This is *NOT* a commit candidate yet (and definitely not until
after release). It's only tentitive... I had to rip up some core
pieces to get the kernel to properly optimize vm_map_entry
structures for the downward-generated-stacks.
http://www.backplane.com/FreeBSD4/
http://www.backplane.com/FreeBSD4/guard-2.diff
http://www.backplane.com/FreeBSD4/guard2.c
With this patch it should be possible to trivially modify
the threads code. You pretty much keep what you have except
use MAP_GUARDED|MAP_STACK, and deal with skipping over the guard
pages.
Resource limits are still an issue. It turns out that the
MAP_STACK code does not deal with the stack resource limit
well at all -- sometimes it catches it, sometimes it doesn't.
At the moment MAP_GUARDED doesn't even try but that does not
prevent you from using the stack memory areas to do your mmaps.
You can use MAP_GUARDED|MAP_STACK now. This is effectively
MAP_GUARDED but will cause the mmap() to fail if you specify
a fixed address that already contains mappings (which is what
you want, I think). i.e. that is what MAP_STACK semantics
normally do - return a failure on a conflict.
There are probably bugs, much more testing is necessary.
-Matt
Matthew Dillon
<[EMAIL PROTECTED]>
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
