home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.hp
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!hobbes.physics.uiowa.edu!news.uiowa.edu!icaen!dsiebert
- From: dsiebert@icaen.uiowa.edu (Doug Siebert)
- Subject: HP-UX 8.07 bug causes crash (& related info on SO_LINGER)
- Message-ID: <1992Dec17.233412.28542@icaen.uiowa.edu>
- Sender: usenet@icaen.uiowa.edu (UseNet News daemon)
- Organization: Iowa Computer Aided Engineering Network, University of Iowa
- Date: Thu, 17 Dec 1992 23:34:12 GMT
- Lines: 38
-
-
- This may be a known problem....if not, could others confirm it so that it can
- hopefully be corrected (and perhaps someone using 9.0 can check if this can
- occur under that rev as well)
-
- The problem occurred when I used the SO_LINGER option of setsockopt and set
- the linger time to a positive value (9 in this case) I had a few processes
- later become stuck while attempting to close the socket, and attached 'xdb'
- to them. That worked fine, but when I attempted a traceback of the process
- (this was as root, I don't know if this occurs as a normal user) using 'T'
- the machine (an HP 9000/710) crashed and rebooted itself. This happened twice
- since I didn't make the connection the first time, because the power to the
- system isn't 100% reliable, and silly fools are storing boxes of paper inches
- from its power cord.
-
- I don't know if this had something to do with the SO_LINGER option, or just
- the fact the process was stuck in kernel code when I attached to it. For
- obvious reasons I haven't tested this in depth -- this system receives a lot
- of use and I can't go randomly crashing it to test my assertions. Obviously
- the workaround I've found to this problem is "don't do a traceback on a
- running process stuck in kernel code" :-)
-
-
- [BTW, this is unrelated, but SO_LINGER doesn't work as the man pages claim it
- does (but then neither does the BSD networking code on which HP-UX's is based)
- Setting SO_LINGER with any non-zero linger time causes the process to get
- stuck indefinitely (I know, I've had 3 of these on my system for the past few
- days) but setting it to 0 does cause the proper behavior of closing the socket
- immediately even with bytes in the send queue, keeping all sockets that are
- "CLOSING" or "CLOSED" from your netstat list, freeing up the buffers they
- would otherwise be wasting.]
- --
- /-----------------------------------------------------------------------------\
- | Doug Siebert | "I don't have to take this abuse |
- | Internet: dsiebert@isca.uiowa.edu | from you - I've got hundreds of |
- | NeXTMail: dsiebert@chop.isca.uiowa.edu | people waiting in line to abuse |
- | ICBM: 41d 39m 55s N, 91d 30m 43s W | me!" Bill Murray, Ghostbusters |
- \-----------------------------------------------------------------------------/
-