have to unlock object and page queues before sleeping

KNF, and 64-bit safety.

null pointer test (from Ralph)

prototype clustered pageout interface,
keep vmmeter stats.

convert to use new queue structures

finish up new device pager: all pagers get offsets

4.4BSD snapshot (revision 8.1); add 1993 to copyright

hack: initialize vm_page_max_wired (max wired user memory) for mlock

mach 2.5/3.0 inspired changes:
1. dynamically compute inactive_target before each invokation of
vm_pageout_scan (as a function of current active/inactive pages)
2. make inactive_target larger to i

mach 2.5/3.0 inspired changes:
1. dynamically compute inactive_target before each invokation of
vm_pageout_scan (as a function of current active/inactive pages)
2. make inactive_target larger to improve chances of reclaim and
provide a better approximation of LRU
3. allow dirty pages to be reclaimed
4. correct computation of page_shortage to avoid excess wakeups
from vm_page_alloc
also some basic code cleanup

show more ...

convert vm_page bit fields to flags

be prepared to handle VM_PAGER_ERROR

new include format; lint

vm_stat is merged into vmmeter structure

numerous statistics move to vm_stat

keep page frame marked busy while in a pager
pmap_remove_all -> pmap_page_protect

update to new CMU copyright and distribution terms

fix includes; new proc struct; rm *_NULL

adopted from Mach 2.5