We have been fighting spinlock abends with WINSOCK for months. Several
weeks ago we installed WSOCK6K on 9 servers and not a single spinlock
abend since. Yippee! That would be great except we now get "Cache memory
allocator out of available memory" from WS2_32 asking for a huge block
(over 4GB...) of memory. Since the request is so big, Netware simply
rejects the request and the server continues working fine. Our SEGSTATS
look OK.

We used to get the spinlock abends only when a backup (EMC/Legato) kicked
off. Now we only get the cache memory error when a backup kicks off
(sometimes...about 20% of the time). Novell has asked for a core dump.
Since the server does not abend they gave me a breakpoint to enter based
on the size of the memory request. Problem is the number keeps
changing...-27066, -28544, -27988 etc. Is there a way to set a breakpoint
to force the server to drop into the debugger on ANY cache memory
allocation error?