Add kern.racct.enable tunable and RACCT_DISABLED config option.

The point of this is to be able to add RACCT (with RACCT_DISABLED)
to GENERIC, to avoid having to rebuild the kernel to use rctl(8).

Differential Revision:	https://reviews.freebsd.org/D2369
Reviewed by:	kib@
MFC after:	1 month
Relnotes:	yes
Sponsored by:	The FreeBSD Foundation

5 files changed, 137 insertions, 102 deletions

diff --git a/sys/vm/swap_pager.c b/sys/vm/swap_pager.c
index 55e02c4..15e5b24 100644
--- a/sys/vm/swap_pager.c
+++ b/sys/vm/swap_pager.c
@@ -198,11 +198,13 @@ swap_reserve_by_cred(vm_ooffset_t incr, struct ucred *cred)
 		panic("swap_reserve: & PAGE_MASK");
 
 #ifdef RACCT
-	PROC_LOCK(curproc);
-	error = racct_add(curproc, RACCT_SWAP, incr);
-	PROC_UNLOCK(curproc);
-	if (error != 0)
-		return (0);
+	if (racct_enable) {
+		PROC_LOCK(curproc);
+		error = racct_add(curproc, RACCT_SWAP, incr);
+		PROC_UNLOCK(curproc);
+		if (error != 0)
+			return (0);
+	}
 #endif
 
 	res = 0;
diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
index b7e668b..bad9994 100644
--- a/sys/vm/vm_map.c
+++ b/sys/vm/vm_map.c
@@ -300,11 +300,11 @@ vmspace_alloc(vm_offset_t min, vm_offset_t max, pmap_pinit_t pinit)
 	return (vm);
 }
 
+#ifdef RACCT
 static void
 vmspace_container_reset(struct proc *p)
 {
 
-#ifdef RACCT
 	PROC_LOCK(p);
 	racct_set(p, RACCT_DATA, 0);
 	racct_set(p, RACCT_STACK, 0);
@@ -312,8 +312,8 @@ vmspace_container_reset(struct proc *p)
 	racct_set(p, RACCT_MEMLOCK, 0);
 	racct_set(p, RACCT_VMEM, 0);
 	PROC_UNLOCK(p);
-#endif
 }
+#endif
 
 static inline void
 vmspace_dofree(struct vmspace *vm)
@@ -415,7 +415,10 @@ vmspace_exit(struct thread *td)
 		pmap_activate(td);
 		vmspace_dofree(vm);
 	}
-	vmspace_container_reset(p);
+#ifdef RACCT
+	if (racct_enable)
+		vmspace_container_reset(p);
+#endif
 }
 
 /* Acquire reference to vmspace owned by another process. */
@@ -3652,14 +3655,16 @@ Retry:
 		return (KERN_NO_SPACE);
 	}
 #ifdef RACCT
-	PROC_LOCK(p);
-	if (is_procstack &&
-	    racct_set(p, RACCT_STACK, ctob(vm->vm_ssize) + grow_amount)) {
+	if (racct_enable) {
+		PROC_LOCK(p);
+		if (is_procstack && racct_set(p, RACCT_STACK,
+		    ctob(vm->vm_ssize) + grow_amount)) {
+			PROC_UNLOCK(p);
+			vm_map_unlock_read(map);
+			return (KERN_NO_SPACE);
+		}
 		PROC_UNLOCK(p);
-		vm_map_unlock_read(map);
-		return (KERN_NO_SPACE);
 	}
-	PROC_UNLOCK(p);
 #endif
 
 	/* Round up the grow amount modulo sgrowsiz */
@@ -3685,15 +3690,17 @@ Retry:
 			goto out;
 		}
 #ifdef RACCT
-		PROC_LOCK(p);
-		if (racct_set(p, RACCT_MEMLOCK,
-		    ptoa(pmap_wired_count(map->pmap)) + grow_amount)) {
+		if (racct_enable) {
+			PROC_LOCK(p);
+			if (racct_set(p, RACCT_MEMLOCK,
+			    ptoa(pmap_wired_count(map->pmap)) + grow_amount)) {
+				PROC_UNLOCK(p);
+				vm_map_unlock_read(map);
+				rv = KERN_NO_SPACE;
+				goto out;
+			}
 			PROC_UNLOCK(p);
-			vm_map_unlock_read(map);
-			rv = KERN_NO_SPACE;
-			goto out;
 		}
-		PROC_UNLOCK(p);
 #endif
 	}
 	/* If we would blow our VMEM resource limit, no go */
@@ -3703,14 +3710,16 @@ Retry:
 		goto out;
 	}
 #ifdef RACCT
-	PROC_LOCK(p);
-	if (racct_set(p, RACCT_VMEM, map->size + grow_amount)) {
+	if (racct_enable) {
+		PROC_LOCK(p);
+		if (racct_set(p, RACCT_VMEM, map->size + grow_amount)) {
+			PROC_UNLOCK(p);
+			vm_map_unlock_read(map);
+			rv = KERN_NO_SPACE;
+			goto out;
+		}
 		PROC_UNLOCK(p);
-		vm_map_unlock_read(map);
-		rv = KERN_NO_SPACE;
-		goto out;
 	}
-	PROC_UNLOCK(p);
 #endif
 
 	if (vm_map_lock_upgrade(map))
@@ -3809,7 +3818,7 @@ Retry:
 
 out:
 #ifdef RACCT
-	if (rv != KERN_SUCCESS) {
+	if (racct_enable && rv != KERN_SUCCESS) {
 		PROC_LOCK(p);
 		error = racct_set(p, RACCT_VMEM, map->size);
 		KASSERT(error == 0, ("decreasing RACCT_VMEM failed"));
diff --git a/sys/vm/vm_mmap.c b/sys/vm/vm_mmap.c
index 02634d6..1dd2479 100644
--- a/sys/vm/vm_mmap.c
+++ b/sys/vm/vm_mmap.c
@@ -1122,16 +1122,18 @@ vm_mlock(struct proc *proc, struct ucred *cred, const void *addr0, size_t len)
 	if (npages + vm_cnt.v_wire_count > vm_page_max_wired)
 		return (EAGAIN);
 #ifdef RACCT
-	PROC_LOCK(proc);
-	error = racct_set(proc, RACCT_MEMLOCK, nsize);
-	PROC_UNLOCK(proc);
-	if (error != 0)
-		return (ENOMEM);
+	if (racct_enable) {
+		PROC_LOCK(proc);
+		error = racct_set(proc, RACCT_MEMLOCK, nsize);
+		PROC_UNLOCK(proc);
+		if (error != 0)
+			return (ENOMEM);
+	}
 #endif
 	error = vm_map_wire(map, start, end,
 	    VM_MAP_WIRE_USER | VM_MAP_WIRE_NOHOLES);
 #ifdef RACCT
-	if (error != KERN_SUCCESS) {
+	if (racct_enable && error != KERN_SUCCESS) {
 		PROC_LOCK(proc);
 		racct_set(proc, RACCT_MEMLOCK,
 		    ptoa(pmap_wired_count(map->pmap)));
@@ -1179,11 +1181,13 @@ sys_mlockall(td, uap)
 		PROC_UNLOCK(td->td_proc);
 	}
 #ifdef RACCT
-	PROC_LOCK(td->td_proc);
-	error = racct_set(td->td_proc, RACCT_MEMLOCK, map->size);
-	PROC_UNLOCK(td->td_proc);
-	if (error != 0)
-		return (ENOMEM);
+	if (racct_enable) {
+		PROC_LOCK(td->td_proc);
+		error = racct_set(td->td_proc, RACCT_MEMLOCK, map->size);
+		PROC_UNLOCK(td->td_proc);
+		if (error != 0)
+			return (ENOMEM);
+	}
 #endif
 
 	if (uap->how & MCL_FUTURE) {
@@ -1205,7 +1209,7 @@ sys_mlockall(td, uap)
 		error = (error == KERN_SUCCESS ? 0 : EAGAIN);
 	}
 #ifdef RACCT
-	if (error != KERN_SUCCESS) {
+	if (racct_enable && error != KERN_SUCCESS) {
 		PROC_LOCK(td->td_proc);
 		racct_set(td->td_proc, RACCT_MEMLOCK,
 		    ptoa(pmap_wired_count(map->pmap)));
@@ -1247,7 +1251,7 @@ sys_munlockall(td, uap)
 	error = vm_map_unwire(map, vm_map_min(map), vm_map_max(map),
 	    VM_MAP_WIRE_USER|VM_MAP_WIRE_HOLESOK);
 #ifdef RACCT
-	if (error == KERN_SUCCESS) {
+	if (racct_enable && error == KERN_SUCCESS) {
 		PROC_LOCK(td->td_proc);
 		racct_set(td->td_proc, RACCT_MEMLOCK, 0);
 		PROC_UNLOCK(td->td_proc);
@@ -1291,7 +1295,7 @@ sys_munlock(td, uap)
 	error = vm_map_unwire(&td->td_proc->p_vmspace->vm_map, start, end,
 	    VM_MAP_WIRE_USER | VM_MAP_WIRE_NOHOLES);
 #ifdef RACCT
-	if (error == KERN_SUCCESS) {
+	if (racct_enable && error == KERN_SUCCESS) {
 		PROC_LOCK(td->td_proc);
 		map = &td->td_proc->p_vmspace->vm_map;
 		racct_set(td->td_proc, RACCT_MEMLOCK,
diff --git a/sys/vm/vm_pageout.c b/sys/vm/vm_pageout.c
index 6f50053..872ffd8 100644
--- a/sys/vm/vm_pageout.c
+++ b/sys/vm/vm_pageout.c
@@ -1787,11 +1787,13 @@ vm_daemon(void)
 
 	while (TRUE) {
 		mtx_lock(&vm_daemon_mtx);
+		msleep(&vm_daemon_needed, &vm_daemon_mtx, PPAUSE, "psleep",
 #ifdef RACCT
-		msleep(&vm_daemon_needed, &vm_daemon_mtx, PPAUSE, "psleep", hz);
+		    racct_enable ? hz : 0
 #else
-		msleep(&vm_daemon_needed, &vm_daemon_mtx, PPAUSE, "psleep", 0);
+		    0
 #endif
+		);
 		swapout_flags = vm_pageout_req_swapout;
 		vm_pageout_req_swapout = 0;
 		mtx_unlock(&vm_daemon_mtx);
@@ -1866,33 +1868,40 @@ again:
 				    &vm->vm_map, limit);
 			}
 #ifdef RACCT
-			rsize = IDX_TO_OFF(size);
-			PROC_LOCK(p);
-			racct_set(p, RACCT_RSS, rsize);
-			ravailable = racct_get_available(p, RACCT_RSS);
-			PROC_UNLOCK(p);
-			if (rsize > ravailable) {
-				/*
-				 * Don't be overly aggressive; this might be
-				 * an innocent process, and the limit could've
-				 * been exceeded by some memory hog.  Don't
-				 * try to deactivate more than 1/4th of process'
-				 * resident set size.
-				 */
-				if (attempts <= 8) {
-					if (ravailable < rsize - (rsize / 4))
-						ravailable = rsize - (rsize / 4);
-				}
-				vm_pageout_map_deactivate_pages(
-				    &vm->vm_map, OFF_TO_IDX(ravailable));
-				/* Update RSS usage after paging out. */
-				size = vmspace_resident_count(vm);
+			if (racct_enable) {
 				rsize = IDX_TO_OFF(size);
 				PROC_LOCK(p);
 				racct_set(p, RACCT_RSS, rsize);
+				ravailable = racct_get_available(p, RACCT_RSS);
 				PROC_UNLOCK(p);
-				if (rsize > ravailable)
-					tryagain = 1;
+				if (rsize > ravailable) {
+					/*
+					 * Don't be overly aggressive; this
+					 * might be an innocent process,
+					 * and the limit could've been exceeded
+					 * by some memory hog.  Don't try
+					 * to deactivate more than 1/4th
+					 * of process' resident set size.
+					 */
+					if (attempts <= 8) {
+						if (ravailable < rsize -
+						    (rsize / 4)) {
+							ravailable = rsize -
+							    (rsize / 4);
+						}
+					}
+					vm_pageout_map_deactivate_pages(
+					    &vm->vm_map,
+					    OFF_TO_IDX(ravailable));
+					/* Update RSS usage after paging out. */
+					size = vmspace_resident_count(vm);
+					rsize = IDX_TO_OFF(size);
+					PROC_LOCK(p);
+					racct_set(p, RACCT_RSS, rsize);
+					PROC_UNLOCK(p);
+					if (rsize > ravailable)
+						tryagain = 1;
+				}
 			}
 #endif
 			vmspace_free(vm);
diff --git a/sys/vm/vm_unix.c b/sys/vm/vm_unix.c
index de9aa78..cc77ff7 100644
--- a/sys/vm/vm_unix.c
+++ b/sys/vm/vm_unix.c
@@ -130,35 +130,39 @@ sys_obreak(td, uap)
 			goto done;
 		}
 #ifdef RACCT
-		PROC_LOCK(td->td_proc);
-		error = racct_set(td->td_proc, RACCT_DATA, new - base);
-		if (error != 0) {
-			PROC_UNLOCK(td->td_proc);
-			error = ENOMEM;
-			goto done;
-		}
-		error = racct_set(td->td_proc, RACCT_VMEM,
-		    map->size + (new - old));
-		if (error != 0) {
-			racct_set_force(td->td_proc, RACCT_DATA, old - base);
-			PROC_UNLOCK(td->td_proc);
-			error = ENOMEM;
-			goto done;
-		}
-		if (!old_mlock && map->flags & MAP_WIREFUTURE) {
-			error = racct_set(td->td_proc, RACCT_MEMLOCK,
-			    ptoa(pmap_wired_count(map->pmap)) + (new - old));
+		if (racct_enable) {
+			PROC_LOCK(td->td_proc);
+			error = racct_set(td->td_proc, RACCT_DATA, new - base);
+			if (error != 0) {
+				PROC_UNLOCK(td->td_proc);
+				error = ENOMEM;
+				goto done;
+			}
+			error = racct_set(td->td_proc, RACCT_VMEM,
+			    map->size + (new - old));
 			if (error != 0) {
 				racct_set_force(td->td_proc, RACCT_DATA,
 				    old - base);
-				racct_set_force(td->td_proc, RACCT_VMEM,
-				    map->size);
 				PROC_UNLOCK(td->td_proc);
 				error = ENOMEM;
 				goto done;
 			}
+			if (!old_mlock && map->flags & MAP_WIREFUTURE) {
+				error = racct_set(td->td_proc, RACCT_MEMLOCK,
+				    ptoa(pmap_wired_count(map->pmap)) +
+				    (new - old));
+				if (error != 0) {
+					racct_set_force(td->td_proc, RACCT_DATA,
+					    old - base);
+					racct_set_force(td->td_proc, RACCT_VMEM,
+					    map->size);
+					PROC_UNLOCK(td->td_proc);
+					error = ENOMEM;
+					goto done;
+				}
+			}
+			PROC_UNLOCK(td->td_proc);
 		}
-		PROC_UNLOCK(td->td_proc);
 #endif
 		prot = VM_PROT_RW;
 #ifdef COMPAT_FREEBSD32
@@ -170,14 +174,19 @@ sys_obreak(td, uap)
 		rv = vm_map_insert(map, NULL, 0, old, new, prot, VM_PROT_ALL, 0);
 		if (rv != KERN_SUCCESS) {
 #ifdef RACCT
-			PROC_LOCK(td->td_proc);
-			racct_set_force(td->td_proc, RACCT_DATA, old - base);
-			racct_set_force(td->td_proc, RACCT_VMEM, map->size);
-			if (!old_mlock && map->flags & MAP_WIREFUTURE) {
-				racct_set_force(td->td_proc, RACCT_MEMLOCK,
-				    ptoa(pmap_wired_count(map->pmap)));
+			if (racct_enable) {
+				PROC_LOCK(td->td_proc);
+				racct_set_force(td->td_proc,
+				    RACCT_DATA, old - base);
+				racct_set_force(td->td_proc,
+				    RACCT_VMEM, map->size);
+				if (!old_mlock && map->flags & MAP_WIREFUTURE) {
+					racct_set_force(td->td_proc,
+					    RACCT_MEMLOCK,
+					    ptoa(pmap_wired_count(map->pmap)));
+				}
+				PROC_UNLOCK(td->td_proc);
 			}
-			PROC_UNLOCK(td->td_proc);
 #endif
 			error = ENOMEM;
 			goto done;
@@ -205,14 +214,16 @@ sys_obreak(td, uap)
 		}
 		vm->vm_dsize -= btoc(old - new);
 #ifdef RACCT
-		PROC_LOCK(td->td_proc);
-		racct_set_force(td->td_proc, RACCT_DATA, new - base);
-		racct_set_force(td->td_proc, RACCT_VMEM, map->size);
-		if (!old_mlock && map->flags & MAP_WIREFUTURE) {
-			racct_set_force(td->td_proc, RACCT_MEMLOCK,
-			    ptoa(pmap_wired_count(map->pmap)));
+		if (racct_enable) {
+			PROC_LOCK(td->td_proc);
+			racct_set_force(td->td_proc, RACCT_DATA, new - base);
+			racct_set_force(td->td_proc, RACCT_VMEM, map->size);
+			if (!old_mlock && map->flags & MAP_WIREFUTURE) {
+				racct_set_force(td->td_proc, RACCT_MEMLOCK,
+				    ptoa(pmap_wired_count(map->pmap)));
+			}
+			PROC_UNLOCK(td->td_proc);
 		}
-		PROC_UNLOCK(td->td_proc);
 #endif
 	}
 done: