kernel: generalize awake

Awake can now interrupt several blocking syscalls (even
during note handling).

Among others, it can interrupt await, pread and pwrite.

It cannot interrupt several others for different reasons:

- awake cannot be interrupted by awake;
- syscalls like remove and create can be used for kernel comunication
  and it would be hard to know if the effect occurred in the
  receiving fs if they were interrupted;
- other syscalls do not need awake since they just provide access
  to kernel infos (eg seek or fd2path)

NOTE: awakes registered before a note cannot occur during the note
handling and will be deferred till the next call to noted.

qa/kern/nested_note.c

@@ -27,15 +27,15 @@ handler(void *v, char *s)
 	int i;
 	if(strcmp(s, "stop") == 0){
 		done = 1;
-		noted(NCONT);
+	}else{
+		print("waiting after %s", s);
+		for(i = 0; i < 1000*1000; ++i)
+			if(i % 4999 == 0)
+				print(".");
+		print("\n");
+		print("wait after %s terminated\n", s);
+		waited++;
 	}
-	print("waiting after %s", s);
-	for(i = 0; i < 1000*1000; ++i)
-		if(i % 4999 == 0)
-			print(".");
-	print("\n");
-	print("wait after %s terminated\n", s);
-	waited++;
 	noted(NCONT);
 }
 
@@ -60,6 +60,8 @@ main(int argc, char**argv)
 		sleep(100);
 	if(waited == 2){
 		print("PASS\n");
-		exits(0);
+		exits("PASS");
 	}
+	print("%d notes received\n");
+	exits("FAIL");
 }

qa/kern/nested_note.runner

@@ -21,8 +21,8 @@ echo -n stop > /proc/$testpid/note
 
 wait $testpid
 
-if ( cat $test_output | grep 'waiting after first.........................................................................................................................................................................................................' > /dev/null ) {
+if ( cat $test_output | grep 'waiting after first.......................' > /dev/null ) {
-if ( cat $test_output | grep 'waiting after second.........................................................................................................................................................................................................' > /dev/null ) {
+if ( cat $test_output | grep 'waiting after second.......................' > /dev/null ) {
 if ( cat $test_output | grep 'PASS' > /dev/null ) {
 	rm $test_output
 	echo PASS

sys/src/kern/amd64/syscall.c

@@ -237,6 +237,7 @@ syscall(Syscalls scallnr, Ureg* ureg)
 
 	m->syscall++;
 	up->insyscall = 1;
+	up->cursyscall = (Syscalls)scallnr;
 	up->pc = ureg->ip;
 	up->dbgreg = ureg;
 	if(up->trace && (pt = proctrace) != nil)
@@ -345,6 +346,7 @@ syscall(Syscalls scallnr, Ureg* ureg)
 
 	up->insyscall = 0;
 	up->psstate = 0;
+	up->cursyscall = 0;
 
 	if(scallnr == SysNoted)
 		noted(ureg, ureg->di);
@@ -352,6 +354,8 @@ syscall(Syscalls scallnr, Ureg* ureg)
 	splhi();
 	if(scallnr != SysRfork && (up->procctl || up->nnote))
 		notify(ureg);
+	else if(canwakeup(scallnr))
+		awokeproc(up);
 
 	/* if we delayed sched because we held a lock, sched now */
 	if(up->delaysched){

sys/src/kern/port/awake.c

@@ -1,7 +1,7 @@
 /*
  * This file is part of Jehanne.
  *
- * Copyright (C) 2015-2016 Giacomo Tesio <giacomo@tesio.it>
+ * Copyright (C) 2015-2017 Giacomo Tesio <giacomo@tesio.it>
  *
  * Jehanne is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -24,6 +24,7 @@
 typedef struct PendingWakeup PendingWakeup;
 struct PendingWakeup
 {
+	short notified;
 	uint64_t time;
 	Proc *p;
 	PendingWakeup *next;
@@ -39,11 +40,53 @@ static QLock l;
 
 static Rendez	awaker;
 
+int
+canwakeup(Syscalls scall)
+{
+	if(scall == 0){
+		/* fault set insyscall to 1 */
+		return 0;
+	}
+	switch(scall){
+	default:
+		panic("canwakeup: unknown scall %d\n", scall);
+	case SysFstat:
+	case SysFwstat:
+	case SysOpen:
+	case SysPread:
+	case SysPwrite:
+	case SysRendezvous:
+	case SysSemacquire:
+	case SysSemrelease:
+	case SysAwait:
+		return 1;
+	case SysAwake:
+	case SysBind:
+	case SysClose:
+	case SysCreate:
+	case SysErrstr:
+	case SysExec:
+	case Sys_exits:
+	case SysFauth:
+	case SysFd2path:
+	case SysFversion:
+	case SysMount:
+	case SysNoted:
+	case SysNotify:
+	case SysRemove:
+	case SysRfork:
+	case SysSeek:
+	case SysUnmount:
+	case SysAlarm:
+		return 0;
+	}
+}
+
 /*
  * Actually wakeup a process
  */
 static void
-wakeupProc(Proc *p)
+wakeupProc(Proc *p, unsigned long t)
 {
 	Mpl pl;
 	Rendez *r;
@@ -78,6 +121,7 @@ wakeupProc(Proc *p)
 	unlock(&p->rlock);
 	splx(pl);
 
+	p->pendingWakeup = t;
 	if(p->state != Rendezvous)
 		return;
 
@@ -102,28 +146,50 @@ void
 awakekproc(void* v)
 {
 	Proc *p;
-	PendingWakeup *toAwake, *tail;
+	PendingWakeup *tail, *toAwake, **toAwakeEnd, *toDefer, **toDeferEnd;
 	uint64_t now;
 
 	for(;;){
 		now = sys->ticks;
 		toAwake = nil;
+		toAwakeEnd = &toAwake;
+		toDefer = nil;
+		toDeferEnd = &toDefer;
 
 		/* search for processes to wakeup */
 		qlock(&l);
 		tail = alarms;
 		while(tail && tail->time <= now){
-			if(tail->p->pendingWakeup > tail->p->lastWakeup && tail->p->state >= Ready)
+			if(tail->p->pendingWakeup > tail->p->lastWakeup
-				break;
+			&& tail->p->state >= Ready){
-			toAwake = tail;
+				*toDeferEnd = tail;
-			--toAwake->p->wakeups;
+				toDeferEnd = &tail->next;
-			toAwake->p->pendingWakeup = toAwake->time;
+			} else if (!tail->notified && tail->p->notified){
+				/* If an awake was requested outside of
+				 * a note handler, it cannot expire
+				 * while executing a note handler.
+				 * On the other hand if an awake was
+				 * requeted while executing a note handler,
+				 * it's up to the note handler to
+				 * forgive it if it's not needed anymore.
+				 */
+				*toDeferEnd = tail;
+				toDeferEnd = &tail->next;
+			} else {
+				*toAwakeEnd = tail;
+				toAwakeEnd = &tail->next;
+				--tail->p->wakeups;
+			}
 			tail = tail->next;
 		}
 		if(toAwake != nil){
-			toAwake->next = nil;
+			*toAwakeEnd = nil;
-			toAwake = alarms;
+			if(toDefer != nil){
-			alarms = tail;
+				*toDeferEnd = tail;
+				alarms = toDefer;
+			} else {
+				alarms = tail;
+			}
 		}
 		qunlock(&l);
 
@@ -136,7 +202,7 @@ awakekproc(void* v)
 				 */
 				if(canqlock(&p->debug)){
 					if(!waserror()){
-						wakeupProc(p);
+						wakeupProc(p, toAwake->time);
 						poperror();
 					}
 					qunlock(&p->debug);
@@ -225,6 +291,7 @@ wakeupafter(int64_t ms)
 	if(new == nil)
 		return 0;
 	new->p = up;
+	new->notified = up->notified;
 	new->time = when;
 
 	qlock(&l);

sys/src/kern/port/portdat.h

@@ -609,6 +609,7 @@ struct Proc
 	Profbuf*	prof;
 
 	int		insyscall;
+	Syscalls	cursyscall;
 	int32_t		blockingfd;	/* fd currenly read/written */
 
 	QLock		debug;		/* to access debugging elements of User */

sys/src/kern/port/portfns.h

@@ -48,6 +48,7 @@ char*		chanpath(Chan*);
 int		canlock(Lock*);
 int		canqlock(QLock*);
 int		canrlock(RWlock*);
+int		canwakeup(Syscalls scall);
 Chan*		cclone(Chan*);
 void		cclose(Chan*);
 void		ccloseq(Chan*);

sys/src/kern/port/proc.c

@@ -568,6 +568,8 @@ newproc(void)
 	p = psalloc();
 
 	p->state = Scheding;
+	p->insyscall = 0;
+	p->cursyscall = 0;
 	p->psstate = "New";
 	p->mach = 0;
 	p->qnext = 0;
@@ -710,7 +712,9 @@ sleep(Rendez *r, int (*f)(void*), void *arg)
 	 */
 	r->p = up;
 
-	if((*f)(arg) || up->notepending && !up->notedeferred){
+	if((*f)(arg)
+	|| (up->notepending && !up->notedeferred)
+	|| (up->insyscall && awakeOnBlock(up) && canwakeup(up->cursyscall))){
 		/*
 		 *  if condition happened or a note is pending
 		 *  never mind
@@ -751,11 +755,14 @@ sleep(Rendez *r, int (*f)(void*), void *arg)
 
 	if(up->notepending && !up->notedeferred) {
 		up->notepending = 0;
+SleepAwakened:
 		splx(s);
 		if(up->procctl == Proc_exitme && up->closingfgrp)
 			forceclosefgrp();
 		error(Eintr);
 	}
+	if(up->insyscall && awakeOnBlock(up) && canwakeup(up->cursyscall))
+		goto SleepAwakened;
 
 	splx(s);
 }

sys/src/kern/port/sysproc.c

@@ -1,7 +1,7 @@
 /*
  * This file is part of Jehanne.
  *
- * Copyright (C) 2015-2016 Giacomo Tesio <giacomo@tesio.it>
+ * Copyright (C) 2015-2017 Giacomo Tesio <giacomo@tesio.it>
  *
  * Jehanne is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
@@ -858,7 +858,6 @@ sysrendezvous(void* tagp, void* rendvalp)
 
 	result = UINT2PTR(up->rendval);
 rendezvousDone:
-	awokeproc(up);
 	return result;
 }
 
@@ -1061,50 +1060,6 @@ semacquire(ProcSegment* s, int* addr, int block)
 	return 1;
 }
 
-/* Acquire semaphore or time-out */
-static int
-tsemacquire(ProcSegment* s, int* addr, uint64_t ms)
-{
-	int acquired, timedout;
-	uint64_t t, elms;
-	Sema phore;
-
-	if(canacquire(addr))
-		return 1;
-	if(ms == 0)
-		return 0;
-	acquired = timedout = 0;
-	semqueue(s, addr, &phore);
-	for(;;){
-		phore.waiting = 1;
-		coherence();
-		if(canacquire(addr)){
-			acquired = 1;
-			break;
-		}
-		if(waserror())
-			break;
-		t = sys->ticks;
-		tsleep(&phore.rend, semawoke, &phore, ms);
-		elms = TK2MS(sys->ticks - t);
-		poperror();
-		if(elms >= ms){
-			timedout = 1;
-			break;
-		}
-		ms -= elms;
-	}
-	semdequeue(s, &phore);
-	coherence();	/* not strictly necessary due to lock in semdequeue */
-	if(!phore.waiting)
-		semwakeup(s, addr, 1);
-	if(timedout)
-		return 0;
-	if(!acquired)
-		nexterror();
-	return 1;
-}
-
 int
 syssemacquire(int* addr, int block)
 {
@@ -1124,26 +1079,6 @@ syssemacquire(int* addr, int block)
 	return semacquire(s, addr, block);
 }
 
-int
-systsemacquire(int* addr, uint64_t ms)
-{
-	ProcSegment *s;
-
-	addr = validaddr(addr, sizeof(int), 1);
-
-	evenaddr(PTR2UINT(addr));
-
-	s = proc_segment(up, PTR2UINT(addr));
-	if(s == nil || (s->permissions&SgWrite) == 0 || (uintptr_t)addr+sizeof(int) > s->top){
-		validaddr(addr, sizeof(int), 1);
-		error(Ebadarg);
-	}
-	if(*addr < 0)
-		error(Ebadarg);
-
-	return tsemacquire(s, addr, ms);
-}
-
 int
 syssemrelease(int* addr, int delta)
 {

sys/src/sysconf.json

@@ -25,7 +25,7 @@
 				"char*",
 				"int"
 			],
-			"Id": 0,
+			"Id": 1,
 			"Name": "await",
 			"Ret": [
 				"int"
@@ -35,7 +35,7 @@
 			"Args": [
 				"long"
 			],
-			"Id": 1,
+			"Id": 2,
 			"Name": "awake",
 			"Ret": [
 				"long"
@@ -47,7 +47,7 @@
 				"const char*",
 				"int"
 			],
-			"Id": 2,
+			"Id": 3,
 			"Name": "bind",
 			"Ret": [
 				"int"
@@ -57,7 +57,7 @@
 			"Args": [
 				"int"
 			],
-			"Id": 3,
+			"Id": 4,
 			"Name": "close",
 			"Ret": [
 				"long"
@@ -69,7 +69,7 @@
 				"long",
 				"long"
 			],
-			"Id": 4,
+			"Id": 5,
 			"Name": "create",
 			"Ret": [
 				"long"
@@ -80,7 +80,7 @@
 				"char*",
 				"int"
 			],
-			"Id": 5,
+			"Id": 6,
 			"Name": "errstr",
 			"Ret": [
 				"int"
@@ -91,7 +91,7 @@
 				"const char*",
 				"const char**"
 			],
-			"Id": 6,
+			"Id": 7,
 			"Name": "exec",
 			"Ret": [
 				"uintptr_t"
@@ -101,7 +101,7 @@
 			"Args": [
 				"const char*"
 			],
-			"Id": 7,
+			"Id": 8,
 			"Name": "_exits",
 			"Ret": [
 				"int"
@@ -112,7 +112,7 @@
 				"int",
 				"const char*"
 			],
-			"Id": 8,
+			"Id": 9,
 			"Name": "fauth",
 			"Ret": [
 				"int"
@@ -124,7 +124,7 @@
 				"const char*",
 				"uint32_t"
 			],
-			"Id": 9,
+			"Id": 10,
 			"Name": "fd2path",
 			"Ret": [
 				"int32_t"
@@ -136,7 +136,7 @@
 				"uint8_t*",
 				"int"
 			],
-			"Id": 10,
+			"Id": 11,
 			"Name": "fstat",
 			"Ret": [
 				"long"
@@ -149,7 +149,7 @@
 				"const char*",
 				"int"
 			],
-			"Id": 11,
+			"Id": 12,
 			"Name": "fversion",
 			"Ret": [
 				"int"
@@ -161,7 +161,7 @@
 				"const uint8_t*",
 				"uint32_t"
 			],
-			"Id": 12,
+			"Id": 13,
 			"Name": "fwstat",
 			"Ret": [
 				"long"
@@ -176,7 +176,7 @@
 				"const char*",
 				"int"
 			],
-			"Id": 13,
+			"Id": 14,
 			"Name": "mount",
 			"Ret": [
 				"int"
@@ -186,7 +186,7 @@
 			"Args": [
 				"int"
 			],
-			"Id": 14,
+			"Id": 15,
 			"Name": "noted",
 			"Ret": [
 				"int"
@@ -196,7 +196,7 @@
 			"Args": [
 				"const void*"
 			],
-			"Id": 15,
+			"Id": 16,
 			"Name": "notify",
 			"Ret": [
 				"int"
@@ -207,7 +207,7 @@
 				"const char*",
 				"uint32_t"
 			],
-			"Id": 16,
+			"Id": 17,
 			"Name": "open",
 			"Ret": [
 				"long"
@@ -220,7 +220,7 @@
 				"long",
 				"long"
 			],
-			"Id": 17,
+			"Id": 18,
 			"Name": "pread",
 			"Ret": [
 				"long"
@@ -233,7 +233,7 @@
 				"long",
 				"long"
 			],
-			"Id": 18,
+			"Id": 19,
 			"Name": "pwrite",
 			"Ret": [
 				"long"
@@ -243,7 +243,7 @@
 			"Args": [
 				"const char*"
 			],
-			"Id": 19,
+			"Id": 20,
 			"Name": "remove",
 			"Ret": [
 				"long"
@@ -254,7 +254,7 @@
 				"const void*",
 				"void*"
 			],
-			"Id": 20,
+			"Id": 21,
 			"Name": "rendezvous",
 			"Ret": [
 				"void*"
@@ -264,7 +264,7 @@
 			"Args": [
 				"uint32_t"
 			],
-			"Id": 21,
+			"Id": 22,
 			"Name": "rfork",
 			"Ret": [
 				"int"
@@ -276,7 +276,7 @@
 				"long",
 				"int"
 			],
-			"Id": 22,
+			"Id": 23,
 			"Name": "seek",
 			"Ret": [
 				"long"
@@ -287,7 +287,7 @@
 				"int*",
 				"int"
 			],
-			"Id": 23,
+			"Id": 24,
 			"Name": "semacquire",
 			"Ret": [
 				"int"
@@ -298,7 +298,7 @@
 				"int*",
 				"int"
 			],
-			"Id": 24,
+			"Id": 25,
 			"Name": "semrelease",
 			"Ret": [
 				"int"