svn commit: r313730 - in head/sys: kern vm

[Andriy Gapon]

Author: avg
Date: Tue Feb 14 13:54:05 2017
New Revision: 313730
URL: https://svnweb.freebsd.org/changeset/base/313730

Log:
  try to fix RACCT_RSS accounting
  
  There could be a race between the vm daemon setting RACCT_RSS based on
  the vm space and vmspace_exit (called from exit1) resetting RACCT_RSS to
  zero.  In that case we can get a zombie process with non-zero RACCT_RSS.
  If the process is jailed, that may break accounting for the jail.
  There could be other consequences.
  
  Fix this race in the vm daemon by updating RACCT_RSS only when a process
  is in the normal state.  Also, make accounting a little bit more
  accurate by refreshing the page resident count after calling
  vm_pageout_map_deactivate_pages().
  Finally, add an assert that the RSS is zero when a process is reaped.
  
  PR:		210315
  Reviewed by:	trasz
  Differential Revision: https://reviews.freebsd.org/D9464

Modified:
  head/sys/kern/kern_racct.c
  head/sys/vm/vm_pageout.c

Modified: head/sys/kern/kern_racct.c
==============================================================================
--- head/sys/kern/kern_racct.c	Tue Feb 14 13:45:38 2017	(r313729)
+++ head/sys/kern/kern_racct.c	Tue Feb 14 13:54:05 2017	(r313730)
@@ -1010,10 +1010,13 @@ racct_proc_exit(struct proc *p)
 	racct_set_locked(p, RACCT_CPU, runtime, 0);
 	racct_add_cred_locked(p->p_ucred, RACCT_PCTCPU, pct);
 
+	KASSERT(p->p_racct->r_resources[RACCT_RSS] == 0,
+	    ("process reaped with %ju allocated for RSS\n",
+	    p->p_racct->r_resources[RACCT_RSS]));
 	for (i = 0; i <= RACCT_MAX; i++) {
 		if (p->p_racct->r_resources[i] == 0)
 			continue;
-	    	if (!RACCT_IS_RECLAIMABLE(i))
+		if (!RACCT_IS_RECLAIMABLE(i))
 			continue;
 		racct_set_locked(p, i, 0, 0);
 	}

Modified: head/sys/vm/vm_pageout.c
==============================================================================
--- head/sys/vm/vm_pageout.c	Tue Feb 14 13:45:38 2017	(r313729)
+++ head/sys/vm/vm_pageout.c	Tue Feb 14 13:54:05 2017	(r313730)
@@ -2268,12 +2268,14 @@ again:
 			if (size >= limit) {
 				vm_pageout_map_deactivate_pages(
 				    &vm->vm_map, limit);
+				size = vmspace_resident_count(vm);
 			}
 #ifdef RACCT
 			if (racct_enable) {
 				rsize = IDX_TO_OFF(size);
 				PROC_LOCK(p);
-				racct_set(p, RACCT_RSS, rsize);
+				if (p->p_state == PRS_NORMAL)
+					racct_set(p, RACCT_RSS, rsize);
 				ravailable = racct_get_available(p, RACCT_RSS);
 				PROC_UNLOCK(p);
 				if (rsize > ravailable) {
@@ -2299,7 +2301,8 @@ again:
 					size = vmspace_resident_count(vm);
 					rsize = IDX_TO_OFF(size);
 					PROC_LOCK(p);
-					racct_set(p, RACCT_RSS, rsize);
+					if (p->p_state == PRS_NORMAL)
+						racct_set(p, RACCT_RSS, rsize);
 					PROC_UNLOCK(p);
 					if (rsize > ravailable)
 						tryagain = 1;