static __inline__ int r300_emit_raw_packet3(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf)
{
u32 header;
int count;
RING_LOCALS;
if (4 > cmdbuf->bufsz)
return DRM_ERR(EINVAL);
/* Fixme !! This simply emits a packet without much checking.
We need to be smarter. */
/* obtain first word - actual packet3 header */
header = *(u32 __user*)cmdbuf->buf;
/* Is it packet 3 ? */
if( (header>>30)!=0x3 ) {
DRM_ERROR("Not a packet3 header (0x%08x)\n", header);
return DRM_ERR(EINVAL);
}
count=(header>>16) & 0x3fff;
/* Check again now that we know how much data to expect */
if ((count+2)*4 > cmdbuf->bufsz){
DRM_ERROR("Expected packet3 of length %d but have only %d bytes left\n",
(count+2)*4, cmdbuf->bufsz);
return DRM_ERR(EINVAL);
}
/* Is it a packet type we know about ? */
switch(header & 0xff00){
case RADEON_3D_LOAD_VBPNTR: /* load vertex array pointers */
return r300_emit_3d_load_vbpntr(dev_priv, cmdbuf, header);
case RADEON_CP_3D_DRAW_IMMD_2: /* triggers drawing using in-packet vertex data */
case RADEON_CP_3D_DRAW_VBUF_2: /* triggers drawing of vertex buffers setup elsewhere */
case RADEON_CP_3D_DRAW_INDX_2: /* triggers drawing using indices to vertex buffer */
case RADEON_CP_INDX_BUFFER: /* DRAW_INDX_2 without INDX_BUFFER seems to lock up the gpu */
case RADEON_WAIT_FOR_IDLE:
case RADEON_CP_NOP:
/* these packets are safe */
break;
default:
DRM_ERROR("Unknown packet3 header (0x%08x)\n", header);
return DRM_ERR(EINVAL);
}
BEGIN_RING(count+2);
OUT_RING(header);
OUT_RING_TABLE( (int __user*)(cmdbuf->buf+4), count+1);
ADVANCE_RING();
cmdbuf->buf += (count+2)*4;
cmdbuf->bufsz -= (count+2)*4;
return 0;
}
/**
* Uploads user-supplied vertex program instructions or parameters onto
* the graphics card.
* Called by r300_do_cp_cmdbuf.
*/
static __inline__ int r300_emit_vpu(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf,
drm_r300_cmd_header_t header)
{
int sz;
int addr;
RING_LOCALS;
sz = header.vpu.count;
addr = (header.vpu.adrhi << 8) | header.vpu.adrlo;
if (!sz)
return 0;
if (sz*16 > cmdbuf->bufsz)
return DRM_ERR(EINVAL);
BEGIN_RING(5+sz*4);
/* Wait for VAP to come to senses.. */
/* there is no need to emit it multiple times, (only once before VAP is programmed,
but this optimization is for later */
OUT_RING_REG( R300_VAP_PVS_WAITIDLE, 0 );
OUT_RING_REG( R300_VAP_PVS_UPLOAD_ADDRESS, addr );
OUT_RING( CP_PACKET0_TABLE( R300_VAP_PVS_UPLOAD_DATA, sz*4 - 1 ) );
OUT_RING_TABLE( (int __user*)cmdbuf->buf, sz*4 );
ADVANCE_RING();
cmdbuf->buf += sz*16;
cmdbuf->bufsz -= sz*16;
return 0;
}
static __inline__ int r300_emit_3d_load_vbpntr(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf,
u32 header)
{
int count, i,k;
#define MAX_ARRAY_PACKET 64
u32 payload[MAX_ARRAY_PACKET];
u32 narrays;
RING_LOCALS;
count=(header>>16) & 0x3fff;
if((count+1)>MAX_ARRAY_PACKET){
DRM_ERROR("Too large payload in 3D_LOAD_VBPNTR (count=%d)\n", count);
return DRM_ERR(EINVAL);
}
memset(payload, 0, MAX_ARRAY_PACKET*4);
memcpy(payload, cmdbuf->buf+4, (count+1)*4);
/* carefully check packet contents */
narrays=payload[0];
k=0;
i=1;
while((k<narrays) && (i<(count+1))){
i++; /* skip attribute field */
if(r300_check_offset(dev_priv, payload[i])){
DRM_ERROR("Offset failed range check (k=%d i=%d) while processing 3D_LOAD_VBPNTR packet.\n", k, i);
return DRM_ERR(EINVAL);
}
k++;
i++;
if(k==narrays)break;
/* have one more to process, they come in pairs */
if(r300_check_offset(dev_priv, payload[i])){
DRM_ERROR("Offset failed range check (k=%d i=%d) while processing 3D_LOAD_VBPNTR packet.\n", k, i);
return DRM_ERR(EINVAL);
}
k++;
i++;
}
/* do the counts match what we expect ? */
if((k!=narrays) || (i!=(count+1))){
DRM_ERROR("Malformed 3D_LOAD_VBPNTR packet (k=%d i=%d narrays=%d count+1=%d).\n", k, i, narrays, count+1);
return DRM_ERR(EINVAL);
}
/* all clear, output packet */
BEGIN_RING(count+2);
OUT_RING(header);
OUT_RING_TABLE(payload, count+1);
ADVANCE_RING();
cmdbuf->buf += (count+2)*4;
cmdbuf->bufsz -= (count+2)*4;
return 0;
}
/* Emit blit with arbitrary source and destination offsets and pitches */
static void
RADEONBlitChunk(ScrnInfoPtr pScrn, struct radeon_bo *src_bo,
struct radeon_bo *dst_bo, uint32_t datatype,
uint32_t src_pitch_offset, uint32_t dst_pitch_offset,
int srcX, int srcY, int dstX, int dstY, int w, int h,
uint32_t src_domain, uint32_t dst_domain)
{
RADEONInfoPtr info = RADEONPTR(pScrn);
if (src_bo && dst_bo) {
BEGIN_ACCEL_RELOC(6, 2);
} else if (src_bo && dst_bo == NULL) {
BEGIN_ACCEL_RELOC(6, 1);
} else {
BEGIN_RING(2*6);
}
OUT_RING_REG(RADEON_DP_GUI_MASTER_CNTL,
RADEON_GMC_DST_PITCH_OFFSET_CNTL |
RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
RADEON_GMC_BRUSH_NONE |
(datatype << 8) |
RADEON_GMC_SRC_DATATYPE_COLOR |
RADEON_ROP3_S |
RADEON_DP_SRC_SOURCE_MEMORY |
RADEON_GMC_CLR_CMP_CNTL_DIS |
RADEON_GMC_WR_MSK_DIS);
OUT_RING_REG(RADEON_SRC_PITCH_OFFSET, src_pitch_offset);
if (src_bo) {
OUT_RING_RELOC(src_bo, src_domain, 0);
}
OUT_RING_REG(RADEON_DST_PITCH_OFFSET, dst_pitch_offset);
if (dst_bo) {
OUT_RING_RELOC(dst_bo, 0, dst_domain);
}
OUT_RING_REG(RADEON_SRC_Y_X, (srcY << 16) | srcX);
OUT_RING_REG(RADEON_DST_Y_X, (dstY << 16) | dstX);
OUT_RING_REG(RADEON_DST_HEIGHT_WIDTH, (h << 16) | w);
ADVANCE_RING();
BEGIN_RING(2*2);
OUT_RING_REG(RADEON_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL);
OUT_RING_REG(RADEON_WAIT_UNTIL,
RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_DMA_GUI_IDLE);
ADVANCE_RING();
}
/**
* Emit the sequence to pacify R300.
*/
static __inline__ void r300_pacify(drm_radeon_private_t* dev_priv)
{
RING_LOCALS;
BEGIN_RING(6);
OUT_RING( CP_PACKET0( R300_RB3D_DSTCACHE_CTLSTAT, 0 ) );
OUT_RING( 0xa );
OUT_RING( CP_PACKET0( 0x4f18, 0 ) );
OUT_RING( 0x3 );
OUT_RING( CP_PACKET3( RADEON_CP_NOP, 0 ) );
OUT_RING( 0x0 );
ADVANCE_RING();
}
static void
RADEONFlush2D(PixmapPtr pPix)
{
RINFO_FROM_SCREEN(pPix->drawable.pScreen);
TRACE;
BEGIN_RING(2*2);
OUT_RING_REG(RADEON_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL);
OUT_RING_REG(RADEON_WAIT_UNTIL,
RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_DMA_GUI_IDLE);
ADVANCE_RING();
}
BOOL copybox810( GMABitMap_t *bm_dst, GMABitMap_t *bm_src,
ULONG dst_x,ULONG dst_y,ULONG dst_width, ULONG dst_height,
ULONG src_x,ULONG src_y,ULONG src_width, ULONG src_height, ULONG mode )
{
uint32_t br00, br13, br22, br23, br09, br11, br26, br12;
D(bug("[IntelGMA:HW] %s()\n", __func__));
LOCK_HW
{
br00 = (2 << 29) | (0x53 << 22) | (6);
if (bm_dst->bpp == 4)
br00 |= 3 << 20;
br13 = bm_dst->pitch | ROP_table[mode].rop;
if (bm_dst->bpp == 4)
br13 |= 3 << 24;
else if (bm_dst->bpp == 2)
br13 |= 1 << 24;
br22 = dst_x | (dst_y << 16);
br23 = (dst_x + dst_width) | (dst_y + dst_height) << 16;
br09 = bm_dst->framebuffer;
br11 = bm_src->pitch;
br26 = src_x | (src_y << 16);
br12 = bm_src->framebuffer;
START_RING(8);
OUT_RING(br00);
OUT_RING(br13);
OUT_RING(br22);
OUT_RING(br23);
OUT_RING(br09);
OUT_RING(br26);
OUT_RING(br11);
OUT_RING(br12);
ADVANCE_RING();
DO_FLUSH();
}
UNLOCK_HW
return TRUE;
}
static int radeon_emit_irq(struct drm_device * dev)
{
drm_radeon_private_t *dev_priv = dev->dev_private;
unsigned int ret;
RING_LOCALS;
atomic_inc(&dev_priv->swi_emitted);
ret = atomic_read(&dev_priv->swi_emitted);
BEGIN_RING(4);
OUT_RING_REG(RADEON_LAST_SWI_REG, ret);
OUT_RING_REG(RADEON_GEN_INT_STATUS, RADEON_SW_INT_FIRE);
ADVANCE_RING();
COMMIT_RING();
return ret;
}
static __inline__ int r300_emit_carefully_checked_packet0(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf,
drm_r300_cmd_header_t header)
{
int reg;
int sz;
int i;
int values[64];
RING_LOCALS;
sz = header.packet0.count;
reg = (header.packet0.reghi << 8) | header.packet0.reglo;
if((sz>64)||(sz<0)){
DRM_ERROR("Cannot emit more than 64 values at a time (reg=%04x sz=%d)\n", reg, sz);
return DRM_ERR(EINVAL);
}
for(i=0;i<sz;i++){
values[i]=((int __user*)cmdbuf->buf)[i];
switch(r300_reg_flags[(reg>>2)+i]){
case MARK_SAFE:
break;
case MARK_CHECK_OFFSET:
if(r300_check_offset(dev_priv, (u32)values[i])){
DRM_ERROR("Offset failed range check (reg=%04x sz=%d)\n", reg, sz);
return DRM_ERR(EINVAL);
}
break;
default:
DRM_ERROR("Register %04x failed check as flag=%02x\n", reg+i*4, r300_reg_flags[(reg>>2)+i]);
return DRM_ERR(EINVAL);
}
}
BEGIN_RING(1+sz);
OUT_RING( CP_PACKET0( reg, sz-1 ) );
OUT_RING_TABLE( values, sz );
ADVANCE_RING();
cmdbuf->buf += sz*4;
cmdbuf->bufsz -= sz*4;
return 0;
}
/**
* Emit a clear packet from userspace.
* Called by r300_emit_packet3.
*/
static __inline__ int r300_emit_clear(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf)
{
RING_LOCALS;
if (8*4 > cmdbuf->bufsz)
return DRM_ERR(EINVAL);
BEGIN_RING(10);
OUT_RING( CP_PACKET3( R200_3D_DRAW_IMMD_2, 8 ) );
OUT_RING( R300_PRIM_TYPE_POINT|R300_PRIM_WALK_RING|
(1<<R300_PRIM_NUM_VERTICES_SHIFT) );
OUT_RING_TABLE( (int __user*)cmdbuf->buf, 8 );
ADVANCE_RING();
cmdbuf->buf += 8*4;
cmdbuf->bufsz -= 8*4;
return 0;
}
static void
RADEONSolid(PixmapPtr pPix, int x1, int y1, int x2, int y2)
{
RINFO_FROM_SCREEN(pPix->drawable.pScreen);
TRACE;
if (CS_FULL(info->cs)) {
RADEONFlush2D(info->accel_state->dst_pix);
radeon_cs_flush_indirect(pScrn);
}
if (info->accel_state->vsync)
RADEONWaitForVLine(pScrn, pPix,
radeon_pick_best_crtc(pScrn, FALSE, x1, x2, y1, y2),
y1, y2);
BEGIN_RING(2*2);
OUT_RING_REG(RADEON_DST_Y_X, (y1 << 16) | x1);
OUT_RING_REG(RADEON_DST_HEIGHT_WIDTH, ((y2 - y1) << 16) | (x2 - x1));
ADVANCE_RING();
}
static void Emit2DState(ScrnInfoPtr pScrn, int op)
{
RADEONInfoPtr info = RADEONPTR(pScrn);
int has_src;
/* don't emit if no operation in progress */
if (info->state_2d.op == 0 && op == 0)
return;
has_src = info->state_2d.src_pitch_offset || info->state_2d.src_bo;
if (has_src) {
BEGIN_ACCEL_RELOC(10, 2);
} else {
BEGIN_ACCEL_RELOC(9, 1);
}
OUT_RING_REG(RADEON_DEFAULT_SC_BOTTOM_RIGHT, info->state_2d.default_sc_bottom_right);
OUT_RING_REG(RADEON_DP_GUI_MASTER_CNTL, info->state_2d.dp_gui_master_cntl);
OUT_RING_REG(RADEON_DP_BRUSH_FRGD_CLR, info->state_2d.dp_brush_frgd_clr);
OUT_RING_REG(RADEON_DP_BRUSH_BKGD_CLR, info->state_2d.dp_brush_bkgd_clr);
OUT_RING_REG(RADEON_DP_SRC_FRGD_CLR, info->state_2d.dp_src_frgd_clr);
OUT_RING_REG(RADEON_DP_SRC_BKGD_CLR, info->state_2d.dp_src_bkgd_clr);
OUT_RING_REG(RADEON_DP_WRITE_MASK, info->state_2d.dp_write_mask);
OUT_RING_REG(RADEON_DP_CNTL, info->state_2d.dp_cntl);
OUT_RING_REG(RADEON_DST_PITCH_OFFSET, info->state_2d.dst_pitch_offset);
OUT_RING_RELOC(info->state_2d.dst_bo, 0, info->state_2d.dst_domain);
if (has_src) {
OUT_RING_REG(RADEON_SRC_PITCH_OFFSET, info->state_2d.src_pitch_offset);
OUT_RING_RELOC(info->state_2d.src_bo, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0);
}
ADVANCE_RING();
if (op)
info->state_2d.op = op;
info->reemit_current2d = Emit2DState;
}
/**
* Emits a packet0 setting arbitrary registers.
* Called by r300_do_cp_cmdbuf.
*
* Note that checks are performed on contents and addresses of the registers
*/
static __inline__ int r300_emit_packet0(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf,
drm_r300_cmd_header_t header)
{
int reg;
int sz;
RING_LOCALS;
sz = header.packet0.count;
reg = (header.packet0.reghi << 8) | header.packet0.reglo;
if (!sz)
return 0;
if (sz*4 > cmdbuf->bufsz)
return DRM_ERR(EINVAL);
if (reg+sz*4 >= 0x10000){
DRM_ERROR("No such registers in hardware reg=%04x sz=%d\n", reg, sz);
return DRM_ERR(EINVAL);
}
if(r300_check_range(reg, sz)){
/* go and check everything */
return r300_emit_carefully_checked_packet0(dev_priv, cmdbuf, header);
}
/* the rest of the data is safe to emit, whatever the values the user passed */
BEGIN_RING(1+sz);
OUT_RING( CP_PACKET0( reg, sz-1 ) );
OUT_RING_TABLE( (int __user*)cmdbuf->buf, sz );
ADVANCE_RING();
cmdbuf->buf += sz*4;
cmdbuf->bufsz -= sz*4;
return 0;
}
static void
RADEONCopy(PixmapPtr pDst,
int srcX, int srcY,
int dstX, int dstY,
int w, int h)
{
RINFO_FROM_SCREEN(pDst->drawable.pScreen);
TRACE;
if (CS_FULL(info->cs)) {
RADEONFlush2D(info->accel_state->dst_pix);
radeon_cs_flush_indirect(pScrn);
}
if (info->accel_state->xdir < 0) {
srcX += w - 1;
dstX += w - 1;
}
if (info->accel_state->ydir < 0) {
srcY += h - 1;
dstY += h - 1;
}
if (info->accel_state->vsync)
RADEONWaitForVLine(pScrn, pDst,
radeon_pick_best_crtc(pScrn, FALSE, dstX, dstX + w, dstY, dstY + h),
dstY, dstY + h);
BEGIN_RING(2*3);
OUT_RING_REG(RADEON_SRC_Y_X, (srcY << 16) | srcX);
OUT_RING_REG(RADEON_DST_Y_X, (dstY << 16) | dstX);
OUT_RING_REG(RADEON_DST_HEIGHT_WIDTH, (h << 16) | w);
ADVANCE_RING();
}
/**
* Parses and validates a user-supplied command buffer and emits appropriate
* commands on the DMA ring buffer.
* Called by the ioctl handler function radeon_cp_cmdbuf.
*/
int r300_do_cp_cmdbuf(drm_device_t* dev,
DRMFILE filp,
drm_file_t* filp_priv,
drm_radeon_cmd_buffer_t* cmdbuf)
{
drm_radeon_private_t *dev_priv = dev->dev_private;
drm_device_dma_t *dma = dev->dma;
drm_buf_t *buf = NULL;
int emit_dispatch_age = 0;
int ret = 0;
DRM_DEBUG("\n");
/* See the comment above r300_emit_begin3d for why this call must be here,
* and what the cleanup gotos are for. */
r300_pacify(dev_priv);
if (cmdbuf->nbox <= R300_SIMULTANEOUS_CLIPRECTS) {
ret = r300_emit_cliprects(dev_priv, cmdbuf, 0);
if (ret)
goto cleanup;
}
while(cmdbuf->bufsz >= sizeof(drm_r300_cmd_header_t)) {
int idx;
drm_r300_cmd_header_t header;
header.u = *(unsigned int *)cmdbuf->buf;
cmdbuf->buf += sizeof(header);
cmdbuf->bufsz -= sizeof(header);
switch(header.header.cmd_type) {
case R300_CMD_PACKET0:
DRM_DEBUG("R300_CMD_PACKET0\n");
ret = r300_emit_packet0(dev_priv, cmdbuf, header);
if (ret) {
DRM_ERROR("r300_emit_packet0 failed\n");
goto cleanup;
}
break;
case R300_CMD_VPU:
DRM_DEBUG("R300_CMD_VPU\n");
ret = r300_emit_vpu(dev_priv, cmdbuf, header);
if (ret) {
DRM_ERROR("r300_emit_vpu failed\n");
goto cleanup;
}
break;
case R300_CMD_PACKET3:
DRM_DEBUG("R300_CMD_PACKET3\n");
ret = r300_emit_packet3(dev_priv, cmdbuf, header);
if (ret) {
DRM_ERROR("r300_emit_packet3 failed\n");
goto cleanup;
}
break;
case R300_CMD_END3D:
DRM_DEBUG("R300_CMD_END3D\n");
/* TODO:
Ideally userspace driver should not need to issue this call,
i.e. the drm driver should issue it automatically and prevent
lockups.
In practice, we do not understand why this call is needed and what
it does (except for some vague guesses that it has to do with cache
coherence) and so the user space driver does it.
Once we are sure which uses prevent lockups the code could be moved
into the kernel and the userspace driver will not
need to use this command.
Note that issuing this command does not hurt anything
except, possibly, performance */
r300_pacify(dev_priv);
break;
case R300_CMD_CP_DELAY:
/* simple enough, we can do it here */
DRM_DEBUG("R300_CMD_CP_DELAY\n");
{
int i;
RING_LOCALS;
BEGIN_RING(header.delay.count);
for(i=0;i<header.delay.count;i++)
OUT_RING(RADEON_CP_PACKET2);
ADVANCE_RING();
}
break;
case R300_CMD_DMA_DISCARD:
DRM_DEBUG("RADEON_CMD_DMA_DISCARD\n");
idx = header.dma.buf_idx;
if (idx < 0 || idx >= dma->buf_count) {
DRM_ERROR("buffer index %d (of %d max)\n",
idx, dma->buf_count - 1);
ret = DRM_ERR(EINVAL);
goto cleanup;
}
buf = dma->buflist[idx];
if (buf->filp != filp || buf->pending) {
DRM_ERROR("bad buffer %p %p %d\n",
buf->filp, filp, buf->pending);
ret = DRM_ERR(EINVAL);
goto cleanup;
}
emit_dispatch_age = 1;
r300_discard_buffer(dev, buf);
break;
case R300_CMD_WAIT:
/* simple enough, we can do it here */
DRM_DEBUG("R300_CMD_WAIT\n");
if(header.wait.flags==0)break; /* nothing to do */
{
RING_LOCALS;
BEGIN_RING(2);
OUT_RING( CP_PACKET0( RADEON_WAIT_UNTIL, 0 ) );
OUT_RING( (header.wait.flags & 0xf)<<14 );
ADVANCE_RING();
}
break;
default:
DRM_ERROR("bad cmd_type %i at %p\n",
header.header.cmd_type,
cmdbuf->buf - sizeof(header));
ret = DRM_ERR(EINVAL);
goto cleanup;
}
}
DRM_DEBUG("END\n");
cleanup:
r300_pacify(dev_priv);
/* We emit the vertex buffer age here, outside the pacifier "brackets"
* for two reasons:
* (1) This may coalesce multiple age emissions into a single one and
* (2) more importantly, some chips lock up hard when scratch registers
* are written inside the pacifier bracket.
*/
if (emit_dispatch_age) {
RING_LOCALS;
/* Emit the vertex buffer age */
BEGIN_RING(2);
RADEON_DISPATCH_AGE(dev_priv->sarea_priv->last_dispatch);
ADVANCE_RING();
}
COMMIT_RING();
return ret;
}
BOOL copybox3d( GMABitMap_t *bm_dst, GMABitMap_t *bm_src,
ULONG dst_x,ULONG dst_y,ULONG dst_width, ULONG dst_height,
ULONG src_x,ULONG src_y,ULONG src_width, ULONG src_height )
{
uint32_t dst_format;
uint32_t src_format;
if( !copybox3d_supported() )
{
return FALSE;
}
// buffers in gfx memory ?
if( ! (bm_src->fbgfx && bm_src->fbgfx) )
{
return FALSE;
}
// Max texture size, src and dst must be differend (at least if overlaps)
if( bm_src->pitch/4 > 2048 || bm_src->height > 2048 || bm_dst == bm_src )
{
return FALSE;
}
// src pitch must be long aligmented.
if( bm_src->pitch & 0x3 )
{
bug("[IntelGMA] copybox3d: ERROR bm_src->pitch=%d/n",bm_src->pitch);
return FALSE;
}
if(bm_src->bpp == 4)
{
src_format = MAPSURF_32BIT | MT_32BIT_ARGB8888;
}
else if(bm_src->bpp == 2)
{
src_format = MAPSURF_16BIT | MT_16BIT_RGB565;
}
else
{
bug("[IntelGMA] copybox3d: ERROR src_bpp=%d/n",bm_src->bpp);
return FALSE;
}
if(bm_dst->bpp == 4)
{
dst_format = COLOR_BUF_ARGB8888;
}
else if(bm_dst->bpp == 2)
{
dst_format = COLOR_BUF_RGB565;
}
else
{
bug("[IntelGMA] copybox3d: ERROR dst_bpp=%d/n",bm_dst->bpp);
return FALSE;
}
D(bug("[IntelGMA:HW] %s()\n", __func__));
LOCK_HW
START_RING(72);
/* invariant state */
OUT_RING( _3DSTATE_AA_CMD |
AA_LINE_ECAAR_WIDTH_ENABLE |
AA_LINE_ECAAR_WIDTH_1_0 |
AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0 );
OUT_RING( _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD |
IAB_MODIFY_ENABLE |
IAB_MODIFY_FUNC | (BLENDFUNC_ADD << IAB_FUNC_SHIFT) |
IAB_MODIFY_SRC_FACTOR | (BLENDFACT_ONE <<
IAB_SRC_FACTOR_SHIFT) |
IAB_MODIFY_DST_FACTOR | (BLENDFACT_ZERO <<
IAB_DST_FACTOR_SHIFT) );
OUT_RING( _3DSTATE_DFLT_DIFFUSE_CMD );
OUT_RING( 0 );
OUT_RING( _3DSTATE_DFLT_SPEC_CMD );
OUT_RING( 0 );
OUT_RING( _3DSTATE_DFLT_Z_CMD );
OUT_RING( 0 );
OUT_RING( _3DSTATE_COORD_SET_BINDINGS |
CSB_TCB(0, 0) |
CSB_TCB(1, 1) |
CSB_TCB(2, 2) |
CSB_TCB(3, 3) |
CSB_TCB(4, 4) |
CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7) );
OUT_RING( _3DSTATE_RASTER_RULES_CMD |
ENABLE_POINT_RASTER_RULE |
OGL_POINT_RASTER_RULE |
ENABLE_LINE_STRIP_PROVOKE_VRTX |
ENABLE_TRI_FAN_PROVOKE_VRTX |
LINE_STRIP_PROVOKE_VRTX(1) |
TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D );
OUT_RING( _3DSTATE_MODES_4_CMD |
ENABLE_LOGIC_OP_FUNC | LOGIC_OP_FUNC(LOGICOP_COPY) |
ENABLE_STENCIL_WRITE_MASK | STENCIL_WRITE_MASK(0xff) |
ENABLE_STENCIL_TEST_MASK | STENCIL_TEST_MASK(0xff) );
OUT_RING( _3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | I1_LOAD_S(4) | I1_LOAD_S(5) | 2 );
OUT_RING( 0x00000000 ); /* Disable texture coordinate wrap-shortest */
OUT_RING( (1 << S4_POINT_WIDTH_SHIFT) |
S4_LINE_WIDTH_ONE |
S4_CULLMODE_NONE |
S4_VFMT_XY );
OUT_RING( 0x00000000 ); /* Stencil. */
OUT_RING( _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT );
OUT_RING( _3DSTATE_SCISSOR_RECT_0_CMD );
OUT_RING( 0 );
OUT_RING( 0 );
OUT_RING( _3DSTATE_DEPTH_SUBRECT_DISABLE );
OUT_RING( _3DSTATE_LOAD_INDIRECT | 0 ); /* disable indirect state */
OUT_RING( 0 );
OUT_RING( _3DSTATE_STIPPLE );
OUT_RING( 0x00000000 );
OUT_RING( _3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0 );
/* samler state */
#define TEX_COUNT 1
OUT_RING( _3DSTATE_MAP_STATE | (3 * TEX_COUNT) );
OUT_RING( (1 << TEX_COUNT) - 1 );
// Source buffer
OUT_RING( bm_src->framebuffer );
OUT_RING( src_format |
(bm_src->height - 1) << MS3_HEIGHT_SHIFT |
(bm_src->pitch/bm_src->bpp - 1) << MS3_WIDTH_SHIFT );
OUT_RING( (bm_src->pitch/4 -1) << MS4_PITCH_SHIFT );
OUT_RING( _3DSTATE_SAMPLER_STATE | (3 * TEX_COUNT) );
OUT_RING( (1 << TEX_COUNT) - 1 );
OUT_RING( MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT |
FILTER_NEAREST << SS2_MAG_FILTER_SHIFT |
FILTER_NEAREST << SS2_MIN_FILTER_SHIFT );
OUT_RING( TEXCOORDMODE_WRAP << SS3_TCX_ADDR_MODE_SHIFT |
TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT |
0 << SS3_TEXTUREMAP_INDEX_SHIFT );
OUT_RING( 0x00000000 );
/* render target state */
// Destination buffer
OUT_RING( _3DSTATE_BUF_INFO_CMD );
OUT_RING( BUF_3D_ID_COLOR_BACK | bm_dst->pitch );
OUT_RING( bm_dst->framebuffer );
OUT_RING( _3DSTATE_DST_BUF_VARS_CMD );
OUT_RING( dst_format |
DSTORG_HORT_BIAS(0x8) |
DSTORG_VERT_BIAS(0x8) );
/* draw rect is unconditional */
OUT_RING( _3DSTATE_DRAW_RECT_CMD );
OUT_RING( 0x00000000 );
OUT_RING( 0x00000000 ); // ymin, xmin
OUT_RING( DRAW_YMAX(dst_y + dst_height - 1) |
DRAW_XMAX(dst_x + dst_width - 1) );
/* yorig, xorig (relate to color buffer?) */
OUT_RING( 0x00000000 );
/* texfmt */
OUT_RING( _3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(1) | I1_LOAD_S(2) | I1_LOAD_S(6) | 2 );
OUT_RING( (4 << S1_VERTEX_WIDTH_SHIFT) | (4 << S1_VERTEX_PITCH_SHIFT) );
OUT_RING( ~S2_TEXCOORD_FMT(0, TEXCOORDFMT_NOT_PRESENT) |
S2_TEXCOORD_FMT(0, TEXCOORDFMT_2D) );
OUT_RING( S6_CBUF_BLEND_ENABLE | S6_COLOR_WRITE_ENABLE |
BLENDFUNC_ADD << S6_CBUF_BLEND_FUNC_SHIFT |
BLENDFACT_ONE << S6_CBUF_SRC_BLEND_FACT_SHIFT |
BLENDFACT_ZERO << S6_CBUF_DST_BLEND_FACT_SHIFT );
/* pixel shader */
OUT_RING( _3DSTATE_PIXEL_SHADER_PROGRAM | (1 + 3*3 - 2) );
/* decl FS_T0 */
OUT_RING( D0_DCL |
GET_UREG_TYPE(FS_T0) << D0_TYPE_SHIFT |
GET_UREG_NR(FS_T0) << D0_NR_SHIFT |
((GET_UREG_TYPE(FS_T0) != GET_UREG_TYPE_S) ? D0_CHANNEL_ALL : 0) );
OUT_RING( 0 );
OUT_RING( 0 );
/* decl FS_S0 */
OUT_RING( D0_DCL |
(GET_UREG_TYPE(FS_S0) << D0_TYPE_SHIFT) |
(GET_UREG_NR(FS_S0) << D0_NR_SHIFT) |
((GET_UREG_TYPE(FS_S0) != GET_UREG_TYPE_S) ? D0_CHANNEL_ALL : 0) );
OUT_RING( 0 );
OUT_RING( 0 );
/* texld(FS_OC, FS_S0, FS_T0 */
OUT_RING( T0_TEXLD |
(GET_UREG_TYPE(FS_OC) << T0_DEST_TYPE_SHIFT) |
(GET_UREG_NR(FS_OC) << T0_DEST_NR_SHIFT) |
(GET_UREG_NR(FS_S0) << T0_SAMPLER_NR_SHIFT) );
OUT_RING( (GET_UREG_TYPE(FS_T0) << T1_ADDRESS_GET_UREG_TYPE_SHIFT) |
(GET_UREG_NR(FS_T0) << T1_ADDRESS_GET_UREG_NR_SHIFT) );
OUT_RING( 0 );
// rectangle
// 3--x
// | |
// 2--1
OUT_RING( PRIM3D_RECTLIST | (3*4 - 1) );
OUT_RING( pack_float( dst_x + dst_width) );
OUT_RING( pack_float( dst_y + dst_height) );
OUT_RING( pack_float(src_x + src_width) );
OUT_RING( pack_float(src_y + src_height) );
OUT_RING( pack_float( dst_x + 0 ) );
OUT_RING( pack_float( dst_y +dst_height) );
OUT_RING( pack_float(src_x + 0) );
OUT_RING( pack_float(src_y + src_height) );
OUT_RING( pack_float( dst_x + 0 ) );
OUT_RING( pack_float( dst_y + 0 ) );
OUT_RING( pack_float(src_x + 0) );
OUT_RING( pack_float(src_y + 0) );
ADVANCE_RING();
DO_FLUSH();
UNLOCK_HW
return TRUE;
}
/**
* Emit up to R300_SIMULTANEOUS_CLIPRECTS cliprects from the given command
* buffer, starting with index n.
*/
static int r300_emit_cliprects(drm_radeon_private_t* dev_priv,
drm_radeon_cmd_buffer_t* cmdbuf,
int n)
{
drm_clip_rect_t box;
int nr;
int i;
RING_LOCALS;
nr = cmdbuf->nbox - n;
if (nr > R300_SIMULTANEOUS_CLIPRECTS)
nr = R300_SIMULTANEOUS_CLIPRECTS;
DRM_DEBUG("%i cliprects\n", nr);
if (nr) {
BEGIN_RING(6 + nr*2);
OUT_RING( CP_PACKET0( R300_RE_CLIPRECT_TL_0, nr*2 - 1 ) );
for(i = 0; i < nr; ++i) {
if (DRM_COPY_FROM_USER_UNCHECKED(&box, &cmdbuf->boxes[n+i], sizeof(box))) {
DRM_ERROR("copy cliprect faulted\n");
return DRM_ERR(EFAULT);
}
box.x1 = (box.x1 + R300_CLIPRECT_OFFSET) & R300_CLIPRECT_MASK;
box.y1 = (box.y1 + R300_CLIPRECT_OFFSET) & R300_CLIPRECT_MASK;
box.x2 = (box.x2 + R300_CLIPRECT_OFFSET) & R300_CLIPRECT_MASK;
box.y2 = (box.y2 + R300_CLIPRECT_OFFSET) & R300_CLIPRECT_MASK;
OUT_RING((box.x1 << R300_CLIPRECT_X_SHIFT) |
(box.y1 << R300_CLIPRECT_Y_SHIFT));
OUT_RING((box.x2 << R300_CLIPRECT_X_SHIFT) |
(box.y2 << R300_CLIPRECT_Y_SHIFT));
}
OUT_RING_REG( R300_RE_CLIPRECT_CNTL, r300_cliprect_cntl[nr-1] );
/* TODO/SECURITY: Force scissors to a safe value, otherwise the
* client might be able to trample over memory.
* The impact should be very limited, but I'd rather be safe than
* sorry.
*/
OUT_RING( CP_PACKET0( R300_RE_SCISSORS_TL, 1 ) );
OUT_RING( 0 );
OUT_RING( R300_SCISSORS_X_MASK | R300_SCISSORS_Y_MASK );
ADVANCE_RING();
} else {
/* Why we allow zero cliprect rendering:
* There are some commands in a command buffer that must be submitted
* even when there are no cliprects, e.g. DMA buffer discard
* or state setting (though state setting could be avoided by
* simulating a loss of context).
*
* Now since the cmdbuf interface is so chaotic right now (and is
* bound to remain that way for a bit until things settle down),
* it is basically impossible to filter out the commands that are
* necessary and those that aren't.
*
* So I choose the safe way and don't do any filtering at all;
* instead, I simply set up the engine so that all rendering
* can't produce any fragments.
*/
BEGIN_RING(2);
OUT_RING_REG( R300_RE_CLIPRECT_CNTL, 0 );
ADVANCE_RING();
}
return 0;
}
