diff --git a/headers/private/kernel/vm/VMCache.h b/headers/private/kernel/vm/VMCache.h
index 8bfc57e323..9101ff75b9 100644
--- a/headers/private/kernel/vm/VMCache.h
+++ b/headers/private/kernel/vm/VMCache.h
@@ -135,7 +135,7 @@ public:
 	virtual	status_t			Adopt(VMCache* source, off_t offset, off_t size,
 									off_t newOffset);
 
-	virtual	status_t			Discard(off_t offset, off_t size);
+	virtual	ssize_t				Discard(off_t offset, off_t size);
 
 			status_t			FlushAndRemoveAllPages();
 
@@ -220,7 +220,7 @@ private:
 			void				_RemoveConsumer(VMCache* consumer);
 
 			bool				_FreePageRange(VMCachePagesTree::Iterator it,
-									page_num_t* toPage);
+									page_num_t* toPage, page_num_t* discarded);
 
 private:
 			int32				fRefCount;
diff --git a/src/system/kernel/vm/VMAnonymousCache.cpp b/src/system/kernel/vm/VMAnonymousCache.cpp
index b77f825e75..eff47ce2f6 100644
--- a/src/system/kernel/vm/VMAnonymousCache.cpp
+++ b/src/system/kernel/vm/VMAnonymousCache.cpp
@@ -604,11 +604,14 @@ VMAnonymousCache::Rebase(off_t newBase, int priority)
 }
 
 
-status_t
+ssize_t
 VMAnonymousCache::Discard(off_t offset, off_t size)
 {
 	_FreeSwapPageRange(offset, offset + size);
-	return VMCache::Discard(offset, size);
+	const ssize_t discarded = VMCache::Discard(offset, size);
+	if (discarded > 0 && fCanOvercommit)
+		Commit(committed_size - discarded, VM_PRIORITY_USER);
+	return discarded;
 }
 
 
@@ -720,6 +723,7 @@ VMAnonymousCache::Commit(off_t size, int priority)
 	TRACE("%p->VMAnonymousCache::Commit(%" B_PRIdOFF ")\n", this, size);
 
 	AssertLocked();
+	ASSERT(size >= (page_count * B_PAGE_SIZE));
 
 	// If we can overcommit, we don't commit here, but in Fault(). We always
 	// unreserve memory, if we're asked to shrink our commitment, though.
diff --git a/src/system/kernel/vm/VMAnonymousCache.h b/src/system/kernel/vm/VMAnonymousCache.h
index b312e52df0..1d1142869d 100644
--- a/src/system/kernel/vm/VMAnonymousCache.h
+++ b/src/system/kernel/vm/VMAnonymousCache.h
@@ -47,7 +47,7 @@ public:
 	virtual	status_t			Adopt(VMCache* source, off_t offset,
 									off_t size, off_t newOffset);
 
-	virtual	status_t			Discard(off_t offset, off_t size);
+	virtual	ssize_t				Discard(off_t offset, off_t size);
 
 	virtual	bool				CanOvercommit();
 	virtual	status_t			Commit(off_t size, int priority);
diff --git a/src/system/kernel/vm/VMAnonymousNoSwapCache.cpp b/src/system/kernel/vm/VMAnonymousNoSwapCache.cpp
index 09bc70f082..d5d7276f0f 100644
--- a/src/system/kernel/vm/VMAnonymousNoSwapCache.cpp
+++ b/src/system/kernel/vm/VMAnonymousNoSwapCache.cpp
@@ -58,10 +58,21 @@ VMAnonymousNoSwapCache::Init(bool canOvercommit, int32 numPrecommittedPages,
 }
 
 
+ssize_t
+VMAnonymousNoSwapCache::Discard(off_t offset, off_t size)
+{
+	const ssize_t discarded = VMCache::Discard(offset, size);
+	if (discarded > 0 && fCanOvercommit)
+		Commit(committed_size - discarded, VM_PRIORITY_USER);
+	return discarded;
+}
+
+
 status_t
 VMAnonymousNoSwapCache::Commit(off_t size, int priority)
 {
 	AssertLocked();
+	ASSERT(size >= (page_count * B_PAGE_SIZE));
 
 	// If we can overcommit, we don't commit here, but in Fault(). We always
 	// unreserve memory, if we're asked to shrink our commitment, though.
diff --git a/src/system/kernel/vm/VMAnonymousNoSwapCache.h b/src/system/kernel/vm/VMAnonymousNoSwapCache.h
index 615f20f026..8acda57f92 100644
--- a/src/system/kernel/vm/VMAnonymousNoSwapCache.h
+++ b/src/system/kernel/vm/VMAnonymousNoSwapCache.h
@@ -22,6 +22,8 @@ public:
 									int32 numGuardPages,
 									uint32 allocationFlags);
 
+	virtual	ssize_t				Discard(off_t offset, off_t size);
+
 	virtual	bool				CanOvercommit();
 	virtual	status_t			Commit(off_t size, int priority);
 	virtual	bool				HasPage(off_t offset);
diff --git a/src/system/kernel/vm/VMCache.cpp b/src/system/kernel/vm/VMCache.cpp
index d152ad47bc..49c586e5c7 100644
--- a/src/system/kernel/vm/VMCache.cpp
+++ b/src/system/kernel/vm/VMCache.cpp
@@ -1106,7 +1106,7 @@ VMCache::SetMinimalCommitment(off_t commitment, int priority)
 
 bool
 VMCache::_FreePageRange(VMCachePagesTree::Iterator it,
-	page_num_t* toPage = NULL)
+	page_num_t* toPage = NULL, page_num_t* discarded = NULL)
 {
 	for (vm_page* page = it.Next();
 		page != NULL && (toPage == NULL || page->cache_offset < *toPage);
@@ -1118,6 +1118,8 @@ VMCache::_FreePageRange(VMCachePagesTree::Iterator it,
 				// as we might cause a deadlock this way
 				page->busy_writing = false;
 					// this will notify the writer to free the page
+				if (discarded != NULL)
+					(*discarded)++;
 				continue;
 			}
 
@@ -1138,6 +1140,8 @@ VMCache::_FreePageRange(VMCachePagesTree::Iterator it,
 			// removing the current node is safe.
 
 		vm_page_free(this, page);
+		if (discarded != NULL)
+			(*discarded)++;
 	}
 
 	return false;
@@ -1241,15 +1245,16 @@ VMCache::Adopt(VMCache* source, off_t offset, off_t size, off_t newOffset)
 
 
 /*! Discards pages in the given range. */
-status_t
+ssize_t
 VMCache::Discard(off_t offset, off_t size)
 {
+	page_num_t discarded = 0;
 	page_num_t startPage = offset >> PAGE_SHIFT;
 	page_num_t endPage = (offset + size + B_PAGE_SIZE - 1) >> PAGE_SHIFT;
-	while (_FreePageRange(pages.GetIterator(startPage, true, true), &endPage))
+	while (_FreePageRange(pages.GetIterator(startPage, true, true), &endPage, &discarded))
 		;
 
-	return B_OK;
+	return (discarded * B_PAGE_SIZE);
 }