressBuffer /* This is the new 3rd parameter returned consisting of a buffer that has the network address type in the first byte, the network address length in the second byte, and then the network address according to the length of the 2nd byte */ This needs to be assigned to Doug now, I think. Either that or marked resolved and verified, once Doug is finished with the fix Doug Eddy has worked this through with the NW OS folks. They provided another callback to get the proper information. With the new connmgr from 10/25/05 and the auditnw.nlm for 1.0.3 of 10/25/05 this appears to be fixed. The same fix must be merged into the 2.0 tree. The 1.0.3 bug is fixed. The bug is now set to monitor getting this merged into the 2.0 FCS product. now merged into 2.0 *** Bug 105016 has been marked as a duplicate of this bug. *** |