[PATCH] Cygwin: mmap: use 64K pages for bookkeeping

[Corinna Vinschen]

On Jan 20 10:36, Ken Brown wrote:
> On 1/20/2025 6:49 AM, Corinna Vinschen wrote:
> > Nice idea, but this may not do what is expected if the mapping is an
> > anonymous mapping, leaving the protection or mapping of trailing pages
> > in a wrong state, isn't it?
> > 
> > Can we easily make sure the type of mapping (file vs anon) is known
> > at the time of rounding, so the rounding is performed differently?
> I hadn't thought of that.  Actually, I *think* the record length is already
> a multiple of 64K for an anonymous mapping.

You're right, of course.  For an anonymous mapping, mmap() forces len to
be a multiple of wincap.allocation_granularity() anyway.

So yeah, looks like your previous patch does the trick.  Feel free to
push it, but I would love to see a comment in mmap_record::match() to
explain why we do it this way.


Thanks,
Corinna