vxd_svga.c split

5 files changed, 1597 insertions, 1429 deletions

diff --git a/makefile b/makefile
index 20cf9b3..028cf83 100644
--- a/makefile
+++ b/makefile
@@ -8,7 +8,7 @@ OBJS += &
   dbgprint32.obj svga.obj pci.obj vxd_fbhda.obj vxd_lib.obj vxd_main.obj &

   vxd_main_qemu.obj vxd_main_svga.obj vxd_svga.obj vxd_vdd.obj vxd_vdd_qemu.obj &

   vxd_vdd_svga.obj vxd_vbe.obj vxd_vbe_qemu.obj vxd_mouse.obj vxd_svga_st.obj &

-  vxd_mouse_svga.obj vxd_svga_mouse.obj

+  vxd_mouse_svga.obj vxd_svga_mouse.obj vxd_svga_mem.obj vxd_svga_cb.obj

 

 INCS = -I$(%WATCOM)\h\win -Iddk -Ivmware

 

@@ -170,6 +170,12 @@ vxd_svga_st.obj : vxd_svga_st.c .autodepend
 vxd_svga_mouse.obj : vxd_svga_mouse.c .autodepend

 	$(CC32) $(CFLAGS32) $(INCS) $(FLAGS) $<

 

+vxd_svga_mem.obj : vxd_svga_mem.c .autodepend

+	$(CC32) $(CFLAGS32) $(INCS) $(FLAGS) $<

+

+vxd_svga_cb.obj : vxd_svga_cb.c .autodepend

+	$(CC32) $(CFLAGS32) $(INCS) $(FLAGS) $<

+

 vxd_vbe.obj : vxd_vbe.c .autodepend

 	$(CC32) $(CFLAGS32) $(INCS) $(FLAGS) $<

 

@@ -434,6 +440,8 @@ file vxd_lib.obj
 file vxd_svga.obj

 file vxd_svga_st.obj

 file vxd_svga_mouse.obj

+file vxd_svga_mem.obj

+file vxd_svga_cb.obj

 file vxd_vdd_svga.obj

 file vxd_mouse_svga.obj

 segment '_LTEXT' PRELOAD NONDISCARDABLE 

diff --git a/vxd_svga.c b/vxd_svga.c
index 1a9cb34..73b8bda 100644
--- a/vxd_svga.c
+++ b/vxd_svga.c
@@ -48,24 +48,7 @@ THE SOFTWARE.
 /*

  * consts

  */

-//#define SVGA3D_MAX_MOBS 33280UL /* SVGA3D_MAX_CONTEXT_IDS + SVGA3D_MAX_CONTEXT_IDS + SVGA3D_MAX_SURFACE_IDS */

 #define SVGA3D_MAX_MOBS (SVGA3D_MAX_CONTEXT_IDS + SVGA3D_MAX_CONTEXT_IDS + SVGA3D_MAX_SURFACE_IDS)

-#define VBOX_VIDEO_MAX_SCREENS 64

-

-#define PTONPAGE (P_SIZE/sizeof(DWORD))

-

-#define SPARE_REGIONS_LARGE 2

-#define SPARE_REGION_LARGE_SIZE (32*1024*1024)

-

-#define SPARE_REGIONS_MEDIUM 16

-#define SPARE_REGION_MEDIUM_SIZE (1024*1024)

-

-#define SPARE_REGIONS_SMALL 128

-//#define SPARE_REGION_SMALL_SIZE 4096

-

-#define SPARE_REGIONS_CNT (SPARE_REGIONS_LARGE+SPARE_REGIONS_MEDIUM+SPARE_REGIONS_SMALL)

-

-#define CACHE_THRESHOLD 128

 

 #define SVGA_MIN_VRAM 8

 

@@ -73,66 +56,6 @@ THE SOFTWARE.
 

 BOOL CB_queue_check(SVGACBHeader *tracked);

 

-#define WAIT_FOR_CB(_cb, _forcesync) \

-	do{ \

-		/*dbg_printf(dbg_wait_cb, __LINE__);*/ \

-		while(!CB_queue_check(_cb)){ \

-			WAIT_FOR_CB_SYNC_ ## _forcesync \

-		} \

-	}while(0)

-

-/* expansions of WAIT_FOR_CB */

-#define WAIT_FOR_CB_SYNC_0

-#define WAIT_FOR_CB_SYNC_1 SVGA_Sync();

-

-static void SVGA_CB_start();

-static void SVGA_CB_restart();

-

-/*

- * types

- */

-

-#pragma pack(push)

-#pragma pack(1)

-typedef struct _cb_enable_t

-{

-	uint32             cmd;

-	SVGADCCmdStartStop cbstart;

-} cb_enable_t;

-

-typedef struct _cb_queue_t

-{

-	struct _cb_queue_t *next;

-	DWORD  flags;

-	DWORD  data_size;

-	DWORD  pad[13];

-} cb_queue_t;

-

-#pragma pack(pop)

-

-typedef struct _cb_queue_info_t

-{

-	cb_queue_t *first;

-	cb_queue_t *last;

-} cb_queue_info_t;

-

-typedef struct spare_region

-{

-	BOOL      used;

-	uint32    missed;

-	SVGA_region_info_t region;

-} spare_region_t;

-

-typedef struct svga_cache_state

-{

-	int free_index_min;

-	int free_index_max;

-	int cnt_large;

-	int cnt_medium;

-	int cnt_small;

-	spare_region_t spare_region[SPARE_REGIONS_CNT];

-} svga_cache_state_t;

-

 /*

  * globals

  */

@@ -142,44 +65,23 @@ static SVGA_DB_t *svga_db = NULL;
 

 extern LONG fb_lock_cnt;

 

-static BOOL gb_support = FALSE;

-static BOOL cb_support = FALSE;

-static BOOL cb_context0 = FALSE;

+       BOOL gb_support = FALSE;

+       BOOL cb_support = FALSE;

+       BOOL cb_context0 = FALSE;

 

 /* for GPU9 is FIFO more stable (VMWARE) or faster (VBOX) */

 static DWORD prefer_fifo = 1;

 

 static BOOL SVGA_is_valid = FALSE;

 

-static uint64 cb_next_id = {0, 0};

 static DWORD fence_next_id = 1;

 void *cmdbuf = NULL;

 void *ctlbuf = NULL;

-static DWORD cb_sem;

-

-static cb_queue_info_t cb_queue_info = {NULL, NULL};

-

-static svga_cache_state_t cache_state = {0, 0};

-

-static BOOL cache_enabled = FALSE;

 

 /* guest frame buffer is dirty */

-static BOOL ST_FB_invalid = FALSE;

-

-static DWORD present_fence = 0;

-

-/* object table for GPU10 */

-static SVGA_OT_info_entry_t otable_setup[SVGA_OTABLE_DX_MAX] = {

-	{0, NULL, RoundTo4k(VMWGFX_NUM_MOB*sizeof(SVGAOTableMobEntry)),                 0, 0}, /* SVGA_OTABLE_MOB */

-	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SURFACE*sizeof(SVGAOTableSurfaceEntry)),      0, 0}, /* SVGA_OTABLE_SURFACE */

-	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_CONTEXT*sizeof(SVGAOTableContextEntry)),      0, 0}, /* SVGA_OTABLE_CONTEXT */

-	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SHADER *sizeof(SVGAOTableShaderEntry)),       0, 0}, /* SVGA_OTABLE_SHADER - not used */

-	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SCREEN_TARGET*sizeof(SVGAOTableScreenTargetEntry)), 0, 0}, /* SVGA_OTABLE_SCREENTARGET (VBOX_VIDEO_MAX_SCREENS) */

-	{0, NULL, RoundTo4k(VMWGFX_NUM_DXCONTEXT*sizeof(SVGAOTableDXContextEntry)),    0, 0}, /* SVGA_OTABLE_DXCONTEXT */

-};

-

-static SVGA_OT_info_entry_t *otable = NULL;

+BOOL ST_FB_invalid = FALSE;

 

+DWORD present_fence = 0;

 /*

  * strings

  */

@@ -197,249 +99,14 @@ static char SVGA_conf_st_size[]    = "STSize";
 static char SVGA_conf_st_flags[]   = "STOptions";

 static char SVGA_vxd_name[]        = "vmwsmini.vxd";

 

-/* VM handle */

-extern DWORD ThisVM;

-

-/* prototypes */

-DWORD SVGA_GetDevCap(DWORD search_id);

-

-/**

- * Return PPN (physical page number) from virtual address

- * 

- **/

-static DWORD getPPN(DWORD virtualaddr)

-{

-	DWORD phy = 0;

-	_CopyPageTable(virtualaddr/P_SIZE, 1, &phy, 0);

-	return phy/P_SIZE;

-}

-

-static DWORD getPtrPPN(void *ptr)

-{

-	return getPPN((DWORD)ptr);

-}

-

-/**

- * Delete item from cache

- **/

-static void cache_delete(int index)

-{

-	if(cache_state.spare_region[index].used != FALSE)

-	{

-		SVGA_region_info_t *rinfo = &cache_state.spare_region[index].region;

-		BYTE *free_ptr = (BYTE*)rinfo->address;

-		uint32 region_size = rinfo->size;;

-		

-		if(rinfo->region_address != NULL)

-		{

-			dbg_printf(dbg_pagefree, rinfo->region_address);

-			_PageFree((PVOID)rinfo->region_address, 0);

-		}

-		else

-		{

-			free_ptr -= P_SIZE;

-		}

-			

-		if(rinfo->mob_address != NULL)

-		{

-			dbg_printf(dbg_pagefree, rinfo->mob_address);

-			_PageFree((PVOID)rinfo->mob_address, 0);

-		}

-			

-		dbg_printf(dbg_pagefree, free_ptr);

-		_PageFree((PVOID)free_ptr, 0);

-			

-		cache_state.spare_region[index].used = FALSE;

-		

-		if((index+1) == cache_state.free_index_max)

-		{

-			cache_state.free_index_max--;

-		}

-		

-		if(index < cache_state.free_index_min)

-		{

-			cache_state.free_index_min = index;

-		}

-		

-		if(region_size >= SPARE_REGION_LARGE_SIZE)

-			cache_state.cnt_large--;

-		else if(region_size >= SPARE_REGION_MEDIUM_SIZE)

-			cache_state.cnt_medium--;

-		else

-			cache_state.cnt_small--;

-	}

-}

-

-/**

- * Insert region to cache

- **/

-static BOOL cache_insert(SVGA_region_info_t *region)

-{

-	int i;

-	BOOL rc = FALSE;

-	

-	if(!cache_enabled)

-	{

-		return FALSE;

-	}

-	

-	if(region->size >= SPARE_REGION_LARGE_SIZE)

-	{

-		if(cache_state.cnt_large > SPARE_REGIONS_LARGE) return FALSE;

-	}

-	else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

-	{

-		if(cache_state.cnt_medium > SPARE_REGIONS_MEDIUM) return FALSE;

-	}

-	else

-	{

-		if(cache_state.cnt_small > SPARE_REGIONS_SMALL) return FALSE;

-	}

-	

-	for(i = cache_state.free_index_min; i < SPARE_REGIONS_CNT; i++)

-	{		

-		if(!cache_state.spare_region[i].used)

-		{

-			memcpy(&cache_state.spare_region[i].region, region, sizeof(SVGA_region_info_t));

-			cache_state.spare_region[i].used = TRUE;

-			cache_state.spare_region[i].missed = 0;

-			rc = TRUE;

-			break;

-		}

-	}

-	

-	for(; i < SPARE_REGIONS_CNT; i++)

-	{

-		if(!cache_state.spare_region[i].used)

-		{

-			cache_state.free_index_min = i;

-			break;

-		}

-	}

-	

-	if(i == SPARE_REGIONS_CNT)

-	{

-		cache_state.free_index_min = i;

-		cache_state.free_index_max = SPARE_REGIONS_CNT;

-	}

-	else if(i >= cache_state.free_index_max)

-	{

-		cache_state.free_index_max = i+1;

-	}

-

-	if(rc)

-	{

-		if(region->size >= SPARE_REGION_LARGE_SIZE)

-			cache_state.cnt_large++;

-		else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

-			cache_state.cnt_medium++;

-		else

-			cache_state.cnt_small++;

-

-		dbg_printf(dbg_cache_insert, region->region_id, region->size);

-	}

-

-	return rc;

-}

-

-/**

- * Use region from cache

- **/

-static BOOL cache_use(SVGA_region_info_t *region)

-{

-	int i;

-	BOOL rc = FALSE;

-	

-	if(!cache_enabled)

-	{

-		return FALSE;

-	}

-	

-	dbg_printf(dbg_cache_search, region->size);

-	

-	for(i = cache_state.free_index_max-1; i >= 0; i--)

-	{

-		if(cache_state.spare_region[i].used != FALSE)

-		{

-			if(cache_state.spare_region[i].region.size == region->size)

-			{

-				SVGA_region_info_t *ptr = &cache_state.spare_region[i].region;

-				

-				region->address        = ptr->address;

-				region->region_address = ptr->region_address;

-				region->region_ppn     = ptr->region_ppn;

-				region->mob_address    = ptr->mob_address;

-				region->mob_ppn        = ptr->mob_ppn;

-				

-				cache_state.spare_region[i].used = FALSE;

-				rc = TRUE;

-				break;

-			}

-			else

-			{

-				if(++cache_state.spare_region[i].missed >= CACHE_THRESHOLD)

-				{

-					dbg_printf(dbg_cache_delete, cache_state.spare_region[i].region.size);

-					cache_delete(i);

-				}

-			}

-		}

-	}

-	

-	if(rc)

-	{

-		if((i+1) == cache_state.free_index_max)

-		{

-			cache_state.free_index_max--;

-		}

-		

-		if(i < cache_state.free_index_min)

-		{

-			cache_state.free_index_min = i;

-		}

-		

-		if(region->size >= SPARE_REGION_LARGE_SIZE)

-			cache_state.cnt_large--;

-		else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

-			cache_state.cnt_medium--;

-		else

-			cache_state.cnt_small--;

-	

-		dbg_printf(dbg_cache_used, region->region_id, region->size);

-	}

-		

-	return rc;

-}

-

 /**

  * Notify virtual HW that is some work to do

  **/

-static void SVGA_Sync()

+void SVGA_Sync()

 {

 	SVGA_WriteReg(SVGA_REG_SYNC, 1);

 }

 

-static void SVGA_Flush_CB_critical()

-{

-	/* wait for actual CB */

-	while(!CB_queue_check(NULL))

-	{

-		SVGA_Sync();

-	}

-	

-	/* drain FIFO */

-	SVGA_Flush();

-}

-

-#if 0

-static void SVGA_Flush_CB()

-{

-	Begin_Critical_Section(0);

-	SVGA_Flush_CB_critical();

-	End_Critical_Section();

-}

-#endif

-

 /* map physical addresses to system linear space */

 void SVGA_MapIO()

 {

@@ -450,132 +117,6 @@ void SVGA_MapIO()
 	}

 }

 

-static DWORD PT_count(DWORD size)

-{

-	DWORD pt1_entries = 0;

-	DWORD pt2_entries = 0;

-

-	pt1_entries = RoundToPages(size);

-	pt2_entries = ((pt1_entries + PTONPAGE - 1)/PTONPAGE);

-	

-	dbg_printf(dbg_pt_build_2, size, pt1_entries, pt2_entries);

-	

-	if(pt2_entries > 1)

-	{

-		return pt2_entries + 1;

-	}

-	

-	return 1;

-}

-

-static void PT_build(DWORD size, void *buf, DWORD *outBase, DWORD *outType, void **outUserPtr)

-{

-	DWORD pt1_entries = 0;

-	DWORD pt2_entries = 0;

-	DWORD i;

-	

-	pt1_entries = RoundToPages(size);

-	pt2_entries = ((pt1_entries + PTONPAGE - 1)/PTONPAGE);

-

-	if(pt2_entries > 1)

-	{

-		DWORD *ptbuf = buf;

-		BYTE *ptr = ((BYTE*)buf)+((pt2_entries+1)*P_SIZE);

-		

-		memset(buf, 0, (pt2_entries+1)*P_SIZE);

-		

-		/* build PT2 */

-		for(i = 0; i < pt2_entries; i++)

-		{

-			ptbuf[i] = getPtrPPN(ptbuf + ((i+1)*PTONPAGE));

-		}

-		

-		ptbuf += PTONPAGE;

-		

-		/* build PT2 */

-		for(i = 0; i < pt1_entries; i++)

-		{

-			ptbuf[i] = getPtrPPN(ptr + i*P_SIZE);

-		}

-		

-		*outBase = getPtrPPN(buf);

-		*outType = SVGA3D_MOBFMT_PTDEPTH_2;

-		*outUserPtr = (void*)ptr;

-	}

-	else

-	{

-		DWORD *ptbuf = buf;

-		BYTE *ptr = ((BYTE*)buf)+P_SIZE;

-		

-		memset(buf, 0, P_SIZE);

-		

-		for(i = 0; i < pt1_entries; i++)

-		{

-			ptbuf[i] = getPtrPPN(ptr + i*P_SIZE);

-		}

-		

-		*outBase = getPtrPPN(buf);

-		*outType = SVGA3D_MOBFMT_PTDEPTH_1;

-		*outUserPtr = (void*)ptr;

-	}

-	

-	dbg_printf(dbg_pt_build, size, *outBase, *outType, *outUserPtr);

-}

-

-/**

- * Allocate OTable for GB objects

- **/

-static void SVGA_OTable_alloc(BOOL screentargets)

-{

-	int i;

-	

-	if(otable == NULL)

-	{

-		otable = (SVGA_OT_info_entry_t *)_PageAllocate(RoundToPages(sizeof(otable_setup)), PG_VM, ThisVM, 0, 0x0, 0x100000, NULL, PAGEFIXED);

-		if(otable)

-		{

-			memcpy(otable, &(otable_setup[0]), sizeof(otable_setup));

-		}

-	}

-	

-	/* not allocate DX when there is no DX support */

-	if(!SVGA_GetDevCap(SVGA3D_DEVCAP_DXCONTEXT))

-	{

-		otable[SVGA_OTABLE_DXCONTEXT].size = 0;

-	}

-	

-/*	if(!screentargets)

-	{

-		otable[SVGA_OTABLE_SCREENTARGET].size = 0;

-	}*/

-

-	if(otable)

-	{

-		for(i = 0; i < SVGA_OTABLE_DX_MAX; i++)

-		{

-			SVGA_OT_info_entry_t *entry = &otable[i];

-			if(entry->size != 0 && (entry->flags & SVGA_OT_FLAG_ALLOCATED) == 0)

-			{

-				void *ptr = (void*)_PageAllocate(RoundToPages(entry->size)+PT_count(entry->size), PG_VM, ThisVM, 0, 0x0, 0x100000, NULL, PAGEFIXED);

-				

-				if(ptr)

-				{

-					dbg_printf(dbg_mob_allocate, i);

-					PT_build(entry->size, ptr, &entry->ppn, &entry->pt_depth, &entry->lin);

-					

-					entry->flags |= SVGA_OT_FLAG_ALLOCATED;

-					entry->flags |= SVGA_OT_FLAG_DIRTY;

-				}

-			}

-		}

-	}

-}

-

-SVGA_OT_info_entry_t *SVGA_OT_setup()

-{

-	return otable;

-}

-

 static char db_mutexname[] = "svga_db_mux";

 

 static void SVGA_DB_alloc()

@@ -651,7 +192,7 @@ SVGA_DB_t *SVGA_DB_setup()
 	return svga_db;

 }

 

-static DWORD SVGA_fence_passed()

+DWORD SVGA_fence_passed()

 {

 	if(SVGA_IsSVGA3())

 	{

@@ -693,7 +234,7 @@ void SVGA_fence_query(DWORD FBPTR ptr_fence_passed, DWORD FBPTR ptr_fence_last)
 	}

 }

 

-static BOOL SVGA_fence_is_passed(DWORD fence_id)

+BOOL SVGA_fence_is_passed(DWORD fence_id)

 {

 	DWORD last_pased;

 	DWORD last_fence;

@@ -735,58 +276,6 @@ void SVGA_fence_wait(DWORD fence_id)
 	}*/

 }

 

-/**

- * Allocate memory for command buffer

- **/

-static DWORD *SVGA_CMB_alloc_size(DWORD datasize)

-{

-	DWORD phy;

-	SVGACBHeader *cb;

-	cb_queue_t *q;

-	

-	Begin_Critical_Section(0);

-	

-	q = (cb_queue_t*)_PageAllocate(RoundToPages(datasize+sizeof(SVGACBHeader)+sizeof(cb_queue_t)), PG_SYS, 0, 0, 0x0, 0x100000, &phy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

-	

-	if(q)

-	{

-		q->next = NULL;

-		q->flags = 0;

-		q->data_size = 0;

-		

-		cb = (SVGACBHeader*)(q+1);

-		

-		memset(cb, 0, sizeof(SVGACBHeader));

-		cb->status = SVGA_CB_STATUS_COMPLETED; /* important to sync between commands */

-		cb->ptr.pa.hi   = 0;

-		cb->ptr.pa.low  = phy + sizeof(cb_queue_t) + sizeof(SVGACBHeader);	

-		

-		End_Critical_Section();

-		return (DWORD*)(cb+1);

-	}

-	End_Critical_Section();

-	

-	return NULL;

-}

-

-void SVGA_CMB_free(DWORD *cmb)

-{

-	SVGACBHeader *cb = ((SVGACBHeader *)cmb)-1;

-	

-	Begin_Critical_Section(0);

-	

-	WAIT_FOR_CB(cb, 1);

-		

-	_PageFree(cb-1, 0);

-	

-	End_Critical_Section();

-}

-

-DWORD *SVGA_CMB_alloc()

-{

-	return SVGA_CMB_alloc_size(SVGA_CB_MAX_SIZE);

-}

-

 void *SVGA_cmd_ptr(DWORD *buf, DWORD *pOffset, DWORD cmd, DWORD cmdsize)

 {

 	DWORD pp = (*pOffset)/sizeof(DWORD);

@@ -806,331 +295,6 @@ void *SVGA_cmd3d_ptr(DWORD *buf, DWORD *pOffset, DWORD cmd, DWORD cmdsize)
 	return (void*)(buf + pp + 2);

 }

 

-// FIXME: inline?

-static void SVGA_cb_id_inc()

-{

-	_asm

-	{

-		inc dword ptr [cb_next_id]

-		adc dword ptr [cb_next_id+4], 0

-	}

-}

-

-#ifdef DBGPRINT

-#include "vxd_svga_debug.h"

-#endif

-

-/*

- * @param tracked: check specific CB, or NULL to check full queue

- *

- * @return: TRUE if tracked is complete or TRUE id queue is empty

- 

- */

- 

-BOOL CB_queue_check(SVGACBHeader *tracked)

-{

-	cb_queue_t *last = NULL;

-	cb_queue_t *item = cb_queue_info.first;

-	BOOL in_queue = FALSE;

-	

-	while(item != NULL)

-	{

-		SVGACBHeader *cb = (SVGACBHeader*)(item+1);

-		if(cb->status >= SVGA_CB_STATUS_COMPLETED)

-		{

-			if(last == NULL)

-			{

-				cb_queue_info.first = item->next;

-			}

-			else

-			{

-				last->next = item->next;

-			}

-			

-			if(cb->status > SVGA_CB_STATUS_COMPLETED)

-			{

-				SVGA_CB_restart();

-			}

-			

-			item = item->next;

-		}

-		else

-		{

-			if(tracked == cb)

-			{

-				in_queue = TRUE;

-			}

-			last = item;

-			item = item->next;

-		}

-	}

-	

-	cb_queue_info.last = last;

-	

-	if(cb_queue_info.first == NULL)

-	{

-		return TRUE;

-	}

-	

-	if(tracked != NULL && in_queue == FALSE)

-	{

-		return TRUE;

-	}

-	

-	return FALSE;	

-}

-

-void CB_queue_insert(SVGACBHeader *cb)

-{

-	cb_queue_t *item = (cb_queue_t*)(cb-1);

-	item->next = NULL;

-	

-	if(cb_queue_info.last != NULL)

-	{

-		cb_queue_info.last->next = item;

-		cb_queue_info.last = item;

-	}

-	else

-	{

-		cb_queue_info.first = item;

-		cb_queue_info.last  = item;

-	}

-}

-

-

-void SVGA_CMB_submit_critical(DWORD FBPTR cmb, DWORD cmb_size, SVGA_CMB_status_t FBPTR status, DWORD flags, DWORD DXCtxId)

-{

-	DWORD fence = 0;

-	SVGACBHeader *cb = ((SVGACBHeader *)cmb)-1;

-	

-	CB_queue_check(NULL);

-	

-	if(status)

-	{

-		cb->status = SVGA_CB_STATUS_NONE;

-		status->sStatus = SVGA_PROC_NONE;

-		status->qStatus = (volatile DWORD*)&cb->status;

-	}

-

-#ifdef DBGPRINT

-	debug_cmdbuf(cmb, cmb_size);

-//	debug_cmdbuf_trace(cmb, cmb_size, SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN);

-//		debug_draw(cmb, cmb_size);

-#endif

-

-	if(flags & SVGA_CB_PRESENT_ASYNC)

-	{

-		if(present_fence != 0)

-		{

-			SVGA_fence_wait(present_fence);

-		}

-		

-		flags |= SVGA_CB_FORCE_FENCE;

-	}

-	else if(present_fence)

-	{

-		if(SVGA_fence_is_passed(present_fence))

-		{

-			present_fence = 0;

-		}

-	}

-	

-	if(flags & SVGA_CB_PRESENT_GPU)

-	{

-		ST_FB_invalid = TRUE;

-	}

-	

-	if(  /* CB are supported and enabled */

-		cb_support &&

-		(cb_context0 || (flags & SVGA_CB_USE_CONTEXT_DEVICE)) &&

-		(flags & SVGA_CB_FORCE_FIFO) == 0

-	)

-	{

-		DWORD cbhwctxid = SVGA_CB_CONTEXT_0;

-		

-		if(flags & SVGA_CB_USE_CONTEXT_DEVICE)

-		{

-			cbhwctxid = SVGA_CB_CONTEXT_DEVICE;

-		}

-		

-		if(flags & SVGA_CB_FORCE_FENCE)

-		{

-			DWORD dwords = cmb_size/sizeof(DWORD);

-			fence = SVGA_fence_get();

-			cmb[dwords]   = SVGA_CMD_FENCE;

-			cmb[dwords+1] = fence;

-			cmb_size += sizeof(DWORD)*2;

-		}

-		

-		if(cmb_size == 0)

-		{

-			cb->status = SVGA_PROC_COMPLETED;

-			if(status)

-			{

-				status->sStatus = SVGA_PROC_COMPLETED;

-				status->qStatus = NULL;

-				status->fifo_fence_used = 0;

-			}

-		}

-		else

-		{

-			cb->status = SVGA_CB_STATUS_NONE;

-			cb->errorOffset = 0;

-			cb->offset = 0; /* VMware modified this, needs to be clear */

-			cb->flags  = SVGA_CB_FLAG_NO_IRQ;

-			

-			if(flags & SVGA_CB_FLAG_DX_CONTEXT)

-			{

-				cb->flags |= SVGA_CB_FLAG_DX_CONTEXT;

-				cb->dxContext = DXCtxId;

-			}

-			else

-			{

-				cb->dxContext = 0;

-			}

-			

-			cb->id.low = cb_next_id.low;

-			cb->id.hi  = cb_next_id.hi;

-			cb->length = cmb_size;

-			

-			CB_queue_insert(cb);			

-			

-			SVGA_WriteReg(SVGA_REG_COMMAND_HIGH, 0); // high part of 64-bit memory address...

-			SVGA_WriteReg(SVGA_REG_COMMAND_LOW, (cb->ptr.pa.low - sizeof(SVGACBHeader)) | cbhwctxid);

-			SVGA_Sync(); /* notify HV to read registers (VMware needs it) */

-			

-			SVGA_cb_id_inc();	

-			

-			if(flags & SVGA_CB_SYNC)

-			{

-				WAIT_FOR_CB(cb, 0);

-				

-				if(cb->status != SVGA_CB_STATUS_COMPLETED)

-				{

-					dbg_printf(dbg_cmd_error, cb->status, cmb[0], cb->errorOffset);

-					if(flags & SVGA_CB_FORCE_FENCE)

-					{

-						/* this may cause freeze, when fence is in buffer and isn't complete. So return some passed fence  */

-						fence = SVGA_fence_passed();

-					}

-				}

-				else

-				{

-					dbg_printf(dbg_cb_suc);

-				}

-				

-				if(status)

-				{

-					status->sStatus = SVGA_PROC_COMPLETED;

-					status->qStatus = NULL;

-					status->fifo_fence_used = fence;

-				}

-			}

-			else

-			{

-				if(status)

-				{

-					status->sStatus = SVGA_PROC_NONE;

-					status->qStatus = (volatile DWORD*)&cb->status;

-					status->fifo_fence_used = fence;

-				}

-			}

-		}	

-	}

-	else /* copy to FIFO */

-	{

-		DWORD *ptr = cmb;

-		DWORD dwords = cmb_size/sizeof(DWORD);

-		DWORD nextCmd, max, min;

-		

-		/* insert fence CMD */

-		fence = SVGA_fence_get();

-		ptr[dwords]   = SVGA_CMD_FENCE;

-		ptr[dwords+1] = fence;

-		

-		dwords += 2;

-		

-		//SVGA_Flush();

-		

-		nextCmd = gSVGA.fifoMem[SVGA_FIFO_NEXT_CMD];

-		max     = gSVGA.fifoMem[SVGA_FIFO_MAX];

-		min     = gSVGA.fifoMem[SVGA_FIFO_MIN];

-		

-		/* copy to fifo */

-		while(dwords > 0)

-		{

-			gSVGA.fifoMem[nextCmd/sizeof(DWORD)] = *ptr;

-			ptr++;

-			

-			nextCmd += sizeof(uint32);

-			if (nextCmd >= max)

-			{

-				nextCmd = min;

-			}

-			gSVGA.fifoMem[SVGA_FIFO_NEXT_CMD] = nextCmd;

-			dwords--;

-		}

-		

-		//SVGA_Sync();

-		

-		if(flags & SVGA_CB_SYNC)

-		{

-			SVGA_fence_wait(fence);

-			if(status)

-			{

-				status->sStatus = SVGA_PROC_COMPLETED;

-				status->qStatus = NULL;

-				status->fifo_fence_used = 0;

-			}

-		}

-		else

-		{

-			if(status)

-			{

-				status->sStatus = SVGA_PROC_FENCE;

-				status->qStatus = NULL;

-				status->fifo_fence_used = fence;

-			}

-		}

-		

-		cb->status = SVGA_PROC_COMPLETED;

-	} /* FIFO */

-	

-	if(status)

-	{

-		status->fifo_fence_last = SVGA_fence_passed();

-	}

-	

-	if(flags & SVGA_CB_PRESENT_ASYNC)

-	{

-		present_fence = fence;

-	}

-	

-	//dbg_printf(dbg_cmd_off, cmb[0]);

-}

-

-void wait_for_cmdbuf()

-{

-	SVGACBHeader *cb;

-	

-	Begin_Critical_Section(0);

-	cb = ((SVGACBHeader *)cmdbuf)-1;

-	WAIT_FOR_CB(cb, 0);

-}

-

-void submit_cmdbuf(DWORD cmdsize, DWORD flags, DWORD dx)

-{

-	SVGA_CMB_submit_critical(cmdbuf, cmdsize, NULL, flags, dx);

-	End_Critical_Section();

-}

-

-void SVGA_CMB_submit(DWORD FBPTR cmb, DWORD cmb_size, SVGA_CMB_status_t FBPTR status, DWORD flags, DWORD DXCtxId)

-{

-	Begin_Critical_Section(0);

-	SVGA_CMB_submit_critical(cmb, cmb_size, status, flags, DXCtxId);

-	End_Critical_Section();

-}

-

 void SVGA_IRQ_proc()

 {

 	dbg_printf(dbg_irq);

@@ -1341,7 +505,7 @@ BOOL SVGA_init_hw()
 			}

 		}

 		

-		cb_sem = Create_Semaphore(1);

+		//cb_sem = Create_Semaphore(1);

 		

 		SVGA_DB_alloc();

 		

@@ -1397,16 +561,16 @@ BOOL SVGA_init_hw()
 		{

 			hda->flags |= FB_ACCEL_VMSVGA10;

 		}

-		

-		memset(&cache_state, 0, sizeof(svga_cache_state_t));

-		

+			

+		cache_init();

+				

 		SVGA_is_valid = TRUE;

 		

 		if(_GetFreePageCount(NULL) >= 0x38000UL) /* 896MB / 4096 */

 		{

 			if(!gb_support) /* mob cache is done by RING-3 DLL */

 			{

-				cache_enabled = TRUE;

+				cache_enable(TRUE);

 			}

 		}

 		

@@ -1421,493 +585,13 @@ BOOL SVGA_init_hw()
 			

 			SVGA_region_create(&testregion);

 		}*/

-		

-		dbg_printf(dbg_cache, cache_enabled);

-		

+				

 		return TRUE;

 	}

 	

 	return FALSE;

 }

 

-

-//#define GMR_CONTIG

-//#define GMR_SYSTEM

-

-//static DWORD pa_flags = PAGEFIXED | PAGEUSEALIGN;

-//static DWORD pa_align = 0x0000000F; // 64k

-

-static DWORD pa_flags = PAGEFIXED;

-static DWORD pa_align = 0x00000000;

-

-/**

- * Allocate guest memory region (GMR) - HW needs know memory physical

- * addressed of pages in (virtual) memory block.

- * Technically this allocate 2 memory block, 1st for data and 2nd as its

- * physical description.

- *

- * @return: TRUE on success

- *

- **/

-BOOL SVGA_region_create(SVGA_region_info_t *rinfo)

-{

-#ifdef GMR_CONTIG

-	ULONG phy = 0;

-#endif

-	ULONG laddr;

-	ULONG pgblk;

-	ULONG new_size = RoundTo4k(rinfo->size);	

-	ULONG nPages = RoundToPages(new_size);

-	//ULONG nPages = RoundToPages64k(rinfo->size);

-	ULONG pa_vm = ThisVM;

-	ULONG pa_type = PG_VM;

-	SVGAGuestMemDescriptor *desc;

-	ULONG size_total = 0;

-

-#ifdef GMR_SYSTEM

-		pa_vm = 0;

-		pa_type = PG_SYS;

-#endif

-	

-	rinfo->size = new_size;

-	

-	Begin_Critical_Section(0);

-

-	if(cache_use(rinfo))

-	{

-		goto spare_region_used;

-	}

-	

-	//dbg_printf(dbg_pages, rinfo->size, nPages, P_SIZE);

-#ifdef GMR_CONTIG

-	/* JH: OK, using PAGECONTIG leads to very fast memory exhaustion, so using only slower way! */

-	/* first try to allocate continuous physical memory space, 1st page is used for GMR descriptor */

-	laddr = _PageAllocate(nPages+1, pa_type, pa_vm, 0, 0x0, 0x100000, &phy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

-	

-	if(laddr)

-	{

-		desc = (SVGAGuestMemDescriptor*)laddr;

-		desc->ppn = (phy/P_SIZE)+1;

-		desc->numPages = nPages;

-		desc++;

-		desc->ppn = 0;

-		desc->numPages = 0;

-		

-		rinfo->address        = (void*)(laddr+P_SIZE);

-		rinfo->region_address = 0;

-		rinfo->region_ppn     = (phy/P_SIZE);

-		rinfo->mob_address    = 0;	

-		rinfo->mob_ppn        = (phy/P_SIZE)+1;

-		

-		//dbg_printf(dbg_region_simple, laddr, rinfo->address);

-		

-		size_total = (nPages+1)*P_SIZE;

-	}

-	else

-#endif

-	{

-		/* memory is too fragmented to create this large continous region */

-		ULONG taddr;

-		ULONG tppn;

-		ULONG pgi;

-		ULONG base_ppn;

-		ULONG base_cnt;

-		ULONG blocks = 1;

-		ULONG blk_pages = 0;

-		ULONG mob_pages = 0;

-		

-		DWORD *mob = NULL;

-		//DWORD mobphy = 0;

-		

-		/* allocate user block */

-		laddr = _PageAllocate(nPages, pa_type, pa_vm, pa_align, 0x0, 0x100000, NULL, pa_flags);

-		

-		if(!laddr)

-		{

-			End_Critical_Section();

-			return FALSE;

-		}

-		

-		/* determine how many physical continuous blocks we have */

-		base_ppn = getPPN(laddr);

-		base_cnt = 1;

-		for(pgi = 1; pgi < nPages; pgi++)

-		{

-			taddr = laddr + pgi*P_SIZE;

-			tppn = getPPN(taddr);

-			

-			if(tppn != base_ppn + base_cnt)

-			{

-				base_ppn = tppn;

-				base_cnt = 1;

-				blocks++;

-			}

-			else

-			{

-				base_cnt++;

-			}

-		}

-			

-		// number of pages to store regions information

-		blk_pages = ((blocks+1)/(P_SIZE/sizeof(SVGAGuestMemDescriptor))) + 1;

-			

-		// add extra descriptors for pages edge

-		blk_pages = ((blocks+blk_pages+1)/(P_SIZE/sizeof(SVGAGuestMemDescriptor))) + 1;

-		

-		/* allocate memory for GMR descriptor */

-		pgblk = _PageAllocate(blk_pages, pa_type, pa_vm, pa_align, 0x0, 0x100000, NULL, pa_flags);

-		if(!pgblk)

-		{

-			_PageFree((PVOID)laddr, 0);

-			End_Critical_Section();

-			return FALSE;

-		}

-		

-		desc = (SVGAGuestMemDescriptor*)pgblk;

-		desc->ppn = getPPN(laddr);

-		desc->numPages = 1;

-		blocks = 1;

-		

-		/* fill GMR physical structure */

-		for(pgi = 1; pgi < nPages; pgi++)

-		{

-			taddr = laddr + pgi*P_SIZE;

-			tppn = getPPN(taddr);

-					

-			if(tppn == desc->ppn + desc->numPages)

-			{

-				desc->numPages++;

-			}

-			else

-			{

-				/* next descriptor is on page edge */

-				if((blocks+1) % (P_SIZE/sizeof(SVGAGuestMemDescriptor)) == 0)

-				{

-					desc++;

-					desc->numPages = 0;

-					desc->ppn = getPPN((DWORD)(desc+1));

-					blocks++;

-				}

-

-				desc++;

-				desc->ppn = tppn;

-				desc->numPages = 1;

-				blocks++;

-			}

-		}

-

-		desc++;

-		desc->ppn = 0;

-		desc->numPages = 0;

-

-		if(gb_support) /* don't create MOBs for Gen9 */

-		{

-			int i;

-			/* for simplicity we're always creating table of depth 2, first page is page of PPN pages */

-			//mob = (DWORD *)_PageAllocate(1 + (nPages + PTONPAGE - 1)/PTONPAGE, pa_type, pa_vm, 0, 0x0, 0x100000, &mobphy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

-			mob_pages = 1 + (nPages + PTONPAGE - 1)/PTONPAGE;

-			mob = (DWORD *)_PageAllocate(mob_pages, pa_type, pa_vm, pa_align, 0x0, 0x100000, 0, pa_flags);

-			

-			if(!mob)

-			{

-				_PageFree((PVOID)laddr, 0);

-				_PageFree((PVOID)pgblk, 0);

-				End_Critical_Section();

-				return FALSE;

-			}

-			

-			memset(mob, 0, mob_pages*P_SIZE);

-			

-			/* dim 1 */

-			for(i = 0; i < mob_pages-1; i++)

-			{

-				//mob[i] = (mobphy/P_SIZE) + i + 1;

-				mob[i] = getPPN(((ULONG)mob) + ((i + 1)*P_SIZE));

-			}

-			/* dim 2 */

-			for(i = 0; i < nPages; i++)

-			{

-				mob[PTONPAGE + i] = getPPN(laddr + i*P_SIZE);

-			}

-		}

-		

-		rinfo->address        = (void*)laddr;

-		rinfo->region_address = (void*)pgblk;

-		rinfo->region_ppn     = getPPN(pgblk);

-		rinfo->mob_address    = (void*)mob;

-		rinfo->mob_ppn        = 0;

-		if(mob)

-		{

-			//rinfo->mob_ppn        = mobphy/P_SIZE;

-			rinfo->mob_ppn        = getPPN((ULONG)mob);

-		}

-		

-		size_total = (nPages + blk_pages + mob_pages)*P_SIZE;

-		

-		//dbg_printf(dbg_region_fragmented);

-	}

-	

-	//dbg_printf(dbg_gmr, rinfo->region_id, rinfo->region_ppn);

-	

-	spare_region_used:

-	//dbg_printf(dbg_mobonly, rinfo->mobonly);

-	

-	// JH: no need here

-	//SVGA_Flush_CB();

-	

-	if(!rinfo->mobonly)

-	{

-		if(rinfo->region_id < SVGA_ReadReg(SVGA_REG_GMR_MAX_IDS))

-		{

-			/* register GMR */

-			SVGA_WriteReg(SVGA_REG_GMR_ID, rinfo->region_id);

-			SVGA_WriteReg(SVGA_REG_GMR_DESCRIPTOR, rinfo->region_ppn);

-			SVGA_Sync(); // notify register change

-			

-			SVGA_Flush_CB_critical();

-		}

-		else

-		{

-			/* GMR is too high, use it as only as MOB (DX) */

-			rinfo->region_ppn = 0;

-		}

-	}

-

-	if(gb_support)

-	{

-		SVGA3dCmdDefineGBMob *mob;

-		DWORD cmdoff = 0;

-		

-		wait_for_cmdbuf();

-  	mob              = SVGA_cmd3d_ptr(cmdbuf, &cmdoff, SVGA_3D_CMD_DEFINE_GB_MOB, sizeof(SVGA3dCmdDefineGBMob));

-  	mob->mobid       = rinfo->region_id;

-  	mob->base        = rinfo->mob_ppn;

-  	mob->sizeInBytes = rinfo->size;

-		if(rinfo->mob_address == NULL) /* continual memory space */

-		{

-			mob->ptDepth     = SVGA3D_MOBFMT_RANGE;

-		}

-		else

-		{

-			mob->ptDepth     = SVGA3D_MOBFMT_PTDEPTH_2;

-		}

-		

-		submit_cmdbuf(cmdoff, SVGA_CB_SYNC, 0);

-		

-  	rinfo->is_mob = 1;

-	}

-	else

-	{

-		rinfo->is_mob = 0;

-	}

-	

-	svga_db->stat_regions_usage += rinfo->size;

-	

-	End_Critical_Section();

-	

-	//dbg_printf(dbg_gmr_succ, rinfo->region_id, rinfo->size);

-	

-	return TRUE;

-}

-

-/**

- * Free data allocated by SVGA_region_create

- *

- **/

-void SVGA_region_free(SVGA_region_info_t *rinfo)

-{

-	BOOL saved_in_cache;

-	BYTE *free_ptr = (BYTE*)rinfo->address;

-	

-	Begin_Critical_Section(0);

-	svga_db->stat_regions_usage -= rinfo->size;

-	

-	saved_in_cache = cache_insert(rinfo);

-	

-	if(gb_support)

-	{

-		SVGA3dCmdDestroyGBMob *mob;

-		DWORD cmdoff = 0;

-		

-  	wait_for_cmdbuf();

-  	mob              = SVGA_cmd3d_ptr(cmdbuf, &cmdoff, SVGA_3D_CMD_DESTROY_GB_MOB, sizeof(SVGA3dCmdDestroyGBMob));

-  	mob->mobid       = rinfo->region_id;

-  	

-  	if(saved_in_cache)

-  		submit_cmdbuf(cmdoff, 0, 0);

-  	else

-			submit_cmdbuf(cmdoff, SVGA_CB_SYNC, 0);

-	}

-	

-	if(!rinfo->mobonly)

-	{

-		SVGA_Sync();

-		SVGA_Flush_CB_critical();

-		

-		SVGA_WriteReg(SVGA_REG_GMR_ID, rinfo->region_id);

-		SVGA_WriteReg(SVGA_REG_GMR_DESCRIPTOR, 0);

-		SVGA_Sync(); // notify register change

-	}

-

-	if(!saved_in_cache)

-	{

-		if(rinfo->region_address != NULL)

-		{

-			//dbg_printf(dbg_pagefree, rinfo->region_address);

-			_PageFree((PVOID)rinfo->region_address, 0);

-		}

-		else

-		{

-			free_ptr -= P_SIZE;

-		}

-			

-		if(rinfo->mob_address != NULL)

-		{

-			//dbg_printf(dbg_pagefree, rinfo->mob_address);

-			_PageFree((PVOID)rinfo->mob_address, 0);

-		}

-			

-		//dbg_printf(dbg_pagefree, free_ptr);

-		_PageFree((PVOID)free_ptr, 0);

-	}

-	End_Critical_Section();

-		

-	//dbg_printf(dbg_pagefree_end, rinfo->region_id, rinfo->size, saved_in_cache);

-		

-	rinfo->address        = NULL;

-	rinfo->region_address = NULL;

-	rinfo->mob_address    = NULL;

-	rinfo->region_ppn     = 0;

-	rinfo->mob_ppn        = 0;

-}

-

-/**

- * Destroy saved regions and free memory

- *

- **/

-void SVGA_flushcache()

-{

-	int i;

-	

-	Begin_Critical_Section(0);

-	

-	for(i = 0; i < cache_state.free_index_max ; i++)

-	{

-		cache_delete(i);

-	}

-	

-	cache_state.free_index_max = 0;

-	cache_state.free_index_min = 0;

-	cache_state.cnt_large      = 0;

-	cache_state.cnt_medium     = 0;

-	cache_state.cnt_small      = 0;

-

-	End_Critical_Section();

-}

-

-static DWORD SVGA_CB_ctr(DWORD data_size)

-{

-	SVGACBHeader *cb = ((SVGACBHeader *)ctlbuf)-1;

-	

-	dbg_printf(dbg_ctr_start);

-

-	//Begin_Critical_Section(0);

-

-	cb->status = SVGA_CB_STATUS_NONE;

-	cb->errorOffset = 0;

-	cb->offset = 0; /* VMware modified this, needs to be clear */

-	cb->flags  = SVGA_CB_FLAG_NO_IRQ;

-	cb->mustBeZero[0] = 0;

-	cb->mustBeZero[1] = 0;

-	cb->mustBeZero[2] = 0;

-	cb->mustBeZero[3] = 0;

-	cb->mustBeZero[4] = 0;

-	cb->mustBeZero[5] = 0;

-	cb->dxContext = 0;

-	cb->id.low = cb_next_id.low;

-	cb->id.hi  = cb_next_id.hi;

-	cb->length = data_size;

-

-	SVGA_WriteReg(SVGA_REG_COMMAND_HIGH, 0);

-	SVGA_WriteReg(SVGA_REG_COMMAND_LOW, (cb->ptr.pa.low - sizeof(SVGACBHeader)) | SVGA_CB_CONTEXT_DEVICE);

-	SVGA_Sync();

-	

-	SVGA_cb_id_inc();

-

-	while(cb->status == SVGA_CB_STATUS_NONE)

-	{

-		SVGA_Sync();

-	}

-	

-	//End_Critical_Section();

-	

-	dbg_printf(dbg_ctr_end);

-	

-	return cb->status;

-}

-

-/**

- * GPU10: start context0

- *

- **/

-static void SVGA_CB_start()

-{

-	if(cb_support && cb_context0 == FALSE)

-	{

-		cb_enable_t *cbe = ctlbuf;

-		DWORD status;

-			

-		memset(cbe, 0, sizeof(cb_enable_t));

-		cbe->cmd = SVGA_DC_CMD_START_STOP_CONTEXT;

-		cbe->cbstart.enable  = 1;

-		cbe->cbstart.context = SVGA_CB_CONTEXT_0;

-		

-		status = SVGA_CB_ctr(sizeof(cb_enable_t));

-		

-		if(status == SVGA_CB_STATUS_COMPLETED)

-		{

-			dbg_printf(dbg_cb_start);

-			cb_context0 = TRUE;

-		}

-		else

-		{

-			cb_support = FALSE;

-			dbg_printf(dbg_cb_start_err, status);

-		}

-	}

-}

-

-/**

- * GPU10: stop context0

- *

- **/

-static void SVGA_CB_stop()

-{

-	cb_context0 = FALSE;

-	

-	if(cb_support)

-	{

-		DWORD status;

-		cb_enable_t *cbe = ctlbuf;

-			

-		memset(cbe, 0, sizeof(cb_enable_t));

-		cbe->cmd = SVGA_DC_CMD_START_STOP_CONTEXT;

-		cbe->cbstart.enable  = 0;

-		cbe->cbstart.context = SVGA_CB_CONTEXT_0;

-		

-		status = SVGA_CB_ctr(sizeof(cb_enable_t));

-		dbg_printf(dbg_cb_stop_status, status);

-		

-		dbg_printf(dbg_cb_stop);

-	}

-}

-

-static void SVGA_CB_restart()

-{

-	cb_context0 = FALSE;

-	SVGA_CB_start();

-}

-

 /* Check if screen acceleration is available */

 static BOOL SVGA_hasAccelScreen()

 {

@@ -2026,94 +710,6 @@ static void SVGA_defineScreen(DWORD w, DWORD h, DWORD bpp)
 	submit_cmdbuf(cmdoff, SVGA_CB_SYNC/*|SVGA_CB_FORCE_FIFO*/, 0);

 }

 

-/**

- * Apply tables to VGPU

- **/

-void SVGA_OTable_load()

-{

-	DWORD i;

-	DWORD cmd_offset = 0;

-	SVGA3dCmdSetOTableBase *cmd;

-

-	SVGA_OT_info_entry_t *ot = SVGA_OT_setup();

-	if(ot == NULL)

-	{

-		return;

-	}

-	

-	wait_for_cmdbuf();

-

-	for(i = SVGA_OTABLE_MOB; i < SVGA_OTABLE_DX_MAX; i++)

-	{

-		SVGA_OT_info_entry_t *entry = &(ot[i]);

-		

-		if((entry->flags & SVGA_OT_FLAG_ACTIVE) == 0)

-		{

-			if(entry->size > 0)

-			{

-				cmd = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_SET_OTABLE_BASE, sizeof(SVGA3dCmdSetOTableBase));

-				cmd->type = i;

-				cmd->baseAddress = entry->ppn;

-				cmd->sizeInBytes = entry->size;

-				if((entry->flags & SVGA_OT_FLAG_DIRTY) == 0)

-				{

-			  	cmd->validSizeInBytes = entry->size;

-			  }

-			  else

-			  {

-			  	cmd->validSizeInBytes = 0;			  	

-			  }

-			  cmd->ptDepth = entry->pt_depth;

-

-			  entry->flags |= SVGA_OT_FLAG_ACTIVE;

-			}

-		}

-	}

-

-	submit_cmdbuf(cmd_offset, SVGA_CB_SYNC, 0);

-}

-

-void SVGA_OTable_unload()

-{

-	DWORD i;

-	DWORD cmd_offset = 0;

-	SVGA3dCmdSetOTableBase *cmd;

-	SVGA3dCmdReadbackOTable *cmd_readback;

-

-	SVGA_OT_info_entry_t *ot = SVGA_OT_setup();

-	if(ot == NULL)

-	{

-		return;

-	}

-	

-	wait_for_cmdbuf();

-

-	for(i = SVGA_OTABLE_MOB; i < SVGA_OTABLE_DX_MAX; i++)

-	{

-		SVGA_OT_info_entry_t *entry = &(ot[i]);

-

-		if(entry->flags & SVGA_OT_FLAG_ACTIVE)

-		{

-			if(entry->size > 0)

-			{

-				cmd_readback = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_READBACK_OTABLE, sizeof(SVGA3dCmdReadbackOTable));

-				cmd_readback->type = i;

-				entry->flags &= ~SVGA_OT_FLAG_DIRTY;

-				

-				cmd = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_SET_OTABLE_BASE, sizeof(SVGA3dCmdSetOTableBase));

-				cmd->type = i;

-				cmd->baseAddress = 0;

-				cmd->sizeInBytes = 0;

-			  cmd->validSizeInBytes = 0;

-			  cmd->ptDepth = SVGA3D_MOBFMT_INVALID;

-

-			  entry->flags &= ~SVGA_OT_FLAG_ACTIVE;

-			}

-		}

-	}

-

-	submit_cmdbuf(cmd_offset, SVGA_CB_SYNC, 0);

-}

 

 /**

  * Test if display mode is supported

diff --git a/vxd_svga.h b/vxd_svga.h
index fe7ea05..04c1897 100644
--- a/vxd_svga.h
+++ b/vxd_svga.h
@@ -1,35 +1,47 @@
 #ifndef __VXD_SVGA_H__INCLUDED__

 #define __VXD_SVGA_H__INCLUDED__

 

+/* consts */

+#define ST_REGION_ID 1

+#define ST_SURFACE_ID 1

+

+#define ST_16BPP   1

+#define ST_CURSOR  2

+#define ST_CURSOR_HIDEABLE 4

+

+

+/* VM handle */

+extern DWORD ThisVM;

+

 extern void *cmdbuf;

+extern void *ctlbuf;

 void wait_for_cmdbuf();

 void submit_cmdbuf(DWORD cmdsize, DWORD flags, DWORD dx);

 void *SVGA_cmd_ptr(DWORD *buf, DWORD *pOffset, DWORD cmd, DWORD cmdsize);

 void *SVGA_cmd3d_ptr(DWORD *buf, DWORD *pOffset, DWORD cmd, DWORD cmdsize);

 DWORD SVGA_pitch(DWORD width, DWORD bpp);

 void SVGA_region_usage_reset();

+BOOL SVGA_fence_is_passed(DWORD fence_id);

+DWORD SVGA_fence_passed();

+DWORD SVGA_GetDevCap(DWORD search_id);

 

+/* screen target */

 extern BOOL  st_used;

 extern DWORD st_flags;

 BOOL st_memory_allocate(DWORD size, DWORD *out);

 void st_defineScreen(DWORD w, DWORD h, DWORD bpp);

 void st_destroyScreen();

-void SVGA_OTable_load();

-void SVGA_OTable_unload();

 

 SVGA_DB_surface_t *SVGA_GetSurfaceInfo(DWORD sid);

-

-#define ST_REGION_ID 1

-#define ST_SURFACE_ID 1

-

-#define ST_16BPP   1

-#define ST_CURSOR  2

-#define ST_CURSOR_HIDEABLE 4

-

 BOOL st_useable(DWORD bpp);

 

+/* VDD */

 DWORD map_pm16(DWORD vm, DWORD linear, DWORD size);

 

+void SVGA_Sync();

+void SVGA_Flush_CB();

+void SVGA_Flush_CB_critical();

+

 /* mouse */

 BOOL SVGA_mouse_hw();

 BOOL SVGA_mouse_load();

@@ -37,4 +49,18 @@ void SVGA_mouse_move(int x, int y);
 void SVGA_mouse_show();

 void SVGA_mouse_hide(BOOL invalidate);

 

+/* memory */

+void SVGA_OTable_load();

+void SVGA_OTable_alloc(BOOL screentargets);

+void SVGA_OTable_unload();

+void cache_init();

+void cache_enable(BOOL enabled);

+

+/* CB */

+DWORD *SVGA_CMB_alloc_size(DWORD datasize);

+void SVGA_CMB_free(DWORD *cmb);

+void SVGA_CB_start();

+void SVGA_CB_stop();

+void SVGA_CB_restart();

+

 #endif /* __VXD_SVGA_H__INCLUDED__ */

diff --git a/vxd_svga_cb.c b/vxd_svga_cb.c
new file mode 100644
index 0000000..2fcbc81
--- /dev/null
+++ b/vxd_svga_cb.c
@@ -0,0 +1,606 @@
+/*****************************************************************************

+

+Copyright (c) 2024 Jaroslav Hensl <emulator@emulace.cz>

+

+Permission is hereby granted, free of charge, to any person obtaining a copy

+of this software and associated documentation files (the "Software"), to deal

+in the Software without restriction, including without limitation the rights

+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell

+copies of the Software, and to permit persons to whom the Software is

+furnished to do so, subject to the following conditions:

+

+The above copyright notice and this permission notice shall be included in

+all copies or substantial portions of the Software.

+

+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN

+THE SOFTWARE.

+

+*****************************************************************************/

+

+/* 32 bit RING-0 code for SVGA-II 3D acceleration */

+#define SVGA

+

+#include <limits.h>

+

+#include "winhack.h"

+#include "vmm.h"

+#include "vxd.h"

+#include "vxd_lib.h"

+

+#include "svga_all.h"

+#include "3d_accel.h"

+#include "code32.h"

+#include "vxd_svga.h"

+#include "vxd_strings.h"

+

+#include "svga_ver.h"

+

+/*

+ * types

+ */

+

+#pragma pack(push)

+#pragma pack(1)

+typedef struct _cb_enable_t

+{

+	uint32             cmd;

+	SVGADCCmdStartStop cbstart;

+} cb_enable_t;

+

+typedef struct _cb_queue_t

+{

+	struct _cb_queue_t *next;

+	DWORD  flags;

+	DWORD  data_size;

+	DWORD  pad[13];

+} cb_queue_t;

+

+#pragma pack(pop)

+

+typedef struct _cb_queue_info_t

+{

+	cb_queue_t *first;

+	cb_queue_t *last;

+} cb_queue_info_t;

+

+/*

+ * Globals

+ */

+extern BOOL gb_support;

+extern BOOL cb_support;

+extern BOOL cb_context0;

+

+extern DWORD present_fence;

+extern BOOL ST_FB_invalid;

+

+BOOL CB_queue_check(SVGACBHeader *tracked);

+

+/*

+ * Locals

+ **/

+static cb_queue_info_t cb_queue_info = {NULL, NULL};

+static uint64 cb_next_id = {0, 0};

+

+/*

+ * Macros

+ */

+

+#define WAIT_FOR_CB(_cb, _forcesync) \

+	do{ \

+		/*dbg_printf(dbg_wait_cb, __LINE__);*/ \

+		while(!CB_queue_check(_cb)){ \

+			WAIT_FOR_CB_SYNC_ ## _forcesync \

+		} \

+	}while(0)

+

+/* expansions of WAIT_FOR_CB */

+#define WAIT_FOR_CB_SYNC_0

+#define WAIT_FOR_CB_SYNC_1 SVGA_Sync();

+

+/* wait for all commands */

+void SVGA_Flush_CB_critical()

+{

+	/* wait for actual CB */

+	while(!CB_queue_check(NULL))

+	{

+		SVGA_Sync();

+	}

+	

+	/* drain FIFO */

+	SVGA_Flush();

+}

+

+void SVGA_Flush_CB()

+{

+	Begin_Critical_Section(0);

+	SVGA_Flush_CB_critical();

+	End_Critical_Section();

+}

+

+/**

+ * Allocate memory for command buffer

+ **/

+DWORD *SVGA_CMB_alloc_size(DWORD datasize)

+{

+	DWORD phy;

+	SVGACBHeader *cb;

+	cb_queue_t *q;

+	

+	Begin_Critical_Section(0);

+	

+	q = (cb_queue_t*)_PageAllocate(RoundToPages(datasize+sizeof(SVGACBHeader)+sizeof(cb_queue_t)), PG_SYS, 0, 0, 0x0, 0x100000, &phy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

+	

+	if(q)

+	{

+		q->next = NULL;

+		q->flags = 0;

+		q->data_size = 0;

+		

+		cb = (SVGACBHeader*)(q+1);

+		

+		memset(cb, 0, sizeof(SVGACBHeader));

+		cb->status = SVGA_CB_STATUS_COMPLETED; /* important to sync between commands */

+		cb->ptr.pa.hi   = 0;

+		cb->ptr.pa.low  = phy + sizeof(cb_queue_t) + sizeof(SVGACBHeader);	

+		

+		End_Critical_Section();

+		return (DWORD*)(cb+1);

+	}

+	End_Critical_Section();

+	

+	return NULL;

+}

+

+void SVGA_CMB_free(DWORD *cmb)

+{

+	SVGACBHeader *cb = ((SVGACBHeader *)cmb)-1;

+	

+	Begin_Critical_Section(0);

+	

+	WAIT_FOR_CB(cb, 1);

+		

+	_PageFree(cb-1, 0);

+	

+	End_Critical_Section();

+}

+

+DWORD *SVGA_CMB_alloc()

+{

+	return SVGA_CMB_alloc_size(SVGA_CB_MAX_SIZE);

+}

+

+// FIXME: inline?

+static void SVGA_cb_id_inc()

+{

+	_asm

+	{

+		inc dword ptr [cb_next_id]

+		adc dword ptr [cb_next_id+4], 0

+	}

+}

+

+#ifdef DBGPRINT

+#include "vxd_svga_debug.h"

+#endif

+

+/*

+ * @param tracked: check specific CB, or NULL to check full queue

+ *

+ * @return: TRUE if tracked is complete or TRUE id queue is empty

+ 

+ */

+BOOL CB_queue_check(SVGACBHeader *tracked)

+{

+	cb_queue_t *last = NULL;

+	cb_queue_t *item = cb_queue_info.first;

+	BOOL in_queue = FALSE;

+	

+	while(item != NULL)

+	{

+		SVGACBHeader *cb = (SVGACBHeader*)(item+1);

+		if(cb->status >= SVGA_CB_STATUS_COMPLETED)

+		{

+			if(last == NULL)

+			{

+				cb_queue_info.first = item->next;

+			}

+			else

+			{

+				last->next = item->next;

+			}

+			

+			if(cb->status > SVGA_CB_STATUS_COMPLETED)

+			{

+				SVGA_CB_restart();

+			}

+			

+			item = item->next;

+		}

+		else

+		{

+			if(tracked == cb)

+			{

+				in_queue = TRUE;

+			}

+			last = item;

+			item = item->next;

+		}

+	}

+	

+	cb_queue_info.last = last;

+	

+	if(cb_queue_info.first == NULL)

+	{

+		return TRUE;

+	}

+	

+	if(tracked != NULL && in_queue == FALSE)

+	{

+		return TRUE;

+	}

+	

+	return FALSE;	

+}

+

+void CB_queue_insert(SVGACBHeader *cb)

+{

+	cb_queue_t *item = (cb_queue_t*)(cb-1);

+	item->next = NULL;

+	

+	if(cb_queue_info.last != NULL)

+	{

+		cb_queue_info.last->next = item;

+		cb_queue_info.last = item;

+	}

+	else

+	{

+		cb_queue_info.first = item;

+		cb_queue_info.last  = item;

+	}

+}

+

+

+void SVGA_CMB_submit_critical(DWORD FBPTR cmb, DWORD cmb_size, SVGA_CMB_status_t FBPTR status, DWORD flags, DWORD DXCtxId)

+{

+	DWORD fence = 0;

+	SVGACBHeader *cb = ((SVGACBHeader *)cmb)-1;

+	

+	CB_queue_check(NULL);

+	

+	if(status)

+	{

+		cb->status = SVGA_CB_STATUS_NONE;

+		status->sStatus = SVGA_PROC_NONE;

+		status->qStatus = (volatile DWORD*)&cb->status;

+	}

+

+#ifdef DBGPRINT

+	debug_cmdbuf(cmb, cmb_size);

+//	debug_cmdbuf_trace(cmb, cmb_size, SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN);

+//		debug_draw(cmb, cmb_size);

+#endif

+

+	if(flags & SVGA_CB_PRESENT_ASYNC)

+	{

+		if(present_fence != 0)

+		{

+			SVGA_fence_wait(present_fence);

+		}

+		

+		flags |= SVGA_CB_FORCE_FENCE;

+	}

+	else if(present_fence)

+	{

+		if(SVGA_fence_is_passed(present_fence))

+		{

+			present_fence = 0;

+		}

+	}

+	

+	if(flags & SVGA_CB_PRESENT_GPU)

+	{

+		ST_FB_invalid = TRUE;

+	}

+	

+	if(  /* CB are supported and enabled */

+		cb_support &&

+		(cb_context0 || (flags & SVGA_CB_USE_CONTEXT_DEVICE)) &&

+		(flags & SVGA_CB_FORCE_FIFO) == 0

+	)

+	{

+		DWORD cbhwctxid = SVGA_CB_CONTEXT_0;

+		

+		if(flags & SVGA_CB_USE_CONTEXT_DEVICE)

+		{

+			cbhwctxid = SVGA_CB_CONTEXT_DEVICE;

+		}

+		

+		if(flags & SVGA_CB_FORCE_FENCE)

+		{

+			DWORD dwords = cmb_size/sizeof(DWORD);

+			fence = SVGA_fence_get();

+			cmb[dwords]   = SVGA_CMD_FENCE;

+			cmb[dwords+1] = fence;

+			cmb_size += sizeof(DWORD)*2;

+		}

+		

+		if(cmb_size == 0)

+		{

+			cb->status = SVGA_PROC_COMPLETED;

+			if(status)

+			{

+				status->sStatus = SVGA_PROC_COMPLETED;

+				status->qStatus = NULL;

+				status->fifo_fence_used = 0;

+			}

+		}

+		else

+		{

+			cb->status = SVGA_CB_STATUS_NONE;

+			cb->errorOffset = 0;

+			cb->offset = 0; /* VMware modified this, needs to be clear */

+			cb->flags  = SVGA_CB_FLAG_NO_IRQ;

+			

+			if(flags & SVGA_CB_FLAG_DX_CONTEXT)

+			{

+				cb->flags |= SVGA_CB_FLAG_DX_CONTEXT;

+				cb->dxContext = DXCtxId;

+			}

+			else

+			{

+				cb->dxContext = 0;

+			}

+			

+			cb->id.low = cb_next_id.low;

+			cb->id.hi  = cb_next_id.hi;

+			cb->length = cmb_size;

+			

+			CB_queue_insert(cb);			

+			

+			SVGA_WriteReg(SVGA_REG_COMMAND_HIGH, 0); // high part of 64-bit memory address...

+			SVGA_WriteReg(SVGA_REG_COMMAND_LOW, (cb->ptr.pa.low - sizeof(SVGACBHeader)) | cbhwctxid);

+			SVGA_Sync(); /* notify HV to read registers (VMware needs it) */

+			

+			SVGA_cb_id_inc();	

+			

+			if(flags & SVGA_CB_SYNC)

+			{

+				WAIT_FOR_CB(cb, 0);

+				

+				if(cb->status != SVGA_CB_STATUS_COMPLETED)

+				{

+					dbg_printf(dbg_cmd_error, cb->status, cmb[0], cb->errorOffset);

+					if(flags & SVGA_CB_FORCE_FENCE)

+					{

+						/* this may cause freeze, when fence is in buffer and isn't complete. So return some passed fence  */

+						fence = SVGA_fence_passed();

+					}

+				}

+				else

+				{

+					dbg_printf(dbg_cb_suc);

+				}

+				

+				if(status)

+				{

+					status->sStatus = SVGA_PROC_COMPLETED;

+					status->qStatus = NULL;

+					status->fifo_fence_used = fence;

+				}

+			}

+			else

+			{

+				if(status)

+				{

+					status->sStatus = SVGA_PROC_NONE;

+					status->qStatus = (volatile DWORD*)&cb->status;

+					status->fifo_fence_used = fence;

+				}

+			}

+		}	

+	}

+	else /* copy to FIFO */

+	{

+		DWORD *ptr = cmb;

+		DWORD dwords = cmb_size/sizeof(DWORD);

+		DWORD nextCmd, max, min;

+		

+		/* insert fence CMD */

+		fence = SVGA_fence_get();

+		ptr[dwords]   = SVGA_CMD_FENCE;

+		ptr[dwords+1] = fence;

+		

+		dwords += 2;

+		

+		//SVGA_Flush();

+		

+		nextCmd = gSVGA.fifoMem[SVGA_FIFO_NEXT_CMD];

+		max     = gSVGA.fifoMem[SVGA_FIFO_MAX];

+		min     = gSVGA.fifoMem[SVGA_FIFO_MIN];

+		

+		/* copy to fifo */

+		while(dwords > 0)

+		{

+			gSVGA.fifoMem[nextCmd/sizeof(DWORD)] = *ptr;

+			ptr++;

+			

+			nextCmd += sizeof(uint32);

+			if (nextCmd >= max)

+			{

+				nextCmd = min;

+			}

+			gSVGA.fifoMem[SVGA_FIFO_NEXT_CMD] = nextCmd;

+			dwords--;

+		}

+		

+		//SVGA_Sync();

+		

+		if(flags & SVGA_CB_SYNC)

+		{

+			SVGA_fence_wait(fence);

+			if(status)

+			{

+				status->sStatus = SVGA_PROC_COMPLETED;

+				status->qStatus = NULL;

+				status->fifo_fence_used = 0;

+			}

+		}

+		else

+		{

+			if(status)

+			{

+				status->sStatus = SVGA_PROC_FENCE;

+				status->qStatus = NULL;

+				status->fifo_fence_used = fence;

+			}

+		}

+		

+		cb->status = SVGA_PROC_COMPLETED;

+	} /* FIFO */

+	

+	if(status)

+	{

+		status->fifo_fence_last = SVGA_fence_passed();

+	}

+	

+	if(flags & SVGA_CB_PRESENT_ASYNC)

+	{

+		present_fence = fence;

+	}

+	

+	//dbg_printf(dbg_cmd_off, cmb[0]);

+}

+

+void wait_for_cmdbuf()

+{

+	SVGACBHeader *cb;

+	

+	Begin_Critical_Section(0);

+	cb = ((SVGACBHeader *)cmdbuf)-1;

+	WAIT_FOR_CB(cb, 0);

+}

+

+void submit_cmdbuf(DWORD cmdsize, DWORD flags, DWORD dx)

+{

+	SVGA_CMB_submit_critical(cmdbuf, cmdsize, NULL, flags, dx);

+	End_Critical_Section();

+}

+

+void SVGA_CMB_submit(DWORD FBPTR cmb, DWORD cmb_size, SVGA_CMB_status_t FBPTR status, DWORD flags, DWORD DXCtxId)

+{

+	Begin_Critical_Section(0);

+	SVGA_CMB_submit_critical(cmb, cmb_size, status, flags, DXCtxId);

+	End_Critical_Section();

+}

+

+static DWORD SVGA_CB_ctr(DWORD data_size)

+{

+	SVGACBHeader *cb = ((SVGACBHeader *)ctlbuf)-1;

+	

+	dbg_printf(dbg_ctr_start);

+

+	//Begin_Critical_Section(0);

+

+	cb->status = SVGA_CB_STATUS_NONE;

+	cb->errorOffset = 0;

+	cb->offset = 0; /* VMware modified this, needs to be clear */

+	cb->flags  = SVGA_CB_FLAG_NO_IRQ;

+	cb->mustBeZero[0] = 0;

+	cb->mustBeZero[1] = 0;

+	cb->mustBeZero[2] = 0;

+	cb->mustBeZero[3] = 0;

+	cb->mustBeZero[4] = 0;

+	cb->mustBeZero[5] = 0;

+	cb->dxContext = 0;

+	cb->id.low = cb_next_id.low;

+	cb->id.hi  = cb_next_id.hi;

+	cb->length = data_size;

+

+	SVGA_WriteReg(SVGA_REG_COMMAND_HIGH, 0);

+	SVGA_WriteReg(SVGA_REG_COMMAND_LOW, (cb->ptr.pa.low - sizeof(SVGACBHeader)) | SVGA_CB_CONTEXT_DEVICE);

+	SVGA_Sync();

+	

+	SVGA_cb_id_inc();

+

+	while(cb->status == SVGA_CB_STATUS_NONE)

+	{

+		SVGA_Sync();

+	}

+	

+	//End_Critical_Section();

+	

+	dbg_printf(dbg_ctr_end);

+	

+	return cb->status;

+}

+

+/**

+ * GPU10: start context0

+ *

+ **/

+void SVGA_CB_start()

+{

+	if(cb_support && cb_context0 == FALSE)

+	{

+		cb_enable_t *cbe = ctlbuf;

+		DWORD status;

+			

+		memset(cbe, 0, sizeof(cb_enable_t));

+		cbe->cmd = SVGA_DC_CMD_START_STOP_CONTEXT;

+		cbe->cbstart.enable  = 1;

+		cbe->cbstart.context = SVGA_CB_CONTEXT_0;

+		

+		status = SVGA_CB_ctr(sizeof(cb_enable_t));

+		

+		if(status == SVGA_CB_STATUS_COMPLETED)

+		{

+			dbg_printf(dbg_cb_start);

+			cb_context0 = TRUE;

+		}

+		else

+		{

+			cb_support = FALSE;

+			dbg_printf(dbg_cb_start_err, status);

+		}

+	}

+}

+

+/**

+ * GPU10: stop context0

+ *

+ **/

+void SVGA_CB_stop()

+{

+	cb_context0 = FALSE;

+	

+	if(cb_support)

+	{

+		DWORD status;

+		cb_enable_t *cbe = ctlbuf;

+			

+		memset(cbe, 0, sizeof(cb_enable_t));

+		cbe->cmd = SVGA_DC_CMD_START_STOP_CONTEXT;

+		cbe->cbstart.enable  = 0;

+		cbe->cbstart.context = SVGA_CB_CONTEXT_0;

+		

+		status = SVGA_CB_ctr(sizeof(cb_enable_t));

+		dbg_printf(dbg_cb_stop_status, status);

+		

+		dbg_printf(dbg_cb_stop);

+	}

+}

+

+/**

+ * GPU10: restart context0 after error

+ *

+ **/

+void SVGA_CB_restart()

+{

+	cb_context0 = FALSE;

+	SVGA_CB_start();

+}

diff --git a/vxd_svga_mem.c b/vxd_svga_mem.c
new file mode 100644
index 0000000..319fb53
--- /dev/null
+++ b/vxd_svga_mem.c
@@ -0,0 +1,932 @@
+/*****************************************************************************

+

+Copyright (c) 2024 Jaroslav Hensl <emulator@emulace.cz>

+

+Permission is hereby granted, free of charge, to any person obtaining a copy

+of this software and associated documentation files (the "Software"), to deal

+in the Software without restriction, including without limitation the rights

+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell

+copies of the Software, and to permit persons to whom the Software is

+furnished to do so, subject to the following conditions:

+

+The above copyright notice and this permission notice shall be included in

+all copies or substantial portions of the Software.

+

+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN

+THE SOFTWARE.

+

+*****************************************************************************/

+

+/* 32 bit RING-0 code for SVGA-II 3D acceleration */

+#define SVGA

+

+#include <limits.h>

+

+#include "winhack.h"

+#include "vmm.h"

+#include "vxd.h"

+#include "vxd_lib.h"

+

+#include "svga_all.h"

+#include "3d_accel.h"

+#include "code32.h"

+#include "vxd_svga.h"

+#include "vxd_strings.h"

+

+#include "svga_ver.h"

+

+/*

+ * consts

+ */

+#define PTONPAGE (P_SIZE/sizeof(DWORD))

+

+#define SPARE_REGIONS_LARGE 2

+#define SPARE_REGION_LARGE_SIZE (32*1024*1024)

+

+#define SPARE_REGIONS_MEDIUM 16

+#define SPARE_REGION_MEDIUM_SIZE (1024*1024)

+

+#define SPARE_REGIONS_SMALL 128

+//#define SPARE_REGION_SMALL_SIZE 4096

+

+#define SPARE_REGIONS_CNT (SPARE_REGIONS_LARGE+SPARE_REGIONS_MEDIUM+SPARE_REGIONS_SMALL)

+

+#define CACHE_THRESHOLD 128

+

+/**

+ * types

+ */

+

+typedef struct spare_region

+{

+	BOOL      used;

+	uint32    missed;

+	SVGA_region_info_t region;

+} spare_region_t;

+

+typedef struct svga_cache_state

+{

+	int free_index_min;

+	int free_index_max;

+	int cnt_large;

+	int cnt_medium;

+	int cnt_small;

+	spare_region_t spare_region[SPARE_REGIONS_CNT];

+} svga_cache_state_t;

+

+

+/* object table for GPU10 */

+static SVGA_OT_info_entry_t otable_setup[SVGA_OTABLE_DX_MAX] = {

+	{0, NULL, RoundTo4k(VMWGFX_NUM_MOB*sizeof(SVGAOTableMobEntry)),                 0, 0}, /* SVGA_OTABLE_MOB */

+	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SURFACE*sizeof(SVGAOTableSurfaceEntry)),      0, 0}, /* SVGA_OTABLE_SURFACE */

+	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_CONTEXT*sizeof(SVGAOTableContextEntry)),      0, 0}, /* SVGA_OTABLE_CONTEXT */

+	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SHADER *sizeof(SVGAOTableShaderEntry)),       0, 0}, /* SVGA_OTABLE_SHADER - not used */

+	{0, NULL, RoundTo4k(VMWGFX_NUM_GB_SCREEN_TARGET*sizeof(SVGAOTableScreenTargetEntry)), 0, 0}, /* SVGA_OTABLE_SCREENTARGET (VBOX_VIDEO_MAX_SCREENS) */

+	{0, NULL, RoundTo4k(VMWGFX_NUM_DXCONTEXT*sizeof(SVGAOTableDXContextEntry)),    0, 0}, /* SVGA_OTABLE_DXCONTEXT */

+};

+

+static SVGA_OT_info_entry_t *otable = NULL;

+

+static svga_cache_state_t cache_state = {0, 0};

+

+static BOOL cache_enabled = FALSE;

+

+

+/*

+ * globals

+ */

+extern FBHDA_t *hda;

+extern ULONG hda_sem;

+static SVGA_DB_t *svga_db = NULL;

+

+extern LONG fb_lock_cnt;

+

+extern BOOL gb_support;

+

+

+/**

+ * Return PPN (physical page number) to virtual address

+ *

+ **/

+static DWORD getPPN(DWORD virtualaddr)

+{

+	DWORD phy = 0;

+	_CopyPageTable(virtualaddr/P_SIZE, 1, &phy, 0);

+	return phy/P_SIZE;

+}

+

+/**

+ * Return PPN for pointer

+ *

+ **/

+static DWORD getPtrPPN(void *ptr)

+{

+	return getPPN((DWORD)ptr);

+}

+

+/**

+ * Return page table pages need for buffer of specific size

+ *

+ **/

+static DWORD PT_count(DWORD size)

+{

+	DWORD pt1_entries = 0;

+	DWORD pt2_entries = 0;

+

+	pt1_entries = RoundToPages(size);

+	pt2_entries = ((pt1_entries + PTONPAGE - 1)/PTONPAGE);

+	

+	dbg_printf(dbg_pt_build_2, size, pt1_entries, pt2_entries);

+	

+	if(pt2_entries > 1)

+	{

+		return pt2_entries + 1;

+	}

+	

+	return 1;

+}

+

+/**

+ * Build paget table for specific buffer

+ *

+ **/

+static void PT_build(DWORD size, void *buf, DWORD *outBase, DWORD *outType, void **outUserPtr)

+{

+	DWORD pt1_entries = 0;

+	DWORD pt2_entries = 0;

+	DWORD i;

+	

+	pt1_entries = RoundToPages(size);

+	pt2_entries = ((pt1_entries + PTONPAGE - 1)/PTONPAGE);

+

+	if(pt2_entries > 1)

+	{

+		DWORD *ptbuf = buf;

+		BYTE *ptr = ((BYTE*)buf)+((pt2_entries+1)*P_SIZE);

+		

+		memset(buf, 0, (pt2_entries+1)*P_SIZE);

+		

+		/* build PT2 */

+		for(i = 0; i < pt2_entries; i++)

+		{

+			ptbuf[i] = getPtrPPN(ptbuf + ((i+1)*PTONPAGE));

+		}

+		

+		ptbuf += PTONPAGE;

+		

+		/* build PT2 */

+		for(i = 0; i < pt1_entries; i++)

+		{

+			ptbuf[i] = getPtrPPN(ptr + i*P_SIZE);

+		}

+		

+		*outBase = getPtrPPN(buf);

+		*outType = SVGA3D_MOBFMT_PTDEPTH_2;

+		*outUserPtr = (void*)ptr;

+	}

+	else

+	{

+		DWORD *ptbuf = buf;

+		BYTE *ptr = ((BYTE*)buf)+P_SIZE;

+		

+		memset(buf, 0, P_SIZE);

+		

+		for(i = 0; i < pt1_entries; i++)

+		{

+			ptbuf[i] = getPtrPPN(ptr + i*P_SIZE);

+		}

+		

+		*outBase = getPtrPPN(buf);

+		*outType = SVGA3D_MOBFMT_PTDEPTH_1;

+		*outUserPtr = (void*)ptr;

+	}

+	

+	dbg_printf(dbg_pt_build, size, *outBase, *outType, *outUserPtr);

+}

+

+/**

+ * Allocate OTable for GB objects

+ **/

+void SVGA_OTable_alloc(BOOL screentargets)

+{

+	int i;

+	

+	if(otable == NULL)

+	{

+		otable = (SVGA_OT_info_entry_t *)_PageAllocate(RoundToPages(sizeof(otable_setup)), PG_VM, ThisVM, 0, 0x0, 0x100000, NULL, PAGEFIXED);

+		if(otable)

+		{

+			memcpy(otable, &(otable_setup[0]), sizeof(otable_setup));

+		}

+	}

+	

+	/* not allocate DX when there is no DX support */

+	if(!SVGA_GetDevCap(SVGA3D_DEVCAP_DXCONTEXT))

+	{

+		otable[SVGA_OTABLE_DXCONTEXT].size = 0;

+	}

+	

+	if(!screentargets)

+	{

+		otable[SVGA_OTABLE_SCREENTARGET].size = 0;

+	}

+

+	if(otable)

+	{

+		for(i = 0; i < SVGA_OTABLE_DX_MAX; i++)

+		{

+			SVGA_OT_info_entry_t *entry = &otable[i];

+			if(entry->size != 0 && (entry->flags & SVGA_OT_FLAG_ALLOCATED) == 0)

+			{

+				void *ptr = (void*)_PageAllocate(RoundToPages(entry->size)+PT_count(entry->size), PG_VM, ThisVM, 0, 0x0, 0x100000, NULL, PAGEFIXED);

+				

+				if(ptr)

+				{

+					dbg_printf(dbg_mob_allocate, i);

+					PT_build(entry->size, ptr, &entry->ppn, &entry->pt_depth, &entry->lin);

+					

+					entry->flags |= SVGA_OT_FLAG_ALLOCATED;

+					entry->flags |= SVGA_OT_FLAG_DIRTY;

+				}

+			}

+		}

+	}

+}

+

+SVGA_OT_info_entry_t *SVGA_OT_setup()

+{

+	return otable;

+}

+

+/**

+ * Apply tables to VGPU

+ **/

+void SVGA_OTable_load()

+{

+	DWORD i;

+	DWORD cmd_offset = 0;

+	SVGA3dCmdSetOTableBase *cmd;

+

+	SVGA_OT_info_entry_t *ot = SVGA_OT_setup();

+	if(ot == NULL)

+	{

+		return;

+	}

+	

+	wait_for_cmdbuf();

+

+	for(i = SVGA_OTABLE_MOB; i < SVGA_OTABLE_DX_MAX; i++)

+	{

+		SVGA_OT_info_entry_t *entry = &(ot[i]);

+		

+		if((entry->flags & SVGA_OT_FLAG_ACTIVE) == 0)

+		{

+			if(entry->size > 0)

+			{

+				cmd = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_SET_OTABLE_BASE, sizeof(SVGA3dCmdSetOTableBase));

+				cmd->type = i;

+				cmd->baseAddress = entry->ppn;

+				cmd->sizeInBytes = entry->size;

+				if((entry->flags & SVGA_OT_FLAG_DIRTY) == 0)

+				{

+			  	cmd->validSizeInBytes = entry->size;

+			  }

+			  else

+			  {

+			  	cmd->validSizeInBytes = 0;			  	

+			  }

+			  cmd->ptDepth = entry->pt_depth;

+

+			  entry->flags |= SVGA_OT_FLAG_ACTIVE;

+			}

+		}

+	}

+

+	submit_cmdbuf(cmd_offset, SVGA_CB_SYNC, 0);

+}

+

+void SVGA_OTable_unload()

+{

+	DWORD i;

+	DWORD cmd_offset = 0;

+	SVGA3dCmdSetOTableBase *cmd;

+	SVGA3dCmdReadbackOTable *cmd_readback;

+

+	SVGA_OT_info_entry_t *ot = SVGA_OT_setup();

+	if(ot == NULL)

+	{

+		return;

+	}

+	

+	wait_for_cmdbuf();

+

+	for(i = SVGA_OTABLE_MOB; i < SVGA_OTABLE_DX_MAX; i++)

+	{

+		SVGA_OT_info_entry_t *entry = &(ot[i]);

+

+		if(entry->flags & SVGA_OT_FLAG_ACTIVE)

+		{

+			if(entry->size > 0)

+			{

+				cmd_readback = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_READBACK_OTABLE, sizeof(SVGA3dCmdReadbackOTable));

+				cmd_readback->type = i;

+				entry->flags &= ~SVGA_OT_FLAG_DIRTY;

+				

+				cmd = SVGA_cmd3d_ptr(cmdbuf, &cmd_offset, SVGA_3D_CMD_SET_OTABLE_BASE, sizeof(SVGA3dCmdSetOTableBase));

+				cmd->type = i;

+				cmd->baseAddress = 0;

+				cmd->sizeInBytes = 0;

+			  cmd->validSizeInBytes = 0;

+			  cmd->ptDepth = SVGA3D_MOBFMT_INVALID;

+

+			  entry->flags &= ~SVGA_OT_FLAG_ACTIVE;

+			}

+		}

+	}

+

+	submit_cmdbuf(cmd_offset, SVGA_CB_SYNC, 0);

+}

+

+

+/**/

+/**

+ * Delete item from cache

+ **/

+static void cache_delete(int index)

+{

+	if(cache_state.spare_region[index].used != FALSE)

+	{

+		SVGA_region_info_t *rinfo = &cache_state.spare_region[index].region;

+		BYTE *free_ptr = (BYTE*)rinfo->address;

+		uint32 region_size = rinfo->size;;

+		

+		if(rinfo->region_address != NULL)

+		{

+			dbg_printf(dbg_pagefree, rinfo->region_address);

+			_PageFree((PVOID)rinfo->region_address, 0);

+		}

+		else

+		{

+			free_ptr -= P_SIZE;

+		}

+			

+		if(rinfo->mob_address != NULL)

+		{

+			dbg_printf(dbg_pagefree, rinfo->mob_address);

+			_PageFree((PVOID)rinfo->mob_address, 0);

+		}

+			

+		dbg_printf(dbg_pagefree, free_ptr);

+		_PageFree((PVOID)free_ptr, 0);

+			

+		cache_state.spare_region[index].used = FALSE;

+		

+		if((index+1) == cache_state.free_index_max)

+		{

+			cache_state.free_index_max--;

+		}

+		

+		if(index < cache_state.free_index_min)

+		{

+			cache_state.free_index_min = index;

+		}

+		

+		if(region_size >= SPARE_REGION_LARGE_SIZE)

+			cache_state.cnt_large--;

+		else if(region_size >= SPARE_REGION_MEDIUM_SIZE)

+			cache_state.cnt_medium--;

+		else

+			cache_state.cnt_small--;

+	}

+}

+

+/**

+ * Insert region to cache

+ **/

+static BOOL cache_insert(SVGA_region_info_t *region)

+{

+	int i;

+	BOOL rc = FALSE;

+	

+	if(!cache_enabled)

+	{

+		return FALSE;

+	}

+	

+	if(region->size >= SPARE_REGION_LARGE_SIZE)

+	{

+		if(cache_state.cnt_large > SPARE_REGIONS_LARGE) return FALSE;

+	}

+	else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

+	{

+		if(cache_state.cnt_medium > SPARE_REGIONS_MEDIUM) return FALSE;

+	}

+	else

+	{

+		if(cache_state.cnt_small > SPARE_REGIONS_SMALL) return FALSE;

+	}

+	

+	for(i = cache_state.free_index_min; i < SPARE_REGIONS_CNT; i++)

+	{		

+		if(!cache_state.spare_region[i].used)

+		{

+			memcpy(&cache_state.spare_region[i].region, region, sizeof(SVGA_region_info_t));

+			cache_state.spare_region[i].used = TRUE;

+			cache_state.spare_region[i].missed = 0;

+			rc = TRUE;

+			break;

+		}

+	}

+	

+	for(; i < SPARE_REGIONS_CNT; i++)

+	{

+		if(!cache_state.spare_region[i].used)

+		{

+			cache_state.free_index_min = i;

+			break;

+		}

+	}

+	

+	if(i == SPARE_REGIONS_CNT)

+	{

+		cache_state.free_index_min = i;

+		cache_state.free_index_max = SPARE_REGIONS_CNT;

+	}

+	else if(i >= cache_state.free_index_max)

+	{

+		cache_state.free_index_max = i+1;

+	}

+

+	if(rc)

+	{

+		if(region->size >= SPARE_REGION_LARGE_SIZE)

+			cache_state.cnt_large++;

+		else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

+			cache_state.cnt_medium++;

+		else

+			cache_state.cnt_small++;

+

+		dbg_printf(dbg_cache_insert, region->region_id, region->size);

+	}

+

+	return rc;

+}

+

+/**

+ * Use region from cache

+ **/

+static BOOL cache_use(SVGA_region_info_t *region)

+{

+	int i;

+	BOOL rc = FALSE;

+	

+	if(!cache_enabled)

+	{

+		return FALSE;

+	}

+	

+	dbg_printf(dbg_cache_search, region->size);

+	

+	for(i = cache_state.free_index_max-1; i >= 0; i--)

+	{

+		if(cache_state.spare_region[i].used != FALSE)

+		{

+			if(cache_state.spare_region[i].region.size == region->size)

+			{

+				SVGA_region_info_t *ptr = &cache_state.spare_region[i].region;

+				

+				region->address        = ptr->address;

+				region->region_address = ptr->region_address;

+				region->region_ppn     = ptr->region_ppn;

+				region->mob_address    = ptr->mob_address;

+				region->mob_ppn        = ptr->mob_ppn;

+				

+				cache_state.spare_region[i].used = FALSE;

+				rc = TRUE;

+				break;

+			}

+			else

+			{

+				if(++cache_state.spare_region[i].missed >= CACHE_THRESHOLD)

+				{

+					dbg_printf(dbg_cache_delete, cache_state.spare_region[i].region.size);

+					cache_delete(i);

+				}

+			}

+		}

+	}

+	

+	if(rc)

+	{

+		if((i+1) == cache_state.free_index_max)

+		{

+			cache_state.free_index_max--;

+		}

+		

+		if(i < cache_state.free_index_min)

+		{

+			cache_state.free_index_min = i;

+		}

+		

+		if(region->size >= SPARE_REGION_LARGE_SIZE)

+			cache_state.cnt_large--;

+		else if(region->size >= SPARE_REGION_MEDIUM_SIZE)

+			cache_state.cnt_medium--;

+		else

+			cache_state.cnt_small--;

+	

+		dbg_printf(dbg_cache_used, region->region_id, region->size);

+	}

+		

+	return rc;

+}

+

+void cache_init()

+{

+	memset(&cache_state, 0, sizeof(svga_cache_state_t));

+}

+

+void cache_enable(BOOL enabled)

+{

+	cache_enabled = enabled;

+}

+

+//#define GMR_CONTIG

+//#define GMR_SYSTEM

+

+//static DWORD pa_flags = PAGEFIXED | PAGEUSEALIGN;

+//static DWORD pa_align = 0x0000000F; // 64k

+

+static DWORD pa_flags = PAGEFIXED;

+static DWORD pa_align = 0x00000000;

+

+/**

+ * Allocate guest memory region (GMR) - HW needs know memory physical

+ * addressed of pages in (virtual) memory block.

+ * Technically this allocate 2 memory block, 1st for data and 2nd as its

+ * physical description.

+ *

+ * @return: TRUE on success

+ *

+ **/

+BOOL SVGA_region_create(SVGA_region_info_t *rinfo)

+{

+#ifdef GMR_CONTIG

+	ULONG phy = 0;

+#endif

+	ULONG laddr;

+	ULONG pgblk;

+	ULONG new_size = RoundTo4k(rinfo->size);	

+	ULONG nPages = RoundToPages(new_size);

+	//ULONG nPages = RoundToPages64k(rinfo->size);

+	ULONG pa_vm = ThisVM;

+	ULONG pa_type = PG_VM;

+	SVGAGuestMemDescriptor *desc;

+	ULONG size_total = 0;

+

+#ifdef GMR_SYSTEM

+		pa_vm = 0;

+		pa_type = PG_SYS;

+#endif

+	

+	rinfo->size = new_size;

+	

+	Begin_Critical_Section(0);

+

+	if(cache_use(rinfo))

+	{

+		goto spare_region_used;

+	}

+	

+	//dbg_printf(dbg_pages, rinfo->size, nPages, P_SIZE);

+#ifdef GMR_CONTIG

+	/* JH: OK, using PAGECONTIG leads to very fast memory exhaustion, so using only slower way! */

+	/* first try to allocate continuous physical memory space, 1st page is used for GMR descriptor */

+	laddr = _PageAllocate(nPages+1, pa_type, pa_vm, 0, 0x0, 0x100000, &phy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

+	

+	if(laddr)

+	{

+		desc = (SVGAGuestMemDescriptor*)laddr;

+		desc->ppn = (phy/P_SIZE)+1;

+		desc->numPages = nPages;

+		desc++;

+		desc->ppn = 0;

+		desc->numPages = 0;

+		

+		rinfo->address        = (void*)(laddr+P_SIZE);

+		rinfo->region_address = 0;

+		rinfo->region_ppn     = (phy/P_SIZE);

+		rinfo->mob_address    = 0;	

+		rinfo->mob_ppn        = (phy/P_SIZE)+1;

+		

+		//dbg_printf(dbg_region_simple, laddr, rinfo->address);

+		

+		size_total = (nPages+1)*P_SIZE;

+	}

+	else

+#endif

+	{

+		/* memory is too fragmented to create this large continous region */

+		ULONG taddr;

+		ULONG tppn;

+		ULONG pgi;

+		ULONG base_ppn;

+		ULONG base_cnt;

+		ULONG blocks = 1;

+		ULONG blk_pages = 0;

+		ULONG mob_pages = 0;

+		

+		DWORD *mob = NULL;

+		//DWORD mobphy = 0;

+		

+		/* allocate user block */

+		laddr = _PageAllocate(nPages, pa_type, pa_vm, pa_align, 0x0, 0x100000, NULL, pa_flags);

+		

+		if(!laddr)

+		{

+			End_Critical_Section();

+			return FALSE;

+		}

+		

+		/* determine how many physical continuous blocks we have */

+		base_ppn = getPPN(laddr);

+		base_cnt = 1;

+		for(pgi = 1; pgi < nPages; pgi++)

+		{

+			taddr = laddr + pgi*P_SIZE;

+			tppn = getPPN(taddr);

+			

+			if(tppn != base_ppn + base_cnt)

+			{

+				base_ppn = tppn;

+				base_cnt = 1;

+				blocks++;

+			}

+			else

+			{

+				base_cnt++;

+			}

+		}

+			

+		// number of pages to store regions information

+		blk_pages = ((blocks+1)/(P_SIZE/sizeof(SVGAGuestMemDescriptor))) + 1;

+			

+		// add extra descriptors for pages edge

+		blk_pages = ((blocks+blk_pages+1)/(P_SIZE/sizeof(SVGAGuestMemDescriptor))) + 1;

+		

+		/* allocate memory for GMR descriptor */

+		pgblk = _PageAllocate(blk_pages, pa_type, pa_vm, pa_align, 0x0, 0x100000, NULL, pa_flags);

+		if(!pgblk)

+		{

+			_PageFree((PVOID)laddr, 0);

+			End_Critical_Section();

+			return FALSE;

+		}

+		

+		desc = (SVGAGuestMemDescriptor*)pgblk;

+		desc->ppn = getPPN(laddr);

+		desc->numPages = 1;

+		blocks = 1;

+		

+		/* fill GMR physical structure */

+		for(pgi = 1; pgi < nPages; pgi++)

+		{

+			taddr = laddr + pgi*P_SIZE;

+			tppn = getPPN(taddr);

+					

+			if(tppn == desc->ppn + desc->numPages)

+			{

+				desc->numPages++;

+			}

+			else

+			{

+				/* next descriptor is on page edge */

+				if((blocks+1) % (P_SIZE/sizeof(SVGAGuestMemDescriptor)) == 0)

+				{

+					desc++;

+					desc->numPages = 0;

+					desc->ppn = getPPN((DWORD)(desc+1));

+					blocks++;

+				}

+

+				desc++;

+				desc->ppn = tppn;

+				desc->numPages = 1;

+				blocks++;

+			}

+		}

+

+		desc++;

+		desc->ppn = 0;

+		desc->numPages = 0;

+

+		if(gb_support) /* don't create MOBs for Gen9 */

+		{

+			int i;

+			/* for simplicity we're always creating table of depth 2, first page is page of PPN pages */

+			//mob = (DWORD *)_PageAllocate(1 + (nPages + PTONPAGE - 1)/PTONPAGE, pa_type, pa_vm, 0, 0x0, 0x100000, &mobphy, PAGECONTIG | PAGEUSEALIGN | PAGEFIXED);

+			mob_pages = 1 + (nPages + PTONPAGE - 1)/PTONPAGE;

+			mob = (DWORD *)_PageAllocate(mob_pages, pa_type, pa_vm, pa_align, 0x0, 0x100000, 0, pa_flags);

+			

+			if(!mob)

+			{

+				_PageFree((PVOID)laddr, 0);

+				_PageFree((PVOID)pgblk, 0);

+				End_Critical_Section();

+				return FALSE;

+			}

+			

+			memset(mob, 0, mob_pages*P_SIZE);

+			

+			/* dim 1 */

+			for(i = 0; i < mob_pages-1; i++)

+			{

+				//mob[i] = (mobphy/P_SIZE) + i + 1;

+				mob[i] = getPPN(((ULONG)mob) + ((i + 1)*P_SIZE));

+			}

+			/* dim 2 */

+			for(i = 0; i < nPages; i++)

+			{

+				mob[PTONPAGE + i] = getPPN(laddr + i*P_SIZE);

+			}

+		}

+		

+		rinfo->address        = (void*)laddr;

+		rinfo->region_address = (void*)pgblk;

+		rinfo->region_ppn     = getPPN(pgblk);

+		rinfo->mob_address    = (void*)mob;

+		rinfo->mob_ppn        = 0;

+		if(mob)

+		{

+			//rinfo->mob_ppn        = mobphy/P_SIZE;

+			rinfo->mob_ppn        = getPPN((ULONG)mob);

+		}

+		

+		size_total = (nPages + blk_pages + mob_pages)*P_SIZE;

+		

+		//dbg_printf(dbg_region_fragmented);

+	}

+	

+	//dbg_printf(dbg_gmr, rinfo->region_id, rinfo->region_ppn);

+	

+	spare_region_used:

+	//dbg_printf(dbg_mobonly, rinfo->mobonly);

+	

+	// JH: no need here

+	//SVGA_Flush_CB();

+	

+	if(!rinfo->mobonly)

+	{

+		if(rinfo->region_id < SVGA_ReadReg(SVGA_REG_GMR_MAX_IDS))

+		{

+			/* register GMR */

+			SVGA_WriteReg(SVGA_REG_GMR_ID, rinfo->region_id);

+			SVGA_WriteReg(SVGA_REG_GMR_DESCRIPTOR, rinfo->region_ppn);

+			SVGA_Sync(); // notify register change

+			

+			SVGA_Flush_CB_critical();

+		}

+		else

+		{

+			/* GMR is too high, use it as only as MOB (DX) */

+			rinfo->region_ppn = 0;

+		}

+	}

+

+	if(gb_support)

+	{

+		SVGA3dCmdDefineGBMob *mob;

+		DWORD cmdoff = 0;

+		

+		wait_for_cmdbuf();

+  	mob              = SVGA_cmd3d_ptr(cmdbuf, &cmdoff, SVGA_3D_CMD_DEFINE_GB_MOB, sizeof(SVGA3dCmdDefineGBMob));

+  	mob->mobid       = rinfo->region_id;

+  	mob->base        = rinfo->mob_ppn;

+  	mob->sizeInBytes = rinfo->size;

+		if(rinfo->mob_address == NULL) /* continual memory space */

+		{

+			mob->ptDepth     = SVGA3D_MOBFMT_RANGE;

+		}

+		else

+		{

+			mob->ptDepth     = SVGA3D_MOBFMT_PTDEPTH_2;

+		}

+		

+		submit_cmdbuf(cmdoff, SVGA_CB_SYNC, 0);

+		

+  	rinfo->is_mob = 1;

+	}

+	else

+	{

+		rinfo->is_mob = 0;

+	}

+	

+	svga_db->stat_regions_usage += rinfo->size;

+	

+	End_Critical_Section();

+	

+	//dbg_printf(dbg_gmr_succ, rinfo->region_id, rinfo->size);

+	

+	return TRUE;

+}

+

+/**

+ * Free data allocated by SVGA_region_create

+ *

+ **/

+void SVGA_region_free(SVGA_region_info_t *rinfo)

+{

+	BOOL saved_in_cache;

+	BYTE *free_ptr = (BYTE*)rinfo->address;

+	

+	Begin_Critical_Section(0);

+	svga_db->stat_regions_usage -= rinfo->size;

+	

+	saved_in_cache = cache_insert(rinfo);

+	

+	if(gb_support)

+	{

+		SVGA3dCmdDestroyGBMob *mob;

+		DWORD cmdoff = 0;

+		

+  	wait_for_cmdbuf();

+  	mob              = SVGA_cmd3d_ptr(cmdbuf, &cmdoff, SVGA_3D_CMD_DESTROY_GB_MOB, sizeof(SVGA3dCmdDestroyGBMob));

+  	mob->mobid       = rinfo->region_id;

+  	

+  	if(saved_in_cache)

+  		submit_cmdbuf(cmdoff, 0, 0);

+  	else

+			submit_cmdbuf(cmdoff, SVGA_CB_SYNC, 0);

+	}

+	

+	if(!rinfo->mobonly)

+	{

+		SVGA_Sync();

+		SVGA_Flush_CB_critical();

+		

+		SVGA_WriteReg(SVGA_REG_GMR_ID, rinfo->region_id);

+		SVGA_WriteReg(SVGA_REG_GMR_DESCRIPTOR, 0);

+		SVGA_Sync(); // notify register change

+	}

+

+	if(!saved_in_cache)

+	{

+		if(rinfo->region_address != NULL)

+		{

+			//dbg_printf(dbg_pagefree, rinfo->region_address);

+			_PageFree((PVOID)rinfo->region_address, 0);

+		}

+		else

+		{

+			free_ptr -= P_SIZE;

+		}

+			

+		if(rinfo->mob_address != NULL)

+		{

+			//dbg_printf(dbg_pagefree, rinfo->mob_address);

+			_PageFree((PVOID)rinfo->mob_address, 0);

+		}

+			

+		//dbg_printf(dbg_pagefree, free_ptr);

+		_PageFree((PVOID)free_ptr, 0);

+	}

+	End_Critical_Section();

+		

+	//dbg_printf(dbg_pagefree_end, rinfo->region_id, rinfo->size, saved_in_cache);

+		

+	rinfo->address        = NULL;

+	rinfo->region_address = NULL;

+	rinfo->mob_address    = NULL;

+	rinfo->region_ppn     = 0;

+	rinfo->mob_ppn        = 0;

+}

+

+/**

+ * Destroy saved regions and free memory

+ *

+ **/

+void SVGA_flushcache()

+{

+	int i;

+	

+	Begin_Critical_Section(0);

+	

+	for(i = 0; i < cache_state.free_index_max ; i++)

+	{

+		cache_delete(i);

+	}

+	

+	cache_state.free_index_max = 0;

+	cache_state.free_index_min = 0;

+	cache_state.cnt_large      = 0;

+	cache_state.cnt_medium     = 0;

+	cache_state.cnt_small      = 0;

+

+	End_Critical_Section();

+}

+