[Bug 948] high CPU in sshd after tcp_wrappers deny

[bugzilla-daemon at mindrot.org]

http://bugzilla.mindrot.org/show_bug.cgi?id=948





------- Additional Comments From dtucker at zip.com.au  2005-02-15 09:27 -------
(In reply to comment #14)
> It seems to only be showing calls to read, a few of which take a lot of time,
> and most of which don't take much.

This read loop is probably the cause of the CPU utilization.  It would appears
that something, somewhere keeps retrying a read after a descriptor closes.

A couple of suggestions for more getting more info:
Can you use lsof or similar to find out what the descriptor is?  (It's 3 in this
case but there's no guarantee it'll be the same, so check with truss first).

Attach a debugger to the looping process and get a backtrace to find out where
it's stuck.  With gdb this is something like:
# gdb /path/to/sshd pid
(gdb) backtrace

If you use the sshd before it's stripped (ie from the build dir) then you will
get better info.  You don't have to be running the one with the debug symbols,
you can be running the stripped sshd and use the one with the debug symbols with
gdb as long as the two sshd's match.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.