port/proc.c		1990/0227		Created.
		1990/03013		Print both pids involved in double-sleep.
		1990/03081		Use c to as current process instead of u->p. Add explicit process hierarchy a la Unix.
		1990/0309		Initialize p->pop in newproc.
		1990/03091		Allocate process locks all at once rather than on-demand at possibly bad times.
		1990/0310		Remove lock allocation loop after all.
		1990/0312		Print segments in debugging process list.
		1990/0322		Merge Proc.wait.use and Proc.wait.queue into single lock Proc.wait.
		1990/0324		Split Proc.wait back into Proc.wait.use and Proc.wait.queue.
		1990/0330		Remove segments from debugging process list.
		1990/0424		Code rearrangement: move lastvar declaration in kproc.
		1990/0509		Keep at most NBROKEN Broken processes around.
		1990/0614		Store time in milliseconds, not ticks (BUG FIX?)
		1990/0617		Use kmap in postnote to access Proc.upage.
		1990/0619		Set u->p->state = Wakeme before sleeping for more procs in newproc. (BUG FIX)
		1990/0629		Print p->r pointer in process dump.
		1990/0704		Print user names in process dump.
		1990/0722		Reset p->fpstate in newproc (BUG FIX?). Use kernel pgrp for all kprocs (was using u->p's).
		1990/0726		Cancel alarm after wakeup. (BUG FIX?)
		1990/0728		Don't cancel alarm after wakeup. (BUG FIX?)
		1990/0731		Add pc argument to m->intr function.
		1990/0801		Add unusepage of Proc.upage when freeing process. (BUG FIX?)
		1990/1101		Move broken process code into new addbroken. Add freebroken.
		1990/11211		Remove Chan* argument to error.
		1990/1124		Remove process hierarchy.
		1990/1211		Add invalidateu function, save, restore, clearmmucache.
		1990/1212		Move unusepage out of sched (deadlock BUG FIX). Rename save, restore to procsave, procrestore. Set p->kp for kprocs.
		1990/1220		Proper function prototype for DEBUG function.
		1990/1227		Add anyready.
		1991/0109		Avoid kmap of upage when it's our own.
		1991/0110		Save exit msg before dropping memory space (BUG FIX).
		1991/0209		Stop calling m->intr in sched.
		1991/0317		Clear p->pgrp in newproc (BUG FIX).
		1991/0318		Use memmove instead of memcpy.
		1991/0420		Split runq into runhiq, runloq.
		1991/0425		Add m->cs statistic. Clear p->kp in newproc (BUG FIX? only if any kprocs exit).
		1991/0430		Something about pidonmach and locking XXX.
		1991/0501		Code reformatting: blank line.
		1991/0507		Something about pidonmach. XXX
		1991/0513		Reset u->p->alarm in pexit (BUG FIX?)
		1991/0517		Close file descriptors before freeing memory (why? XXX)
		1991/0523		Handle u==0 in kmap case in postnote (BUG FIX).
		1991/0529		Comment why clearmmucache and flushmmu in fork.
		1991/0608		Rename DEBUG to procdump.
		1991/0705		Add Stopped. Formatting changes, extra comments.
		1991/0710		Move error routines into portable code.
		1991/0712		Don't close nil fgrp, egrp. (BUG FIX)
		1991/0717		Comment fixes. Avoid race in putseg cleanup (BUG FIX?)?
		1991/0720		Avoid deref nil in process dump.
		1991/0721		Remove bracketing prints in process dump.
		1991/0727		New sleep and wakeup: replace wokeup with notepending.
		1991/0805		postnote: only reset r->p if its us (seems dodgy? what's the race?)
		1991/0806		Add Rendez state, break out of rendezvous in postnote.
		1991/0807		More debugging print changes.
		1991/0810		Drop lock before erroring out when process disappears in postnote (BUG FIX).
		1991/0820		Set up->p after clearing up, so that u->p = p not needed after initial setlabel. (pseudo-BUG FIX).
		1991/0830		More debugging prints.
		1991/0906		Check p->mach to see if p is running (BUG FIX).
		1991/0919		Avoid compiler-induced bug because setlabel/gotolabel finds wrong variable. (BUG FIX)
		1991/0926		Remove Zombie, MMUing, Exiting, Inwait states.
		1991/1003		Replace manual upage deallocation with simpleputpage.
		1991/1005		Only prepare wait messages for non-kprocs.
		1991/1006		Only prepare wait messages for non-kprocs (2).
		1991/1007		Remove call to flushvirt when process breaks. (XXX why?)
		1991/1024		Close pgrp, egrp, and dot early (XXX why? just cleaner? actually needed?)
		1991/1105		Remove user name from Pgrp, add to Proc.
		1991/1108		Set Proc.psstate to "Stopped" while stopped.
		1991/1109		Add Proc.procmode. Replace "bootes" with eve.
		1991/1110		Rewrite goto in newproc as for loop. Call resrcwait instead of manual sleeping.
		1991/1112		Set Proc.procmode in kproc (XXX why?).
		1991/1115		Shuffle SET/USED around.
		1991/1120		Clean duplicate code in postnote.
		1991/1127		Silence compiler warning.
		1991/1215		Remove child accumulated times from wait message.
		1991/1216		Make p->debug a qlock.
		1991/1218		Add Proc.text and Proc.pid to non-empty exit status in pexit.
		1992/0103		Remove clearmmucache in fork. (XXX why?)
		1992/0111		Move error.h to ../port. Change errors to actual strings.
		1992/0114		Add Eprocdied error.
		1992/0115		Only save wait information if have a parent.
		1992/0120		Remove unneeded .
		1992/0122		Put process state in Proc structure instead of using stack variable in sched.
		1992/0205		Report sum of children's and process's time in wait message.
		1992/0301		The Grim Reaper lays waste the bodies of the dead.
		1992/0303		Panic if boot process dies.
		1992/0309		Use Ref for pidalloc. Keep wait message in text form.
		1992/0319		Add panic in wakeup if p->sched.pc is bad.
		1992/0321		Move lib.h to ../port.
		1992/0328		Add panic in proctab.
		1992/0428		Add Proc.noteid.
		1992/0519		Add splhi calls to sleep, tsleep around checks of r->p. (BUG FIX?)
		1992/0520		Silence compiler warnings.
		1992/0527		Succeed if process is missing in postnote.
		1992/0602		Rewrite tsleep to avoid explicit Alarm structures. Now alarm info is in Proc. New tsleep can call regular sleep, avoiding duplicate implementation.
		1992/0603		Change runproc to be always called splhi. (BUG FIX?)
		1992/0604		Fix error prototype (char[] to char*).
		1992/0609		Document procctl called splhi, add spllo before pexit, add splhi after sched. (BUG FIXes?)
		1992/0617		Add tfn to handle very quick alarms in tsleep. (BUG FIX)
		1992/0619		Formatting. Call xalloc instead of ialloc.
		1992/0620		Remove explicit wait structure list. Just use malloc.
		1992/0622		Remove busy loop in runproc.
		1992/0701		Take self out of talarm.list if checkalarm didn't. (BUG FIX).
		1992/0703		Move lastvar to last. Drop user from process text name.
		1992/0711		Remove unused variables.
		1992/0805		Add waserror/poperror around smalloc (smalloc can error??? XXX).
		1992/0902		Print exit string of dying boot process.
		1992/0909		Document wakeup requirement that only one process can call it.
		1992/1206		Code rearrangement. Call ready(p) while holding p->pgrp lock. (BUG FIX?)
		1992/1208		Only free a single broken process in freebroken. (BUG: != should be ==.)
		1992/1225		Correct test in freebroken (BUG FIX).
		1993/0112		Print when sending notes to kprocs.
		1993/0216		In postnote, only ready(p) if it is found in the rendezvous lists (BUG fix? avoids double-ready).
		1993/0309		Add qwaitr lock in pwait. (XXX why?)
		1993/0501		Rewrite sched - remove upage. Much code cleanup, converting u->p to up.
		1993/0528		Bus contention games.
		1993/0729		Call kmapinval in sched.
		1993/0731		Don't call kmapinval in sched.
		1993/0801		Call kmapinval in sched.
		1993/0804		Don't call kmapinval in sched.
		1993/0829		Formatting edits.
		1993/0830		Bus contention games.
		1993/0904		No change.
		1993/1123		Add killbig.
		1993/1201		Add per-process counters, including TLB fault counter and countext switch counter. Clear counters and syscall args in newproc.
		1993/1204		Remove Schedq definition (now in portdat.h). Add per-mach runqs.
		1993/1212		Formatting edit. Check m->runq.head in anyready.
		1993/1214		Remove TLB fault counter. Add hi, lo run queue for each mach.
		1993/1215		Remove runloq. Count number ready on each queue.
		1994/0311		Add per-mach callbacks.
		1994/0320		Move anyready up.
		1994/0321		Move anyready down.
		1994/0322		Don't initialize per-mach callbacks in schedinit.
		1994/0324		Return whether runq's are empty in newcallback (XXX why?)
		1994/0325		Don't return scheduling info in newcallback.
		1994/0406		Change process dump.
		1994/0407		Allocate p->syscall buffer for each process (nsyscall determined after build time).
		1994/0508		Remove callbacks.
		1994/0612		Remove counter, syscall arrays.
		1994/0725		Rebalance proc each time we preempt it (call balance).
		1994/0727		Use runhiq for high-priority processes.
		1994/0728		Change scheduler to five-level run queues. No longer per-mach.
		1994/0729		Faster anyready.
		1994/0805		Print debugging info about talarm loops.
		1994/0808		Always adjust priority while running (pseudo-BUG fix).
		1994/0809		Add anyready0, debugging print.
		1994/0810		Add p->readyticks for debugging print.
		1994/0812		Add Proc.rgrp. Remove talarm loop check.
		1994/0816		Set Nrq in portdat.h.
		1994/0817		Add lastreadied to preempt only when higher-priority guys are waiting.
		1994/0830		Don't print the empty run queues.
		1994/0914		Add Schedq structure.
		1994/0915		Invert priorities.
		1994/0917		Use weighted average to compute run-time priorities.
		1994/0918		Rename readyticks to readytime.
		1994/0920		Update priorities for processes sleeping a long time.
		1994/1027		Add renameuser for when eve changes.
		1995/0101		Be more careful about accessing rq->head without a lock.
		1995/0102		New scheduler: back to a single runq, procs can be hard-wired to particular machs.
		1995/0104		Various code cleanup.
		1995/0105		Remove nrdy in favor of runq.n.
		1995/0106		Return boolean in anyhigher (BUG fix? if runq.n somehow got to zero?).
		1995/0107		Go back to simpler anyhigher. Anyhigher is now any higher or same.
		1995/0108		Only clear p->yield if it is nonzero. (How can this matter? XXX)
		1995/0109		More complicated anyhigher.
		1995/0110		New scheduler: back to old multi-run-queue scheduler, with some tweaks.
		1995/0112		Update comment about locks held from pexit. Restore spl state after stopping proc in procctl.
		1995/0115		New Proc.ureg.
		1995/0126		Remove prints about sleeping or erroring splhi.
		1995/0215		Add Proc_exitbig, separate from Proc_exitme, to set exit status.
		1995/0329		Be sure to NUL-terminate wait message (BUG fix).
		1995/0811		Change m to MACHP(m->machno) to accomodate VM tricks.
		1995/0910		Add randomcount to attempt to capture randomness.
		1995/0912		Reformat randomcount.
		1995/0913		Remove randomcount.
		1995/1009		New process dump format.
		1995/1024		New process dump format.
		1995/1030		New process dump format, new scheduling dump.
		1995/1111		Lock palloc going into moribund sched
		1995/1230		Lock palloc for moribund sched.
		1996/0121		Wake up up->waitr when exiting, so that children waiting for us to notice them can give up. (BUG FIX).
		1996/0315		Yet another scheduling policy in runproc].
		1996/0426		Clear p->error in newproc.
		1996/0511		Add PriLock to avoid lock inversion.
		1996/0516		More avoidance of lock inversion.
		1996/0522		Initialize Proc.lockpri in newproc.
		1996/0523		More avoidance of lock inversion.
		1996/0626		Use p->nnote instead of up->nnote in postnote. Three year old BUG FIX!
		1996/1016		Add up->rlock for sleep/wakeup.
		1996/1017		Do not lock(r) in sleep/wakeup.
		1997/0220		Add coherence calls to force correctness on looser memory architectures (bug fix?).
		1997/0327		Formatting edits. Change close to cclose.
		1997/0821		In runproc, void lock(runq) while splhi.
		1997/1104		In pexit, nil the up->fgrp pointer before calling closefgrp, in case closefgrp manages to get back into pexit (bug fix?)
		1998/0325		Add QueueingW.
		1998/0425		Bug fix: lock p->debug while closing fgrp to avoid race with /proc/n/fd.
		1998/0508		Bug fix: in sleep, if we get interrupted between sleep1 and sched, undo the effects of the sleep1. Otherwise an inopportune interrupt during the error(Eintr) processing could call sched for us, putting us to sleep for good.
		1998/0510		Add splx(s) in sleep1. Not needed: in sleep, either notepending == 0 the first time, in which case we call sched and go spllo, or notepending == 1 the second time, in which case we call error and go spllo. Either way we're coming out spllo.
		1998/0511		In sleep, only back out and set up->state = Running if the state is currently Wakeme.
		1998/0512		Whitespace edit. Debugging: print more about panic in postnote.
		1998/0513		Merge sleep1 and sleep. (XXX Bug fixes and other changes?)
		1998/0514		Recomment sleep, wakeup, and postnote. No code changes.
		1998/0603		Tune scheduler: change movetime from HZ/2 to HZ/10.
		1998/0604		Bug fix?: remove livelock priority hack.
		1998/0605		Note that pexit is also holding the palloc lock.
		1998/0606		Lower lock priority. Fairer scheduler. Bug fix.
		1998/0724		Comment edit.
		1998/0725		Add procflushseg.
		1998/0806		Add state QueueingR.
		1998/0825		Bug fix: print format.
		1998/0901		Yet another runproc implementation.
		1998/0923		Bug fix: in the "fair" scheduling case, if we find a process to run, actually run it.
		1999/0108		Bug fix? In pexit, lock up->seglock around segment cleanup.
		1999/0110		More process dump changes.
		1999/0326		Remove lastreadied: check all run queues again. (Bug fix? If priorities changed during a clock tick, lastreadied was not updated.)
		1999/0327		Call idlehands to put the processor to sleep if there's no one to wake (Bug? in the wrong place?).
		1999/0711		Only move processes that aren't wired. Use Proc.wired instead of separate wired array.
		1999/0811		Bug fix: use MACHP(0)->ticks, not m->ticks, for scheduling decisions.
		2000/0115		Bug fix: avoid using MS2TK on huge sleep times.
		2000/0116		Comment edit.
		2000/0129		Zero out up->twhen if sleep errors out.
		2000/0325		Safety first: nil out up->egrp, up->rgrp, up->dot, and up->pgrp after closing them.
		2000/0331		Bug fix? nil out all fields under lock before closing them.
		2000/0519		Comment out kproc postnote print.
		2000/0714		Give Procalloc struct a tag.
		2000/0919		New sleep, wakeup, postnote. Locks r, so can avoid explicit coherence calls once more.
		2000/1018		Set p->newtlb to cause flush. (Bug fix?)
		2000/1026		Rename FPinit to FPINIT.
		2000/1027		Rename FPINIT to FPinit.
		2000/1106		Flush output buffer during process dump.
		2000/1130		Bug fix: move idlehands to correct location.
		2001/0207		In wakeup return the actual proc that woke up instead of a flag.
		2001/0217		Use "kproc" for psstate in kprocs.
		2001/0315		Add Proc.fixedpri to scheduler.
		2001/0510		Add hash table to find proc by pid.
		2001/0527		Make procalloc, pidalloc, and runq static.
		2001/0529		In wakeup, remove unnecessary rv variable.
		2001/0717		Print about kproc notes and ignore them.
		2001/0802		Remove Proc.nchild. Initialize Proc.parent in newproc.
		2001/0819		Wait messages are in binary again.
		2001/0822		Add Proc.syserror to hold last system call error.
		2001/0924		More efficient Proc.syserror: use two buffers and swap pointers.
		2001/1117		Add Proc.privatemem.
		2001/1218		Don't ignore postnote to kprocs anymore.
		2002/0114		Bug fix: print format.
		2002/0228		Reduce number of scheduling quanta.
		2002/0315		Move Schedq back into portdat.h.
		2002/0326		Add safe ms2tk. Bug: rounding is wrong.
		2002/0328		Add Proc.lockwait.
		2002/0329		Bug fix: correct rounding in ms2tk.
		2002/0402		Remove EDF scheduler.
		2002/0403		Delay sched if process would be preempted while holding locks.
		2002/0404		Add EDF scheduler back, along with scheduling event log debugging.
		2002/0406		Remove debugging event log.
		2002/0410		Remove ../port/edf.h. Move EDF functions to regular headers.
		2002/0413		Debugging: save error context in errlab[NERR-1]. (Not guaranteed to be valid.)
		2002/0414		Remove Proc.delaysched, but still delay scheds during locks.
		2002/0415		Select process for moving at move time, not just after.
		2002/0416		Proc.delaysched is back.
		2002/0420		Panic if sched during ilock.
		2002/0501		Store wait time in milliseconds, not ticks.
		2002/0502		Add Proc.noswap.
		2002/0506		Do not kill "kfs" process in killbig.
		2002/0626		More information about ilock sched panic.
		2002/0704		Add new Edfinterface to allow empty EDF stub routines.
		2002/0710		Remove ms2tk (now in XXX.c). Use new safe tk2ms for waitmsg times.
		2002/0821		Keep track of how long each processor spends idling.
		2002/0822		Use perfticks instead of fastticks for idle stats.
		2002/0925		More info in sched/ilock panic.
		2002/1010		Only set m->perf.period if nonzero. (Bug fix?)
		2002/1104		Add Proc.setargs.
		2003/0214		Add unfair to turn off fair scheduling (for debugging).
		2003/0226		Remove unfair scheduling.
		2003/0228		New many-run-queue scheduler.
		2003/0301		Scheduling tweak: even prioritized processes get penalized for running too long.
		2003/0314		Add yield.
		2003/0316		Add rebalance to move processes into new scheduling queues.
		2003/0319		Comment edits.
		2003/0406		Formatting edits.