diff --git a/src/kernel/core/vm/Jamfile b/src/kernel/core/vm/Jamfile
index c5a410ca18..866294ae05 100644
--- a/src/kernel/core/vm/Jamfile
+++ b/src/kernel/core/vm/Jamfile
@@ -1,15 +1,16 @@
 SubDir OBOS_TOP src kernel core vm ;
 
-KernelStaticLibrary libvm : 
-	<$(SOURCE_GRIST)>vm.c 
-	<$(SOURCE_GRIST)>vm_cache.c 
-	<$(SOURCE_GRIST)>vm_daemons.c 
-	<$(SOURCE_GRIST)>vm_page.c 
-	<$(SOURCE_GRIST)>vm_store_anonymous_noswap.c 
-	<$(SOURCE_GRIST)>vm_store_device.c 
-	<$(SOURCE_GRIST)>vm_store_null.c 
-	<$(SOURCE_GRIST)>vm_store_vnode.c 
-	<$(SOURCE_GRIST)>vm_tests.c 
+KernelMergeObject kernel_vm.o :
+	<$(SOURCE_GRIST)>cache.cpp
+	<$(SOURCE_GRIST)>vm.c
+	<$(SOURCE_GRIST)>vm_cache.c
+	<$(SOURCE_GRIST)>vm_daemons.c
+	<$(SOURCE_GRIST)>vm_page.c
+	<$(SOURCE_GRIST)>vm_store_anonymous_noswap.c
+	<$(SOURCE_GRIST)>vm_store_device.c
+	<$(SOURCE_GRIST)>vm_store_null.c
+	<$(SOURCE_GRIST)>vm_store_vnode.c
+	<$(SOURCE_GRIST)>vm_tests.c
 	:
 	-fno-pic -Wno-unused -D_KERNEL_
 	;
diff --git a/src/kernel/core/vm/cache.cpp b/src/kernel/core/vm/cache.cpp
new file mode 100644
index 0000000000..0983de7503
--- /dev/null
+++ b/src/kernel/core/vm/cache.cpp
@@ -0,0 +1,348 @@
+/* 
+** Copyright 2004, Axel Dörfler, axeld@pinc-software.de. All rights reserved.
+** Distributed under the terms of the OpenBeOS License.
+*/
+
+
+#include <KernelExport.h>
+
+#include <cache.h>
+#include <khash.h>
+#include <lock.h>
+
+#include <unistd.h>
+#include <stdlib.h>
+#include <string.h>
+
+
+/* Note, this is just a dump and simple cache implementation targeted to be
+ * compatible with the one found in BeOS.
+ * This won't be the cache we end up using in R1, seriously :-)
+ */
+
+
+struct cache_entry {
+	cache_entry	*next;			// next in hash
+	off_t		block_number;
+	void		*data;
+	int32		ref_count;
+	int32		lock;
+	bool		is_dirty;
+};
+
+struct cache {
+	hash_table	*hash;
+	mutex		lock;
+	off_t		max_blocks;
+	size_t		block_size;
+};
+
+static const int32 kNumCaches = 16;
+struct cache sCaches[kNumCaches];
+	// we can cache the first 16 fds (I said we were dumb, right?)
+
+
+class MutexLocker {
+	public:
+		MutexLocker(int fd)
+			: fMutex(NULL)
+		{
+			if (fd < 0 || fd >= kNumCaches)
+				return;
+
+			fMutex = &sCaches[fd].lock;
+			mutex_lock(fMutex);
+		}
+
+		~MutexLocker()
+		{
+			if (fMutex != NULL)
+				mutex_unlock(fMutex);
+		}
+
+		status_t InitCheck()
+		{
+			return fMutex != NULL ? B_OK : B_ERROR;
+		}
+
+	private:
+		mutex	*fMutex;
+};
+
+
+static int
+cache_entry_compare(void *_cacheEntry, const void *_block)
+{
+	cache_entry *cacheEntry = (cache_entry *)_cacheEntry;
+	const off_t *block = (const off_t *)_block;
+
+	return cacheEntry->block_number - *block;
+}
+
+
+static uint32
+cache_entry_hash(void *_cacheEntry, const void *_block, uint32 range)
+{
+	cache_entry *cacheEntry = (cache_entry *)_cacheEntry;
+	const off_t *block = (const off_t *)_block;
+
+	if (cacheEntry != NULL)
+		return cacheEntry->block_number % range;
+
+	return *block % range;
+}
+
+
+static status_t
+write_cache_entry(int fd, cache_entry *entry)
+{
+	ssize_t blockSize = sCaches[fd].block_size;
+
+	return write_pos(fd, entry->block_number * blockSize,
+				entry->data, blockSize) < blockSize ? B_ERROR : B_OK;
+}
+
+
+void
+put_cache_entry(int fd, cache_entry *entry)
+{
+	if (entry == NULL)
+		return;
+
+	if (--entry->ref_count > 0)
+		return;
+
+	// write out entry if necessary
+	if (entry->is_dirty
+		&& write_cache_entry(fd, entry) != B_OK)
+		panic("could not write back entry\n");
+
+	hash_remove(sCaches[fd].hash, entry);
+	free(entry->data);
+	free(entry);
+}
+
+
+cache_entry *
+new_cache_entry(int fd, off_t blockNumber, size_t blockSize)
+{
+	cache_entry *entry = (cache_entry *)malloc(sizeof(cache_entry));
+	if (entry == NULL)
+		return NULL;
+
+	entry->data = malloc(blockSize);
+	if (entry->data == NULL) {
+		free(entry);
+		return NULL;
+	}
+
+	sCaches[fd].block_size = blockSize;
+		// store block size (once would be sufficient, though)
+
+	entry->block_number = blockNumber;
+	entry->ref_count = 1;
+	entry->lock = 0;
+	entry->is_dirty = false;
+
+	hash_insert(sCaches[fd].hash, entry);
+
+	return entry;
+}
+
+
+cache_entry *
+lookup_cache_entry(int fd, off_t blockNumber)
+{
+	cache_entry *entry = (cache_entry *)hash_lookup(sCaches[fd].hash, &blockNumber);
+	if (entry == NULL)
+		return NULL;
+
+	entry->ref_count++;
+	return entry;
+}
+
+
+//	#pragma mark -
+//	public BeOS compatible interface
+
+
+extern "C" void
+force_cache_flush(int dev, int prefer_log_blocks)
+{
+}
+
+
+extern "C" int
+flush_blocks(int dev, off_t bnum, int nblocks)
+{
+	return 0;
+}
+
+
+extern "C" int
+flush_device(int dev, int warn_locked)
+{
+	return 0;
+}
+
+
+extern "C" int
+init_cache_for_device(int fd, off_t max_blocks)
+{
+	if (fd < 0 || fd >= kNumCaches)
+		return B_ERROR;
+	if (sCaches[fd].hash != NULL)
+		return B_BUSY;
+
+	sCaches[fd].hash = hash_init(32, 0, &cache_entry_compare, &cache_entry_hash);
+	return mutex_init(&sCaches[fd].lock, "block cache");
+}
+
+
+extern "C" int
+remove_cached_device_blocks(int fd, int allowWrite)
+{
+	if (fd < 0 || fd >= kNumCaches || sCaches[fd].hash == NULL)
+		return B_ERROR;
+
+	if (allowWrite) {
+		hash_iterator iterator;
+		hash_open(sCaches[fd].hash, &iterator);
+		
+		cache_entry *entry;
+		while ((entry = (cache_entry *)hash_next(sCaches[fd].hash, &iterator)) != NULL) {
+			// write out entry if necessary
+			if (entry->is_dirty
+				&& write_cache_entry(fd, entry) != B_OK)
+				panic("write failed!\n");
+		}
+		hash_close(sCaches[fd].hash, &iterator, false);
+	}
+
+	hash_uninit(sCaches[fd].hash);
+	mutex_destroy(&sCaches[fd].lock);
+
+	return 0;
+}
+
+
+extern "C" void *
+get_block(int fd, off_t blockNumber, int blockSize)
+{
+	MutexLocker locker(fd);
+	if (locker.InitCheck() != B_OK)
+		return NULL;
+
+	cache_entry *entry = lookup_cache_entry(fd, blockNumber);
+	if (entry == NULL) {
+		// read entry into cache
+		entry = new_cache_entry(fd, blockNumber, blockSize);
+		if (entry == NULL)
+			return NULL;
+
+		if (read_pos(fd, blockNumber * blockSize, entry->data, blockSize) < blockSize) {
+			free(entry->data);
+			free(entry);
+			return NULL;
+		}
+	}
+
+	entry->lock++;
+
+	return entry->data;
+}
+
+
+extern "C" void *
+get_empty_block(int fd, off_t blockNumber, int blockSize)
+{
+	MutexLocker locker(fd);
+	if (locker.InitCheck() != B_OK)
+		return NULL;
+
+	cache_entry *entry = lookup_cache_entry(fd, blockNumber);
+	if (entry == NULL) {
+		// create new cache entry
+		entry = new_cache_entry(fd, blockNumber, blockSize);
+		if (entry == NULL)
+			return NULL;
+	}
+	memset(entry->data, 0, blockSize);
+
+	entry->lock++;
+
+	return entry->data;
+}
+
+
+extern "C" int
+release_block(int fd, off_t blockNumber)
+{
+	MutexLocker locker(fd);
+	if (locker.InitCheck() != B_OK)
+		return B_ERROR;
+
+	cache_entry *entry = lookup_cache_entry(fd, blockNumber);
+	if (entry == NULL)
+		panic("release_block() called on block %Ld that was not cached\n", blockNumber);
+
+	entry->lock--;
+	put_cache_entry(fd, entry);
+	return 0;
+}
+
+
+extern "C" int
+mark_blocks_dirty(int fd, off_t blockNumber, int numBlocks)
+{
+	return -1;
+}
+
+
+extern "C" int
+cached_read(int fd, off_t blockNumber, void *data, off_t numBlocks, int blockSize)
+{
+	MutexLocker locker(fd);
+	if (locker.InitCheck() != B_OK)
+		return B_ERROR;
+
+	ssize_t bytes = numBlocks * blockSize;
+	if (read_pos(fd, blockNumber * blockSize, data, bytes) < bytes)
+		return -1;
+
+	// check if there are any cached blocks in that range shadowing the device contents
+
+	for (off_t blockOffset = 0; blockOffset < numBlocks; blockOffset++) {
+		cache_entry *entry = lookup_cache_entry(fd, blockNumber + blockOffset);
+		if (entry != NULL) {
+			if (entry->is_dirty)
+				memcpy((void *)((addr_t)data + blockOffset * blockSize), entry->data, blockSize);
+			put_cache_entry(fd, entry);
+		}
+	}
+	return 0;
+}
+
+
+extern "C" int
+cached_write(int fd, off_t blockNumber, const void *data, off_t numBlocks, int blockSize)
+{
+	return -1;
+}
+
+
+extern "C" int
+cached_write_locked(int fd, off_t blockNumber, const void *data, off_t numBlocks, int blockSize)
+{
+	return -1;
+}
+
+
+extern "C" int
+set_blocks_info(int fd, off_t *blocks, int numBlocks,
+	void (*func)(off_t bnum, size_t nblocks, void *arg),
+	void *arg)
+{
+	return -1;
+}
+