global_client_failed_oplock_break

[Jeremy Allison]

On Fri, Nov 09, 2007 at 03:54:04PM -0800, Zack Kirsch wrote:
> Hey tech-
>  
> Does anybody know where this variable came from? I understand its
> intent, but I wonder if it actually solves a real problem somebody
> observed, or if it was just put in place to prevent "bad clients" ?

A combination of both, reall. When first writing the oplock code I was paranoid
about screwing them up :-).

There were many instanced of Win9x clients not responding to
oplock breaks (the theory was that the TCP stack on Win9x
could go "deaf" for long enough after sending a packet
that a fast Linux box could hit that window with a quick
reply). I've seen many network traces of clients missing
a break, but never completely repeatable.

Jeremy.