X-Git-Url: https://git.ucc.asn.au/?a=blobdiff_plain;f=Kernel%2Farch%2Fx86_64%2Fmm_virt.c;h=a8d890a1547641d07157d9f34ebf312dadd7bb88;hb=f086aa018b58f23bc15fbee7b2c648e35bb7dc1c;hp=a0b384db95063ac957d040cddd679143e04fd0c0;hpb=273ede03312a1b5c537d250403ff11065106e186;p=tpg%2Facess2.git

diff --git a/Kernel/arch/x86_64/mm_virt.c b/Kernel/arch/x86_64/mm_virt.c
index a0b384db..a8d890a1 100644
--- a/Kernel/arch/x86_64/mm_virt.c
+++ b/Kernel/arch/x86_64/mm_virt.c
@@ -6,10 +6,12 @@
 #define DEBUG	0
 #include <acess.h>
 #include <mm_virt.h>
+#include <threads_int.h>
 #include <proc.h>
 
 // === CONSTANTS ===
 #define PHYS_BITS	52	// TODO: Move out
+#define VIRT_BITS	48
 
 #define PML4_SHIFT	39
 #define PDP_SHIFT	30
@@ -17,32 +19,35 @@
 #define PTAB_SHIFT	12
 
 #define	PADDR_MASK	0x7FFFFFFF##FFFFF000
-#define PAGE_MASK	(((Uint)1 << 36)-1)
-#define TABLE_MASK	(((Uint)1 << 27)-1)
-#define PDP_MASK	(((Uint)1 << 18)-1)
-#define PML4_MASK	(((Uint)1 << 9)-1)
+#define PAGE_MASK	((1LL << 36)-1)
+#define TABLE_MASK	((1LL << 27)-1)
+#define PDP_MASK	((1LL << 18)-1)
+#define PML4_MASK	((1LL << 9)-1)
 
 #define	PF_PRESENT	0x001
 #define	PF_WRITE	0x002
 #define	PF_USER		0x004
-#define	PF_LARGE	0x000
+#define	PF_LARGE	0x080
+#define	PF_GLOBAL	0x100
 #define	PF_COW		0x200
 #define	PF_PAGED	0x400
 #define	PF_NX		0x80000000##00000000
 
 // === MACROS ===
-#define PAGETABLE(idx)	(*((tPAddr*)MM_FRACTAL_BASE+((idx)&PAGE_MASK)))
+#define PAGETABLE(idx)	(*((Uint64*)MM_FRACTAL_BASE+((idx)&PAGE_MASK)))
 #define PAGEDIR(idx)	PAGETABLE((MM_FRACTAL_BASE>>12)+((idx)&TABLE_MASK))
 #define PAGEDIRPTR(idx)	PAGEDIR((MM_FRACTAL_BASE>>21)+((idx)&PDP_MASK))
 #define PAGEMAPLVL4(idx)	PAGEDIRPTR((MM_FRACTAL_BASE>>30)+((idx)&PML4_MASK))
 
 #define TMPCR3()	PAGEMAPLVL4(MM_TMPFRAC_BASE>>39)
-#define TMPTABLE(idx)	(*((tPAddr*)MM_TMPFRAC_BASE+((idx)&PAGE_MASK)))
+#define TMPTABLE(idx)	(*((Uint64*)MM_TMPFRAC_BASE+((idx)&PAGE_MASK)))
 #define TMPDIR(idx)	PAGETABLE((MM_TMPFRAC_BASE>>12)+((idx)&TABLE_MASK))
 #define TMPDIRPTR(idx)	PAGEDIR((MM_TMPFRAC_BASE>>21)+((idx)&PDP_MASK))
 #define TMPMAPLVL4(idx)	PAGEDIRPTR((MM_TMPFRAC_BASE>>30)+((idx)&PML4_MASK))
 
-#define INVLPG(__addr)	__asm__ __volatile__ ("invlpg (%0)"::"r"(__addr));
+#define INVLPG(__addr)	__asm__ __volatile__ ("invlpg (%0)"::"r"(__addr))
+#define INVLPG_ALL()	__asm__ __volatile__ ("mov %cr3,%rax;\n\tmov %rax,%cr3;")
+#define INVLPG_GLOBAL()	__asm__ __volatile__ ("mov %cr4,%rax;\n\txorl $0x80, %eax;\n\tmov %rax,%cr4;\n\txorl $0x80, %eax;\n\tmov %rax,%cr4")
 
 // === CONSTS ===
 //tPAddr	* const gaPageTable = MM_FRACTAL_BASE;
@@ -50,11 +55,14 @@
 // === IMPORTS ===
 extern void	Error_Backtrace(Uint IP, Uint BP);
 extern tPAddr	gInitialPML4[512];
+extern void	Threads_SegFault(tVAddr Addr);
+extern char	_UsertextBase[];
 
 // === PROTOTYPES ===
 void	MM_InitVirt(void);
 //void	MM_FinishVirtualInit(void);
-void	MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs);
+void	MM_int_ClonePageEnt( Uint64 *Ent, void *NextLevel, tVAddr Addr, int bTable );
+ int	MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs);
 void	MM_DumpTables(tVAddr Start, tVAddr End);
  int	MM_GetPageEntryPtr(tVAddr Addr, BOOL bTemp, BOOL bAllocate, BOOL bLargePage, tPAddr **Pointer);
  int	MM_MapEx(tVAddr VAddr, tPAddr PAddr, BOOL bTemp, BOOL bLarge);
@@ -69,57 +77,131 @@ tMutex	glMM_TempFractalLock;
 // === CODE ===
 void MM_InitVirt(void)
 {
-	MM_DumpTables(0, -1L);
+	Log_Debug("MMVirt", "&PAGEMAPLVL4(0) = %p", &PAGEMAPLVL4(0));
+//	MM_DumpTables(0, -1L);
 }
 
 void MM_FinishVirtualInit(void)
 {
+	PAGEMAPLVL4(0) = 0;
 }
 
 /**
+ * \brief Clone a page from an entry
+ * \param Ent	Pointer to the entry in the PML4/PDP/PD/PT
+ * \param NextLevel	Pointer to contents of the entry
+ * \param Addr	Dest address
+ * \note Used in COW
+ */
+void MM_int_ClonePageEnt( Uint64 *Ent, void *NextLevel, tVAddr Addr, int bTable )
+{
+	tPAddr	curpage = *Ent & PADDR_MASK; 
+	if( MM_GetRefCount( curpage ) <= 0 ) {
+		Log_KernelPanic("MMVirt", "Page %P still marked COW, but unreferenced", curpage);
+	}
+//	Log_Debug("MM_Virt", "%P refcount %i", curpage, MM_GetRefCount( curpage ));
+	if( MM_GetRefCount( curpage ) == 1 )
+	{
+		*Ent &= ~PF_COW;
+		*Ent |= PF_PRESENT|PF_WRITE;
+//		Log_Debug("MMVirt", "COW ent at %p (%p), last (%P)", Ent, NextLevel, curpage);
+	}
+	else
+	{
+		void	*tmp;
+		tPAddr	paddr;
+		
+		if( !(paddr = MM_AllocPhys()) ) {
+			Threads_SegFault(Addr);
+			return ;
+		}
+
+		ASSERT(paddr != curpage);
+			
+		tmp = (void*)MM_MapTemp(paddr);
+		memcpy( tmp, NextLevel, 0x1000 );
+		MM_FreeTemp( (tVAddr)tmp );
+		
+//		Log_Debug("MMVirt", "COW ent at %p (%p) from %P to %P", Ent, NextLevel, curpage, paddr);
+
+		MM_DerefPhys( curpage );
+		*Ent &= PF_USER;
+		*Ent |= paddr|PF_PRESENT|PF_WRITE;
+	}
+	INVLPG( (tVAddr)NextLevel );
+	
+	// Mark COW on pages
+	if(bTable) 
+	{
+		Uint64	*dp = NextLevel;
+		 int	i;
+		for( i = 0; i < 512; i ++ )
+		{
+			if( !(dp[i] & PF_PRESENT) )	continue;
+			MM_RefPhys( dp[i] & PADDR_MASK );
+			if( dp[i] & PF_WRITE ) {
+				dp[i] &= ~PF_WRITE;
+				dp[i] |= PF_COW;
+			}
+		}
+	}
+}
+
+/*
  * \brief Called on a page fault
  */
-void MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs)
+int MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs)
 {
 	// TODO: Implement Copy-on-Write
-	#if 0
-	if( gaPageDir  [Addr>>22] & PF_PRESENT
-	 && gaPageTable[Addr>>12] & PF_PRESENT
-	 && gaPageTable[Addr>>12] & PF_COW )
+	#if 1
+	if( PAGEMAPLVL4(Addr>>39) & PF_PRESENT
+	 && PAGEDIRPTR (Addr>>30) & PF_PRESENT
+	 && PAGEDIR    (Addr>>21) & PF_PRESENT
+	 && PAGETABLE  (Addr>>12) & PF_PRESENT )
 	{
-		tPAddr	paddr;
-		if(MM_GetRefCount( gaPageTable[Addr>>12] & ~0xFFF ) == 1)
+		// PML4 Entry
+		if( PAGEMAPLVL4(Addr>>39) & PF_COW )
 		{
-			gaPageTable[Addr>>12] &= ~PF_COW;
-			gaPageTable[Addr>>12] |= PF_PRESENT|PF_WRITE;
+			tPAddr	*dp = &PAGEDIRPTR((Addr>>39)*512);
+			MM_int_ClonePageEnt( &PAGEMAPLVL4(Addr>>39), dp, Addr, 1 );
+//			MM_DumpTables(Addr>>39 << 39, (((Addr>>39) + 1) << 39) - 1);
 		}
-		else
+		// PDP Entry
+		if( PAGEDIRPTR(Addr>>30) & PF_COW )
 		{
-			//Log("MM_PageFault: COW - MM_DuplicatePage(0x%x)", Addr);
-			paddr = MM_DuplicatePage( Addr );
-			MM_DerefPhys( gaPageTable[Addr>>12] & ~0xFFF );
-			gaPageTable[Addr>>12] &= PF_USER;
-			gaPageTable[Addr>>12] |= paddr|PF_PRESENT|PF_WRITE;
+			tPAddr	*dp = &PAGEDIR( (Addr>>30)*512 );
+			MM_int_ClonePageEnt( &PAGEDIRPTR(Addr>>30), dp, Addr, 1 );
+//			MM_DumpTables(Addr>>30 << 30, (((Addr>>30) + 1) << 30) - 1);
+		}
+		// PD Entry
+		if( PAGEDIR(Addr>>21) & PF_COW )
+		{
+			tPAddr	*dp = &PAGETABLE( (Addr>>21)*512 );
+			MM_int_ClonePageEnt( &PAGEDIR(Addr>>21), dp, Addr, 1 );
+//			MM_DumpTables(Addr>>21 << 21, (((Addr>>21) + 1) << 21) - 1);
+		}
+		// PT Entry
+		if( PAGETABLE(Addr>>12) & PF_COW )
+		{
+			MM_int_ClonePageEnt( &PAGETABLE(Addr>>12), (void*)(Addr & ~0xFFF), Addr, 0 );
+			INVLPG( Addr & ~0xFFF );
+			return 0;
 		}
-		
-		INVLPG( Addr & ~0xFFF );
-		return;
 	}
 	#endif
 	
 	// If it was a user, tell the thread handler
 	if(ErrorCode & 4) {
-		Warning("%s %s %s memory%s",
-			(ErrorCode&4?"User":"Kernel"),
+		Warning("User %s %s memory%s",
 			(ErrorCode&2?"write to":"read from"),
 			(ErrorCode&1?"bad/locked":"non-present"),
 			(ErrorCode&16?" (Instruction Fetch)":"")
 			);
-		Warning("User Pagefault: Instruction at %04x:%08x accessed %p",
+		Warning("User Pagefault: Instruction at %04x:%p accessed %p",
 			Regs->CS, Regs->RIP, Addr);
 		__asm__ __volatile__ ("sti");	// Restart IRQs
-//		Threads_SegFault(Addr);
-		return ;
+		Threads_SegFault(Addr);
+		return 0;
 	}
 	
 	// Kernel #PF
@@ -129,8 +211,7 @@ void MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs)
 		Warning("Reserved Bits Trashed!");
 	else
 	{
-		Warning("%s %s %s memory%s",
-			(ErrorCode&4?"User":"Kernel"),
+		Warning("Kernel %s %s memory%s",
 			(ErrorCode&2?"write to":"read from"),
 			(ErrorCode&1?"bad/locked":"non-present"),
 			(ErrorCode&16?" (Instruction Fetch)":"")
@@ -142,10 +223,8 @@ void MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs)
 	Error_Backtrace(Regs->RIP, Regs->RBP);
 	
 	MM_DumpTables(0, -1);
-	
-	__asm__ __volatile__ ("cli");
-	for( ;; )
-		HALT();
+
+	return 1;	
 }
 
 /**
@@ -154,7 +233,7 @@ void MM_PageFault(tVAddr Addr, Uint ErrorCode, tRegs *Regs)
 void MM_DumpTables(tVAddr Start, tVAddr End)
 {
 	#define CANOICAL(addr)	((addr)&0x800000000000?(addr)|0xFFFF000000000000:(addr))
-	const tPAddr	CHANGEABLE_BITS = 0xFF8;
+	const tPAddr	CHANGEABLE_BITS = ~(PF_PRESENT|PF_WRITE|PF_USER|PF_COW|PF_PAGED) & 0xFFF;
 	const tPAddr	MASK = ~CHANGEABLE_BITS;	// Physical address and access bits
 	tVAddr	rangeStart = 0;
 	tPAddr	expected = CHANGEABLE_BITS;	// CHANGEABLE_BITS is used because it's not a vaild value
@@ -180,17 +259,17 @@ void MM_DumpTables(tVAddr Start, tVAddr End)
 		//Debug("&PAGETABLE(%i page) = %p", page, &PAGETABLE(page));
 		
 		// End of a range
-		if(
-			!(PAGEMAPLVL4(page>>27) & PF_PRESENT)
-		||	!(PAGEDIRPTR(page>>18) & PF_PRESENT)
-		||	!(PAGEDIR(page>>9) & PF_PRESENT)
-		||  !(PAGETABLE(page) & PF_PRESENT)
-		||  (PAGETABLE(page) & MASK) != expected)
+		if(!(PAGEMAPLVL4(page>>27) & PF_PRESENT)
+		|| !(PAGEDIRPTR(page>>18) & PF_PRESENT)
+		|| !(PAGEDIR(page>>9) & PF_PRESENT)
+		|| !(PAGETABLE(page) & PF_PRESENT)
+		|| (PAGETABLE(page) & MASK) != expected)
 		{			
-			if(expected != CHANGEABLE_BITS) {
-				Log("%016llx => %013llx : 0x%6llx (%c%c%c%c)",
+			if(expected != CHANGEABLE_BITS)
+			{
+				Log("%016llx => %13llx : 0x%6llx (%c%c%c%c)",
 					CANOICAL(rangeStart),
-					PAGETABLE(rangeStart>>12) & ~0xFFF,
+					PAGETABLE(rangeStart>>12) & PADDR_MASK,
 					curPos - rangeStart,
 					(expected & PF_PAGED ? 'p' : '-'),
 					(expected & PF_COW ? 'C' : '-'),
@@ -202,19 +281,16 @@ void MM_DumpTables(tVAddr Start, tVAddr End)
 			if( !(PAGEMAPLVL4(page>>27) & PF_PRESENT) ) {
 				page += (1 << 27) - 1;
 				curPos += (1L << 39) - 0x1000;
-				//Debug("pml4 ent unset (page = 0x%x now)", page);
 				continue;
 			}
 			if( !(PAGEDIRPTR(page>>18) & PF_PRESENT) ) {
 				page += (1 << 18) - 1;
 				curPos += (1L << 30) - 0x1000;
-				//Debug("pdp ent unset (page = 0x%x now)", page);
 				continue;
 			}
 			if( !(PAGEDIR(page>>9) & PF_PRESENT) ) {
 				page += (1 << 9) - 1;
 				curPos += (1L << 21) - 0x1000;
-				//Debug("pd ent unset (page = 0x%x now)", page);
 				continue;
 			}
 			if( !(PAGETABLE(page) & PF_PRESENT) )	continue;
@@ -227,9 +303,9 @@ void MM_DumpTables(tVAddr Start, tVAddr End)
 	}
 	
 	if(expected != CHANGEABLE_BITS) {
-		Log("%016llx => %013llx : 0x%6llx (%c%c%c%c)",
+		Log("%016llx => %13llx : 0x%6llx (%c%c%c%c)",
 			CANOICAL(rangeStart),
-			PAGETABLE(rangeStart>>12) & ~0xFFF,
+			PAGETABLE(rangeStart>>12) & PADDR_MASK,
 			curPos - rangeStart,
 			(expected & PF_PAGED ? 'p' : '-'),
 			(expected & PF_COW ? 'C' : '-'),
@@ -258,13 +334,22 @@ int MM_GetPageEntryPtr(tVAddr Addr, BOOL bTemp, BOOL bAllocate, BOOL bLargePage,
 	const int	nADDR_SIZES = sizeof(ADDR_SIZES)/sizeof(ADDR_SIZES[0]);
 	 int	i;
 	
+	#define BITMASK(bits)	( (1LL << (bits))-1 )
+
 	if( bTemp )
-		pmlevels[3] = (void*)MM_TMPFRAC_BASE;	// Temporary Page Table
+	{
+		pmlevels[3] = &TMPTABLE(0);	// Page Table
+		pmlevels[2] = &TMPDIR(0);	// PDIR
+		pmlevels[1] = &TMPDIRPTR(0);	// PDPT
+		pmlevels[0] = &TMPMAPLVL4(0);	// PML4
+	}
 	else
+	{
 		pmlevels[3] = (void*)MM_FRACTAL_BASE;	// Page Table
-	pmlevels[2] = &pmlevels[3][(MM_FRACTAL_BASE>>12)&PAGE_MASK];	// PDIR
-	pmlevels[1] = &pmlevels[2][(MM_FRACTAL_BASE>>21)&TABLE_MASK];	// PDPT
-	pmlevels[0] = &pmlevels[1][(MM_FRACTAL_BASE>>30)&PDP_MASK];	// PML4
+		pmlevels[2] = &pmlevels[3][(MM_FRACTAL_BASE>>12)&BITMASK(VIRT_BITS-12)];	// PDIR
+		pmlevels[1] = &pmlevels[2][(MM_FRACTAL_BASE>>21)&BITMASK(VIRT_BITS-21)];	// PDPT
+		pmlevels[0] = &pmlevels[1][(MM_FRACTAL_BASE>>30)&BITMASK(VIRT_BITS-30)];	// PML4
+	}
 	
 	// Mask address
 	Addr &= (1ULL << 48)-1;
@@ -283,11 +368,15 @@ int MM_GetPageEntryPtr(tVAddr Addr, BOOL bTemp, BOOL bAllocate, BOOL bLargePage,
 		if( !(pmlevels[i][Addr >> ADDR_SIZES[i]] & 1) )
 		{
 			if( !bAllocate )	return -4;	// If allocation is not requested, error
-			tmp = MM_AllocPhys();
-			if(!tmp)	return -2;
+			if( !(tmp = MM_AllocPhys()) )	return -2;
 			pmlevels[i][Addr >> ADDR_SIZES[i]] = tmp | 3;
-			INVLPG( &pmlevels[i+1][ (Addr>>ADDR_SIZES[i])<<9 ] );
-			memset( &pmlevels[i+1][ (Addr>>ADDR_SIZES[i])<<9 ], 0, 0x1000 );
+			if( Addr < 0x800000000000 )
+				pmlevels[i][Addr >> ADDR_SIZES[i]] |= PF_USER;
+			INVLPG( &pmlevels[i+1][ (Addr>>ADDR_SIZES[i])*512 ] );
+			memset( &pmlevels[i+1][ (Addr>>ADDR_SIZES[i])*512 ], 0, 0x1000 );
+			LOG("Init PML%i ent 0x%x %p with %P", 4 - i,
+				Addr>>ADDR_SIZES[i],
+				(Addr>>ADDR_SIZES[i])<<ADDR_SIZES[i], tmp);
 		}
 		// Catch large pages
 		else if( pmlevels[i][Addr >> ADDR_SIZES[i]] & PF_LARGE )
@@ -325,7 +414,10 @@ int MM_MapEx(tVAddr VAddr, tPAddr PAddr, BOOL bTemp, BOOL bLarge)
 	if( *ent & 1 )	LEAVE_RET('i', 0);
 	
 	*ent = PAddr | 3;
-	
+
+	if( VAddr < 0x800000000000 )
+		*ent |= PF_USER;
+
 	INVLPG( VAddr );
 
 	LEAVE('i', 1);	
@@ -353,7 +445,7 @@ void MM_Unmap(tVAddr VAddr)
 	if( !(PAGEDIRPTR(VAddr >> 30) & 1) )	return ;
 	// Check Page Dir
 	if( !(PAGEDIR(VAddr >> 21) & 1) )	return ;
-	
+
 	PAGETABLE(VAddr >> PTAB_SHIFT) = 0;
 	INVLPG( VAddr );
 }
@@ -419,7 +511,7 @@ int MM_GetPageEntry(tVAddr Addr, tPAddr *Phys, Uint *Flags)
 	ret = MM_GetPageEntryPtr(Addr, 0, 0, 0, &ptr);
 	if( ret < 0 )	return 0;
 	
-	*Phys = *ptr & ~0xFFF;
+	*Phys = *ptr & PADDR_MASK;
 	*Flags = *ptr & 0xFFF;
 	return ret;
 }
@@ -435,7 +527,9 @@ tPAddr MM_GetPhysAddr(tVAddr Addr)
 	ret = MM_GetPageEntryPtr(Addr, 0, 0, 0, &ptr);
 	if( ret < 0 )	return 0;
 	
-	return (*ptr & ~0xFFF) | (Addr & 0xFFF);
+	if( !(*ptr & 1) )	return 0;
+	
+	return (*ptr & PADDR_MASK) | (Addr & 0xFFF);
 }
 
 /**
@@ -550,6 +644,7 @@ tVAddr MM_MapHWPages(tPAddr PAddr, Uint Number)
 			ret -= 0x1000;
 			PAddr -= 0x1000;
 			MM_Map(ret, PAddr);
+			MM_RefPhys(PAddr);
 		}
 		
 		return ret;
@@ -567,6 +662,7 @@ void MM_UnmapHWPages(tVAddr VAddr, Uint Number)
 //	Log_KernelPanic("MM", "TODO: Implement MM_UnmapHWPages");
 	while( Number -- )
 	{
+		MM_DerefPhys( MM_GetPhysAddr(VAddr) );
 		MM_Unmap(VAddr);
 		VAddr += 0x1000;
 	}
@@ -595,10 +691,7 @@ tVAddr MM_AllocDMA(int Pages, int MaxBits, tPAddr *PhysAddr)
 		phys = MM_AllocPhys();
 		*PhysAddr = phys;
 		ret = MM_MapHWPages(phys, 1);
-		if(ret == 0) {
-			MM_DerefPhys(phys);
-			return 0;
-		}
+		MM_DerefPhys(phys);
 		return ret;
 	}
 	
@@ -609,10 +702,11 @@ tVAddr MM_AllocDMA(int Pages, int MaxBits, tPAddr *PhysAddr)
 	
 	// Allocated successfully, now map
 	ret = MM_MapHWPages(phys, Pages);
+	// MapHWPages references the pages, so deref them back down to 1
+	for(;Pages--;phys+=0x1000)
+		MM_DerefPhys(phys);
 	if( ret == 0 ) {
 		// If it didn't map, free then return 0
-		for(;Pages--;phys+=0x1000)
-			MM_DerefPhys(phys);
 		return 0;
 	}
 	
@@ -623,13 +717,31 @@ tVAddr MM_AllocDMA(int Pages, int MaxBits, tPAddr *PhysAddr)
 // --- Tempory Mappings ---
 tVAddr MM_MapTemp(tPAddr PAddr)
 {
-	Log_KernelPanic("MM", "TODO: Implement MM_MapTemp");
+	const int max_slots = (MM_TMPMAP_END - MM_TMPMAP_BASE) / PAGE_SIZE;
+	tVAddr	ret = MM_TMPMAP_BASE;
+	 int	i;
+	
+	for( i = 0; i < max_slots; i ++, ret += PAGE_SIZE )
+	{
+		tPAddr	*ent;
+		if( MM_GetPageEntryPtr( ret, 0, 1, 0, &ent) < 0 ) {
+			continue ;
+		}
+
+		if( *ent & 1 )
+			continue ;
+
+		*ent = PAddr | 3;
+		MM_RefPhys(PAddr);
+		INVLPG(ret);
+		return ret;
+	}
 	return 0;
 }
 
 void MM_FreeTemp(tVAddr VAddr)
 {
-	Log_KernelPanic("MM", "TODO: Implement MM_FreeTemp");
+	MM_Deallocate(VAddr);
 	return ;
 }
 
@@ -638,7 +750,9 @@ void MM_FreeTemp(tVAddr VAddr)
 tPAddr MM_Clone(void)
 {
 	tPAddr	ret;
-	
+	 int	i;
+	tVAddr	kstackbase;
+
 	// #1 Create a copy of the PML4
 	ret = MM_AllocPhys();
 	if(!ret)	return 0;
@@ -646,82 +760,164 @@ tPAddr MM_Clone(void)
 	// #2 Alter the fractal pointer
 	Mutex_Acquire(&glMM_TempFractalLock);
 	TMPCR3() = ret | 3;
+	INVLPG_ALL();
 	
-	INVLPG(&TMPMAPLVL4(0));
-	memcpy(&TMPMAPLVL4(0), &PAGEMAPLVL4(0), 0x1000);
+	// #3 Set Copy-On-Write to all user pages
+	for( i = 0; i < 256; i ++)
+	{
+		TMPMAPLVL4(i) = PAGEMAPLVL4(i);
+//		Log_Debug("MM", "TMPMAPLVL4(%i) = 0x%016llx", i, TMPMAPLVL4(i));
+		if( !(TMPMAPLVL4(i) & PF_PRESENT) )	continue ;
+		
+		MM_RefPhys( TMPMAPLVL4(i) & PADDR_MASK );
+		
+		if( TMPMAPLVL4(i) & PF_WRITE ) {
+			TMPMAPLVL4(i) |= PF_COW;
+			TMPMAPLVL4(i) &= ~PF_WRITE;
+		}
+	}
 	
-	Log_KernelPanic("MM", "TODO: Implement MM_Clone");
+	// #4 Map in kernel pages
+	for( i = 256; i < 512; i ++ )
+	{
+		// Skip addresses:
+		// 320 0xFFFFA....	- Kernel Stacks
+		if( i == 320 )	continue;
+		// 509 0xFFFFFE0..	- Fractal mapping
+		if( i == 508 )	continue;
+		// 510 0xFFFFFE8..	- Temp fractal mapping
+		if( i == 509 )	continue;
+		
+		TMPMAPLVL4(i) = PAGEMAPLVL4(i);
+		if( TMPMAPLVL4(i) & 1 )
+			MM_RefPhys( TMPMAPLVL4(i) & PADDR_MASK );
+	}
 	
-	// #3 Set Copy-On-Write to all user pages
+	// #5 Set fractal mapping
+	TMPMAPLVL4(508) = ret | 3;
+	TMPMAPLVL4(509) = 0;	// Temp
 	
+	// #6 Create kernel stack
+	//  tThread->KernelStack is the top
+	//  There is 1 guard page below the stack
+	kstackbase = Proc_GetCurThread()->KernelStack - KERNEL_STACK_SIZE;
+
+	Log("MM_Clone: kstackbase = %p", kstackbase);
 	
-	// #4 Return
+	TMPMAPLVL4(MM_KSTACK_BASE >> PML4_SHIFT) = 0;
+	for( i = 1; i < KERNEL_STACK_SIZE/0x1000; i ++ )
+	{
+		tPAddr	phys = MM_AllocPhys();
+		tVAddr	tmpmapping;
+		MM_MapEx(kstackbase+i*0x1000, phys, 1, 0);
+		
+		Log_Debug("MM", "MM_Clone: Cloning stack page %p from %P to %P",
+			kstackbase+i*0x1000, MM_GetPhysAddr( kstackbase+i*0x1000 ), phys
+			);
+		tmpmapping = MM_MapTemp(phys);
+		if( MM_GetPhysAddr( kstackbase+i*0x1000 ) )
+			memcpy((void*)tmpmapping, (void*)(kstackbase+i*0x1000), 0x1000);
+		else
+			memset((void*)tmpmapping, 0, 0x1000);
+//		if( i == 0xF )
+//			Debug_HexDump("MM_Clone: *tmpmapping = ", (void*)tmpmapping, 0x1000);
+		MM_FreeTemp(tmpmapping);
+	}
+	
+//	MAGIC_BREAK();
+
+	// #7 Return
 	TMPCR3() = 0;
-	INVLPG(&TMPMAPLVL4(0));
+	INVLPG_ALL();
 	Mutex_Release(&glMM_TempFractalLock);
-	return 0;
+//	Log("MM_Clone: RETURN %P", ret);
+	return ret;
 }
 
 void MM_ClearUser(void)
 {
 	tVAddr	addr = 0;
-	// #1 Traverse the structure < 2^47, Deref'ing all pages
-	// #2 Free tables/dirs/pdps once they have been cleared
+	 int	pml4, pdpt, pd, pt;
 	
-	for( addr = 0; addr < 0x800000000000; )
+	for( pml4 = 0; pml4 < 256; pml4 ++ )
 	{
-		if( PAGEMAPLVL4(addr >> PML4_SHIFT) & 1 )
+		// Catch an un-allocated PML4 entry
+		if( !(PAGEMAPLVL4(pml4) & 1) ) {
+			addr += 1ULL << PML4_SHIFT;
+			continue ;
+		}
+		
+		// Catch a large COW
+		if( (PAGEMAPLVL4(pml4) & PF_COW) ) {
+			addr += 1ULL << PML4_SHIFT;
+		}
+		else
 		{
-			if( PAGEDIRPTR(addr >> PDP_SHIFT) & 1 )
+			// TODO: Large pages
+			
+			// Child entries
+			for( pdpt = 0; pdpt < 512; pdpt ++ )
 			{
-				if( PAGEDIR(addr >> PDIR_SHIFT) & 1 )
-				{
-					// Page
-					if( PAGETABLE(addr >> PTAB_SHIFT) & 1 ) {
-						MM_DerefPhys( PAGETABLE(addr >> PTAB_SHIFT) & PADDR_MASK );
-						PAGETABLE(addr >> PTAB_SHIFT) = 0;
-					}
-					addr += 1 << PTAB_SHIFT;
-					// Dereference the PDIR Entry
-					if( (addr + (1 << PTAB_SHIFT)) >> PDIR_SHIFT != (addr >> PDIR_SHIFT) ) {
-						MM_DerefPhys( PAGEMAPLVL4(addr >> PDIR_SHIFT) & PADDR_MASK );
-						PAGEDIR(addr >> PDIR_SHIFT) = 0;
-					}
-				}
-				else {
-					addr += 1 << PDIR_SHIFT;
+				// Unallocated
+				if( !(PAGEDIRPTR(addr >> PDP_SHIFT) & 1) ) {
+					addr += 1ULL << PDP_SHIFT;
 					continue;
 				}
-				// Dereference the PDP Entry
-				if( (addr + (1 << PDIR_SHIFT)) >> PDP_SHIFT != (addr >> PDP_SHIFT) ) {
-					MM_DerefPhys( PAGEMAPLVL4(addr >> PDP_SHIFT) & PADDR_MASK );
-					PAGEDIRPTR(addr >> PDP_SHIFT) = 0;
+			
+				// Catch a large COW
+				if( (PAGEDIRPTR(addr >> PDP_SHIFT) & PF_COW) ) {
+					addr += 1ULL << PDP_SHIFT;
 				}
-			}
-			else {
-				addr += 1 << PDP_SHIFT;
-				continue;
-			}
-			// Dereference the PML4 Entry
-			if( (addr + (1 << PDP_SHIFT)) >> PML4_SHIFT != (addr >> PML4_SHIFT) ) {
-				MM_DerefPhys( PAGEMAPLVL4(addr >> PML4_SHIFT) & PADDR_MASK );
-				PAGEMAPLVL4(addr >> PML4_SHIFT) = 0;
+				else {
+					// Child entries
+					for( pd = 0; pd < 512; pd ++ )
+					{
+						// Unallocated PDir entry
+						if( !(PAGEDIR(addr >> PDIR_SHIFT) & 1) ) {
+							addr += 1ULL << PDIR_SHIFT;
+							continue;
+						}
+						
+						// COW Page Table
+						if( PAGEDIR(addr >> PDIR_SHIFT) & PF_COW ) {
+							addr += 1ULL << PDIR_SHIFT;
+						}
+						else
+						{
+							// TODO: Catch large pages
+							
+							// Child entries
+							for( pt = 0; pt < 512; pt ++ )
+							{
+								// Free page
+								if( PAGETABLE(addr >> PTAB_SHIFT) & 1 ) {
+									MM_DerefPhys( PAGETABLE(addr >> PTAB_SHIFT) & PADDR_MASK );
+									PAGETABLE(addr >> PTAB_SHIFT) = 0;
+								}
+								addr += 1ULL << 12;
+							}
+						}
+						// Free page table
+						MM_DerefPhys( PAGEDIR(addr >> PDIR_SHIFT) & PADDR_MASK );
+						PAGEDIR(addr >> PDIR_SHIFT) = 0;
+					}
+				}
+				// Free page directory
+				MM_DerefPhys( PAGEDIRPTR(addr >> PDP_SHIFT) & PADDR_MASK );
+				PAGEDIRPTR(addr >> PDP_SHIFT) = 0;
 			}
 		}
-		else {
-			addr += (tVAddr)1 << PML4_SHIFT;
-			continue;
-		}
+		// Free page directory pointer table (PML4 entry)
+		MM_DerefPhys( PAGEMAPLVL4(pml4) & PADDR_MASK );
+		PAGEMAPLVL4(pml4) = 0;
 	}
 }
 
-tVAddr MM_NewWorkerStack(void)
+tVAddr MM_NewWorkerStack(void *StackData, size_t StackSize)
 {
 	tVAddr	ret;
 	 int	i;
 	
-//	Log_KernelPanic("MM", "TODO: Implement MM_NewWorkerStack");
-	
 	// #1 Set temp fractal to PID0
 	Mutex_Acquire(&glMM_TempFractalLock);
 	TMPCR3() = ((tPAddr)gInitialPML4 - KERNEL_BASE) | 3;
@@ -729,7 +925,9 @@ tVAddr MM_NewWorkerStack(void)
 	// #2 Scan for a free stack addresss < 2^47
 	for(ret = 0x100000; ret < (1ULL << 47); ret += KERNEL_STACK_SIZE)
 	{
-		if( MM_GetPhysAddr(ret) == 0 )	break;
+		tPAddr	*ptr;
+		if( MM_GetPageEntryPtr(ret, 1, 0, 0, &ptr) <= 0 )	break;
+		if( !(*ptr & 1) )	break;
 	}
 	if( ret >= (1ULL << 47) ) {
 		Mutex_Release(&glMM_TempFractalLock);
@@ -748,6 +946,19 @@ tVAddr MM_NewWorkerStack(void)
 		}
 		MM_MapEx(ret + i*0x1000, phys, 1, 0);
 	}
+
+	if( StackSize > 0x1000 ) {
+		Log_Error("MM", "MM_NewWorkerStack: StackSize(0x%x) > 0x1000, cbf handling", StackSize);
+	}
+	else {
+		tPAddr	*ptr, paddr;
+		tVAddr	tmp_addr;
+		MM_GetPageEntryPtr(ret + i*0x1000, 1, 0, 0, &ptr);
+		paddr = *ptr & ~0xFFF;
+		tmp_addr = MM_MapTemp(paddr);
+		memcpy( (void*)(tmp_addr + (0x1000 - StackSize)), StackData, StackSize );
+		MM_FreeTemp(tmp_addr);
+	}
 	
 	Mutex_Release(&glMM_TempFractalLock);
 	
@@ -763,11 +974,11 @@ tVAddr MM_NewKStack(void)
 	Uint	i;
 	for( ; base < MM_KSTACK_TOP; base += KERNEL_STACK_SIZE )
 	{
-		if(MM_GetPhysAddr(base) != 0)
+		if(MM_GetPhysAddr(base+KERNEL_STACK_SIZE-0x1000) != 0)
 			continue;
 		
 		//Log("MM_NewKStack: Found one at %p", base + KERNEL_STACK_SIZE);
-		for( i = 0; i < KERNEL_STACK_SIZE; i += 0x1000)
+		for( i = 0x1000; i < KERNEL_STACK_SIZE; i += 0x1000)
 		{
 			if( !MM_Allocate(base+i) )
 			{