kernel: make noswap flag exclude processes from killbig() if not eve, reset noswap flag on exec

sys/man/3/proc

@@ -301,6 +301,32 @@ This bit is inherited across
 and
 .IR exec (2).
 .TP 10n
+.B nohang
+Clear the hang bit.
+.TP 10n
+.B private
+Make it impossible to read the process's user memory.
+This property is inherited on
+.IR fork (2),
+cleared on
+.IR exec (2),
+and is not otherwise resettable.
+.TP 10n
+.B noswap
+Don't allow this process to be swapped out.  This should
+be used carefully and sparingly or the system could run
+out of memory.  It is meant for processes that can't be
+swapped, like the local fileserver implementing the swap
+device and for processes containing sensitive data.
+This property is inherited on
+.IR fork (2),
+cleared on
+.IR exec (2),
+and is not otherwise resettable.
+.TP 10n
+.B kill
+Kill the process the next time it crosses the user/kernel boundary.
+.TP 10n
 .B "close\ \fIn
 Close file descriptor
 .I n
@@ -309,25 +335,6 @@ in the process.
 .B closefiles
 Close all open file descriptors in the process.
 .TP 10n
-.B nohang
-Clear the hang bit.
-.TP 10n
-.B noswap
-Don't allow this process to be swapped out.  This should
-be used carefully and sparingly or the system could run
-out of memory.  It is meant for processes that can't be
-swapped, like the ones implementing the swap device and for
-processes containing sensitive data.
-.TP 10n
-.B kill
-Kill the process the next time it crosses the user/kernel boundary.
-.TP 10n
-.B private
-Make it impossible to read the process's user memory.
-This property is inherited on fork, cleared on
-.IR exec (2),
-and is not otherwise resettable.
-.TP 10n
 .B "pri\ \fIn
 Set the base priority for the process to the integer
 .IR n .

sys/src/9/port/devproc.c

@@ -492,7 +492,7 @@ procwstat(Chan *c, uchar *db, int n)
 	if(p->pid != PID(c->qid))
 		error(Eprocdied);
 
-	if(strcmp(up->user, p->user) != 0 && strcmp(up->user, eve) != 0)
+	if(strcmp(up->user, p->user) != 0 && !iseve())
 		error(Eperm);
 
 	d = smalloc(sizeof(Dir)+n);
@@ -500,10 +500,9 @@ procwstat(Chan *c, uchar *db, int n)
 	if(n == 0)
 		error(Eshortstat);
 	if(!emptystr(d->uid) && strcmp(d->uid, p->user) != 0){
-		if(strcmp(up->user, eve) != 0)
+		if(!iseve())
 			error(Eperm);
-		else
-			kstrdup(&p->user, d->uid);
+		kstrdup(&p->user, d->uid);
 	}
 	/* p->procmode determines default mode for files in /proc */
 	if(d->mode != ~0UL)
@@ -1530,32 +1529,6 @@ procstopped(void *a)
 	return ((Proc*)a)->state == Stopped;
 }
 
-ulong
-procpagecount(Proc *p)
-{
-	Segment *s;
-	ulong pages;
-	int i;
-
-	eqlock(&p->seglock);
-	if(waserror()){
-		qunlock(&p->seglock);
-		nexterror();
-	}
-	pages = 0;
-	for(i=0; i<NSEG; i++){
-		if((s = p->seg[i]) != nil){
-			eqlock(s);
-			pages += mcountseg(s);
-			qunlock(s);
-		}
-	}
-	qunlock(&p->seglock);
-	poperror();
-
-	return pages;
-}
-
 int
 procctlmemio(Proc *p, uintptr offset, int n, void *va, int read)
 {

sys/src/9/port/error.h

@@ -37,7 +37,6 @@ extern char Econrefused[];	/* connection refused */
 extern char Econinuse[];	/* connection in use */
 extern char Eintr[];		/* interrupted */
 extern char Enomem[];		/* kernel allocate failed */
-extern char Enoswap[];		/* swap space full */
 extern char Esoverlap[];	/* segments overlap */
 extern char Emouseset[];	/* mouse type already set */
 extern char Eshort[];		/* i/o count too small */

sys/src/9/port/portdat.h

@@ -667,7 +667,7 @@ struct Proc
 	QLock	debug;		/* to access debugging elements of User */
 	Proc	*pdbg;		/* the debugging process */
 	ulong	procmode;	/* proc device default file mode */
-	ulong	privatemem;	/* proc does not let anyone read mem */
+	int	privatemem;	/* proc does not let anyone read mem */
 	int	hang;		/* hang at next exec for debug */
 	int	procctl;	/* Control for /proc debugging */
 	uintptr	pc;		/* DEBUG only */

sys/src/9/port/portfns.h

@@ -225,9 +225,10 @@ ulong		procalarm(ulong);
 void		procctl(Proc*);
 void		procdump(void);
 int		procfdprint(Chan*, int, int, char*, int);
+void		procflushseg(Segment*);
 int		procindex(ulong);
 void		procinit0(void);
-void		procflushseg(Segment*);
+ulong		procpagecount(Proc*);
 void		procpriority(Proc*, int, int);
 Proc*		proctab(int);
 extern void	(*proctrace)(Proc*, int, vlong); 

sys/src/9/port/proc.c

@@ -1491,6 +1491,32 @@ exhausted(char *resource)
 	error(buf);
 }
 
+ulong
+procpagecount(Proc *p)
+{
+	Segment *s;
+	ulong pages;
+	int i;
+
+	eqlock(&p->seglock);
+	if(waserror()){
+		qunlock(&p->seglock);
+		nexterror();
+	}
+	pages = 0;
+	for(i=0; i<NSEG; i++){
+		if((s = p->seg[i]) != nil){
+			eqlock(s);
+			pages += mcountseg(s);
+			qunlock(s);
+		}
+	}
+	qunlock(&p->seglock);
+	poperror();
+
+	return pages;
+}
+
 void
 killbig(char *why)
 {
@@ -1503,25 +1529,20 @@ killbig(char *why)
 	kp = nil;
 	ep = procalloc.arena+conf.nproc;
 	for(p = procalloc.arena; p < ep; p++) {
-		if(p->state == Dead || p->kp || !canqlock(&p->seglock))
+		if(p->state == Dead || p->kp)
 			continue;
-		l = 0;
-		for(i=1; i<NSEG; i++) {
-			s = p->seg[i];
-			if(s == nil || !canqlock(s))
-				continue;
-			l += mcountseg(s);
-			qunlock(s);
-		}
-		qunlock(&p->seglock);
-		if(l > max && ((p->procmode&0222) || strcmp(eve, p->user)!=0)) {
+		if((p->noswap || (p->procmode & 0222) == 0) && strcmp(eve, p->user) == 0)
+			continue;
+		l = procpagecount(p);
+		if(l > max){
 			kp = p;
 			max = l;
 		}
 	}
-	if(kp == nil || !canqlock(&kp->seglock))
+	if(kp == nil)
 		return;
 	print("%lud: %s killed: %s\n", kp->pid, kp->text, why);
+	qlock(&kp->seglock);
 	for(p = procalloc.arena; p < ep; p++) {
 		if(p->state == Dead || p->kp)
 			continue;
@@ -1531,7 +1552,8 @@ killbig(char *why)
 	kp->procctl = Proc_exitbig;
 	for(i = 0; i < NSEG; i++) {
 		s = kp->seg[i];
-		if(s != nil && canqlock(s)) {
+		if(s != nil) {
+			qlock(s);
 			mfreeseg(s, s->base, (s->top - s->base)/BY2PG);
 			qunlock(s);
 		}

sys/src/9/port/sysproc.c

@@ -512,6 +512,7 @@ sysexec(va_list list)
 	up->notify = 0;
 	up->notified = 0;
 	up->privatemem = 0;
+	up->noswap = 0;
 	procsetup(up);
 	qunlock(&up->debug);