[omniNotify] Occasional notifd crash

[Mark Zimmerman]

Greetings:

I am starting to see occasional core dumps of notifd under Solaris 9
(sparc). I don't have a lot of information yet but I thought I would
throw this out here in case anyone is still reading this list and has
insight or experience with this one.

The following line showed up in the debug printout:

DBG[lib/RDIOplocks.cc:458]: *** Internal error: RDIOplocks::free_entry
called with an entry that appears to already be on the free list

Then, some time later a core dump occurred. The core file shows this:

t at 15 (l at 15) terminated by signal SEGV (no mapping at the fault address)
0x7fffffff7f9c72dc: acquire+0x0004:     ldx      [%i0 + 80], %i2
Current function is omni_thread_wrapper (optimized)
  451           me->run(me->thread_arg);
(dbx) where
current thread: t at 15
  [1] RDIOplockEntry::acquire(0x8a8d20, 0x1008c9230, 0x0, 0x7fffffff7ed39f40, 0x
1c48, 0x8a8d20), at 0x7fffffff7f9c72dc 
  [2] SupplierAdmin_i::obj_gc(0x7fffffff7fba1fc8, 0x1008c9230, 0x0, 0x0, 0x258, 
0x7fffffff7a3fbae4), at 0x7fffffff7f8d4538 
  [3] EventChannel_i::gcollect(0x100160830, 0x5, 0x7fffffff7a3fbca8, 0x4, 0x10, 
0x5), at 0x7fffffff7f929b54 
=>[4] omni_thread_wrapper(ptr = ???) (optimized), at 0x7fffffff7e203708 (line ~4
51) in "posix.cc"


This was running at a remote site with practically all debug and
reporting turned off, unfortunately.

-- Mark