radeon_cp.c

ati driver

	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
	{ 0000000000, 0000000000 },
};


int RADEON_READ_PLL(drm_device_t *dev, int addr)
{
	drm_radeon_private_t *dev_priv = dev->dev_private;

	RADEON_WRITE8(RADEON_CLOCK_CNTL_INDEX, addr & 0x1f);
	return RADEON_READ(RADEON_CLOCK_CNTL_DATA);
}

#if RADEON_FIFO_DEBUG
static void radeon_status( drm_radeon_private_t *dev_priv )
{
	printk( "%s:\n", __FUNCTION__ );
	printk( "RBBM_STATUS = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_RBBM_STATUS ) );
	printk( "CP_RB_RTPR = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_CP_RB_RPTR ) );
	printk( "CP_RB_WTPR = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_CP_RB_WPTR ) );
	printk( "AIC_CNTL = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_AIC_CNTL ) );
	printk( "AIC_STAT = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_AIC_STAT ) );
	printk( "AIC_PT_BASE = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_AIC_PT_BASE ) );
	printk( "TLB_ADDR = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_AIC_TLB_ADDR ) );
	printk( "TLB_DATA = 0x%08x\n",
		(unsigned int)RADEON_READ( RADEON_AIC_TLB_DATA ) );
}
#endif


/* ================================================================
 * Engine, FIFO control
 */

static int radeon_do_pixcache_flush( drm_radeon_private_t *dev_priv )
{
	u32 tmp;
	int i;

	dev_priv->stats.boxes |= RADEON_BOX_WAIT_IDLE;

	tmp  = RADEON_READ( RADEON_RB2D_DSTCACHE_CTLSTAT );
	tmp |= RADEON_RB2D_DC_FLUSH_ALL;
	RADEON_WRITE( RADEON_RB2D_DSTCACHE_CTLSTAT, tmp );

	for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) {
		if ( !(RADEON_READ( RADEON_RB2D_DSTCACHE_CTLSTAT )
		       & RADEON_RB2D_DC_BUSY) ) {
			return 0;
		}
		DRM_UDELAY( 1 );
	}

#if RADEON_FIFO_DEBUG
	DRM_ERROR( "failed!\n" );
	radeon_status( dev_priv );
#endif
	return DRM_ERR(EBUSY);
}

static int radeon_do_wait_for_fifo( drm_radeon_private_t *dev_priv,
				    int entries )
{
	int i;

	dev_priv->stats.boxes |= RADEON_BOX_WAIT_IDLE;

	for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) {
		int slots = ( RADEON_READ( RADEON_RBBM_STATUS )
			      & RADEON_RBBM_FIFOCNT_MASK );
		if ( slots >= entries ) return 0;
		DRM_UDELAY( 1 );
	}

#if RADEON_FIFO_DEBUG
	DRM_ERROR( "failed!\n" );
	radeon_status( dev_priv );
#endif
	return DRM_ERR(EBUSY);
}

static int radeon_do_wait_for_idle( drm_radeon_private_t *dev_priv )
{
	int i, ret;

	dev_priv->stats.boxes |= RADEON_BOX_WAIT_IDLE;

	ret = radeon_do_wait_for_fifo( dev_priv, 64 );
	if ( ret ) return ret;

	for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) {
		if ( !(RADEON_READ( RADEON_RBBM_STATUS )
		       & RADEON_RBBM_ACTIVE) ) {
			radeon_do_pixcache_flush( dev_priv );
			return 0;
		}
		DRM_UDELAY( 1 );
	}

#if RADEON_FIFO_DEBUG
	DRM_ERROR( "failed!\n" );
	radeon_status( dev_priv );
#endif
	return DRM_ERR(EBUSY);
}


/* ================================================================
 * CP control, initialization
 */

/* Load the microcode for the CP */
static void radeon_cp_load_microcode( drm_radeon_private_t *dev_priv )
{
	int i;
	DRM_DEBUG( "\n" );

	radeon_do_wait_for_idle( dev_priv );

	RADEON_WRITE( RADEON_CP_ME_RAM_ADDR, 0 );

	if (dev_priv->is_r200)
	{
		DRM_INFO("Loading R200 Microcode\n");
		for ( i = 0 ; i < 256 ; i++ ) 
		{
			RADEON_WRITE( RADEON_CP_ME_RAM_DATAH,
				      R200_cp_microcode[i][1] );
			RADEON_WRITE( RADEON_CP_ME_RAM_DATAL,
				      R200_cp_microcode[i][0] );
		}
	}
	else
	{
		for ( i = 0 ; i < 256 ; i++ ) {
			RADEON_WRITE( RADEON_CP_ME_RAM_DATAH,
				      radeon_cp_microcode[i][1] );
			RADEON_WRITE( RADEON_CP_ME_RAM_DATAL,
				      radeon_cp_microcode[i][0] );
		}
	}
}

/* Flush any pending commands to the CP.  This should only be used just
 * prior to a wait for idle, as it informs the engine that the command
 * stream is ending.
 */
static void radeon_do_cp_flush( drm_radeon_private_t *dev_priv )
{
	DRM_DEBUG( "\n" );
#if 0
	u32 tmp;

	tmp = RADEON_READ( RADEON_CP_RB_WPTR ) | (1 << 31);
	RADEON_WRITE( RADEON_CP_RB_WPTR, tmp );
#endif
}

/* Wait for the CP to go idle.
 */
int radeon_do_cp_idle( drm_radeon_private_t *dev_priv )
{
	RING_LOCALS;
	DRM_DEBUG( "\n" );

	BEGIN_RING( 6 );

	RADEON_PURGE_CACHE();
	RADEON_PURGE_ZCACHE();
	RADEON_WAIT_UNTIL_IDLE();

	ADVANCE_RING();
	COMMIT_RING();

	return radeon_do_wait_for_idle( dev_priv );
}

/* Start the Command Processor.
 */
static void radeon_do_cp_start( drm_radeon_private_t *dev_priv )
{
	RING_LOCALS;
	DRM_DEBUG( "\n" );

	radeon_do_wait_for_idle( dev_priv );

	RADEON_WRITE( RADEON_CP_CSQ_CNTL, dev_priv->cp_mode );

	dev_priv->cp_running = 1;

	BEGIN_RING( 6 );

	RADEON_PURGE_CACHE();
	RADEON_PURGE_ZCACHE();
	RADEON_WAIT_UNTIL_IDLE();

	ADVANCE_RING();
	COMMIT_RING();
}

/* Reset the Command Processor.  This will not flush any pending
 * commands, so you must wait for the CP command stream to complete
 * before calling this routine.
 */
static void radeon_do_cp_reset( drm_radeon_private_t *dev_priv )
{
	u32 cur_read_ptr;
	DRM_DEBUG( "\n" );

	cur_read_ptr = RADEON_READ( RADEON_CP_RB_RPTR );
	RADEON_WRITE( RADEON_CP_RB_WPTR, cur_read_ptr );
	*dev_priv->ring.head = cur_read_ptr;
	dev_priv->ring.tail = cur_read_ptr;
}

/* Stop the Command Processor.  This will not flush any pending
 * commands, so you must flush the command stream and wait for the CP
 * to go idle before calling this routine.
 */
static void radeon_do_cp_stop( drm_radeon_private_t *dev_priv )
{
	DRM_DEBUG( "\n" );

	RADEON_WRITE( RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIDIS_INDDIS );

	dev_priv->cp_running = 0;
}

/* Reset the engine.  This will stop the CP if it is running.
 */
static int radeon_do_engine_reset( drm_device_t *dev )
{
	drm_radeon_private_t *dev_priv = dev->dev_private;
	u32 clock_cntl_index, mclk_cntl, rbbm_soft_reset;
	DRM_DEBUG( "\n" );

	radeon_do_pixcache_flush( dev_priv );

	clock_cntl_index = RADEON_READ( RADEON_CLOCK_CNTL_INDEX );
	mclk_cntl = RADEON_READ_PLL( dev, RADEON_MCLK_CNTL );

	RADEON_WRITE_PLL( RADEON_MCLK_CNTL, ( mclk_cntl |
					      RADEON_FORCEON_MCLKA |
					      RADEON_FORCEON_MCLKB |
 					      RADEON_FORCEON_YCLKA |
					      RADEON_FORCEON_YCLKB |
					      RADEON_FORCEON_MC |
					      RADEON_FORCEON_AIC ) );

	rbbm_soft_reset = RADEON_READ( RADEON_RBBM_SOFT_RESET );

	RADEON_WRITE( RADEON_RBBM_SOFT_RESET, ( rbbm_soft_reset |
						RADEON_SOFT_RESET_CP |
						RADEON_SOFT_RESET_HI |
						RADEON_SOFT_RESET_SE |
						RADEON_SOFT_RESET_RE |
						RADEON_SOFT_RESET_PP |
						RADEON_SOFT_RESET_E2 |
						RADEON_SOFT_RESET_RB ) );
	RADEON_READ( RADEON_RBBM_SOFT_RESET );
	RADEON_WRITE( RADEON_RBBM_SOFT_RESET, ( rbbm_soft_reset &
						~( RADEON_SOFT_RESET_CP |
						   RADEON_SOFT_RESET_HI |
						   RADEON_SOFT_RESET_SE |
						   RADEON_SOFT_RESET_RE |
						   RADEON_SOFT_RESET_PP |
						   RADEON_SOFT_RESET_E2 |
						   RADEON_SOFT_RESET_RB ) ) );
	RADEON_READ( RADEON_RBBM_SOFT_RESET );


	RADEON_WRITE_PLL( RADEON_MCLK_CNTL, mclk_cntl );
	RADEON_WRITE( RADEON_CLOCK_CNTL_INDEX, clock_cntl_index );
	RADEON_WRITE( RADEON_RBBM_SOFT_RESET,  rbbm_soft_reset );

	/* Reset the CP ring */
	radeon_do_cp_reset( dev_priv );

	/* The CP is no longer running after an engine reset */
	dev_priv->cp_running = 0;

	/* Reset any pending vertex, indirect buffers */
	radeon_freelist_reset( dev );

	return 0;
}

static void radeon_cp_init_ring_buffer( drm_device_t *dev,
				        drm_radeon_private_t *dev_priv )
{
	u32 ring_start, cur_read_ptr;
	u32 tmp;

	/* Initialize the memory controller */
	RADEON_WRITE( RADEON_MC_FB_LOCATION,
		      (dev_priv->agp_vm_start - 1) & 0xffff0000 );

	if ( !dev_priv->is_pci ) {
		RADEON_WRITE( RADEON_MC_AGP_LOCATION,
			      (((dev_priv->agp_vm_start - 1 +
				 dev_priv->agp_size) & 0xffff0000) |
			       (dev_priv->agp_vm_start >> 16)) );
	}

#if __REALLY_HAVE_AGP
	if ( !dev_priv->is_pci )
		ring_start = (dev_priv->cp_ring->offset
			      - dev->agp->base
			      + dev_priv->agp_vm_start);
       else
#endif
		ring_start = (dev_priv->cp_ring->offset
			      - dev->sg->handle
			      + dev_priv->agp_vm_start);

	RADEON_WRITE( RADEON_CP_RB_BASE, ring_start );

	/* Set the write pointer delay */
	RADEON_WRITE( RADEON_CP_RB_WPTR_DELAY, 0 );

	/* Initialize the ring buffer's read and write pointers */
	cur_read_ptr = RADEON_READ( RADEON_CP_RB_RPTR );
	RADEON_WRITE( RADEON_CP_RB_WPTR, cur_read_ptr );
	*dev_priv->ring.head = cur_read_ptr;
	dev_priv->ring.tail = cur_read_ptr;

	if ( !dev_priv->is_pci ) {
		RADEON_WRITE( RADEON_CP_RB_RPTR_ADDR,
			      dev_priv->ring_rptr->offset );
	} else {
		drm_sg_mem_t *entry = dev->sg;
		unsigned long tmp_ofs, page_ofs;

		tmp_ofs = dev_priv->ring_rptr->offset - dev->sg->handle;
		page_ofs = tmp_ofs >> PAGE_SHIFT;

		RADEON_WRITE( RADEON_CP_RB_RPTR_ADDR,
			     entry->busaddr[page_ofs]);
		DRM_DEBUG( "ring rptr: offset=0x%08x handle=0x%08lx\n",
			   entry->busaddr[page_ofs],
			   entry->handle + tmp_ofs );
	}

	/* Initialize the scratch register pointer.  This will cause
	 * the scratch register values to be written out to memory
	 * whenever they are updated.
	 *
	 * We simply put this behind the ring read pointer, this works
	 * with PCI GART as well as (whatever kind of) AGP GART
	 */
	RADEON_WRITE( RADEON_SCRATCH_ADDR, RADEON_READ( RADEON_CP_RB_RPTR_ADDR )
					 + RADEON_SCRATCH_REG_OFFSET );

	dev_priv->scratch = ((__volatile__ u32 *)
			     dev_priv->ring.head +
			     (RADEON_SCRATCH_REG_OFFSET / sizeof(u32)));

	RADEON_WRITE( RADEON_SCRATCH_UMSK, 0x7 );

	/* Writeback doesn't seem to work everywhere, test it first */
	DRM_WRITE32( &dev_priv->scratch[1], 0 );
	RADEON_WRITE( RADEON_SCRATCH_REG1, 0xdeadbeef );

	for ( tmp = 0 ; tmp < dev_priv->usec_timeout ; tmp++ ) {
		if ( DRM_READ32( &dev_priv->scratch[1] ) == 0xdeadbeef )
			break;
		DRM_UDELAY( 1 );
	}

	if ( tmp < dev_priv->usec_timeout ) {
		dev_priv->writeback_works = 1;
		DRM_DEBUG( "writeback test succeeded, tmp=%d\n", tmp );
	} else {
		dev_priv->writeback_works = 0;
		DRM_DEBUG( "writeback test failed\n" );
	}

	dev_priv->sarea_priv->last_frame = dev_priv->scratch[0] = 0;
	RADEON_WRITE( RADEON_LAST_FRAME_REG,
		      dev_priv->sarea_priv->last_frame );

	dev_priv->sarea_priv->last_dispatch = dev_priv->scratch[1] = 0;
	RADEON_WRITE( RADEON_LAST_DISPATCH_REG,
		      dev_priv->sarea_priv->last_dispatch );

	dev_priv->sarea_priv->last_clear = dev_priv->scratch[2] = 0;
	RADEON_WRITE( RADEON_LAST_CLEAR_REG,
		      dev_priv->sarea_priv->last_clear );

	/* Set ring buffer size */
#ifdef __BIG_ENDIAN
	RADEON_WRITE( RADEON_CP_RB_CNTL, dev_priv->ring.size_l2qw | RADEON_BUF_SWAP_32BIT );
#else
	RADEON_WRITE( RADEON_CP_RB_CNTL, dev_priv->ring.size_l2qw );
#endif

	radeon_do_wait_for_idle( dev_priv );

	/* Turn on bus mastering */
	tmp = RADEON_READ( RADEON_BUS_CNTL ) & ~RADEON_BUS_MASTER_DIS;
	RADEON_WRITE( RADEON_BUS_CNTL, tmp );

	/* Sync everything up */
	RADEON_WRITE( RADEON_ISYNC_CNTL,
		      (RADEON_ISYNC_ANY2D_IDLE3D |
		       RADEON_ISYNC_ANY3D_IDLE2D |
		       RADEON_ISYNC_WAIT_IDLEGUI |
		       RADEON_ISYNC_CPSCRATCH_IDLEGUI) );
}

static int radeon_do_init_cp( drm_device_t *dev, drm_radeon_init_t *init )
{
	drm_radeon_private_t *dev_priv;
	u32 tmp;
	DRM_DEBUG( "\n" );

	dev_priv = DRM(alloc)( sizeof(drm_radeon_private_t), DRM_MEM_DRIVER );
	if ( dev_priv == NULL )
		return DRM_ERR(ENOMEM);

	memset( dev_priv, 0, sizeof(drm_radeon_private_t) );

	dev_priv->is_pci = init->is_pci;

#if !defined(PCIGART_ENABLED)
	/* PCI support is not 100% working, so we disable it here.
	 */
	if ( dev_priv->is_pci ) {
		DRM_ERROR( "PCI GART not yet supported for Radeon!\n" );
		dev->dev_private = (void *)dev_priv;
		radeon_do_cleanup_cp(dev);
		return DRM_ERR(EINVAL);
	}
#endif

	if ( dev_priv->is_pci && !dev->sg ) {
		DRM_ERROR( "PCI GART memory not allocated!\n" );
		dev->dev_private = (void *)dev_priv;
		radeon_do_cleanup_cp(dev);
		return DRM_ERR(EINVAL);
	}

	dev_priv->usec_timeout = init->usec_timeout;
	if ( dev_priv->usec_timeout < 1 ||
	     dev_priv->usec_timeout > RADEON_MAX_USEC_TIMEOUT ) {
		DRM_DEBUG( "TIMEOUT problem!\n" );
		dev->dev_private = (void *)dev_priv;
		radeon_do_cleanup_cp(dev);
		return DRM_ERR(EINVAL);
	}

	dev_priv->is_r200 = (init->func == RADEON_INIT_R200_CP);
	dev_priv->do_boxes = 0;
	dev_priv->cp_mode = init->cp_mode;

	/* We don't support anything other than bus-mastering ring mode,
	 * but the ring can be in either AGP or PCI space for the ring
	 * read pointer.
	 */
	if ( ( init->cp_mode != RADEON_CSQ_PRIBM_INDDIS ) &&
	     ( init->cp_mode != RADEON_CSQ_PRIBM_INDBM ) ) {
		DRM_DEBUG( "BAD cp_mode (%x)!\n", init->cp_mode );
		dev->dev_private = (void *)dev_priv;
		radeon_do_cleanup_cp(dev);
		return DRM_ERR(EINVAL);
	}

	switch ( init->fb_bpp ) {
	case 16:
		dev_priv->color_fmt = RADEON_COLOR_FORMAT_RGB565;
		break;
	case 32:
	default:
		dev_priv->color_fmt = RADEON_COLOR_FORMAT_ARGB8888;
		break;
	}
	dev_priv->front_offset	= init->front_offset;
	dev_priv->front_pitch	= init->front_pitch;
	dev_priv->back_offset	= init->back_offset;
	dev_priv->back_pitch	= init->back_pitch;

	switch ( init->depth_bpp ) {
	case 16:
		dev_priv->depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
		break;
	case 32:
	default:
		dev_priv->depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
		break;
	}
	dev_priv->depth_offset	= init->depth_offset;
	dev_priv->depth_pitch	= init->depth_pitch;

	dev_priv->front_pitch_offset = (((dev_priv->front_pitch/64) << 22) |
					(dev_priv->front_offset >> 10));
	dev_priv->back_pitch_offset = (((dev_priv->back_pitch/64) << 22) |
				       (dev_priv->back_offset >> 10));
	dev_priv->depth_pitch_offset = (((dev_priv->depth_pitch/64) << 22) |
					(dev_priv->depth_offset >> 10));

	/* Hardware state for depth clears.  Remove this if/when we no
	 * longer clear the depth buffer with a 3D rectangle.  Hard-code
	 * all values to prevent unwanted 3D state from slipping through
	 * and screwing with the clear operation.
	 */
	dev_priv->depth_clear.rb3d_cntl = (RADEON_PLANE_MASK_ENABLE |
					   (dev_priv->color_fmt << 10) |
					   (1<<15));

	dev_priv->depth_clear.rb3d_zstencilcntl = 
		(dev_priv->depth_fmt |
		 RADEON_Z_TEST_ALWAYS |
		 RADEON_STENCIL_TEST_ALWAYS |
		 RADEON_STENCIL_S_FAIL_REPLACE |
		 RADEON_STENCIL_ZPASS_REPLACE |
		 RADEON_STENCIL_ZFAIL_REPLACE |
		 RADEON_Z_WRITE_ENABLE);

	dev_priv->depth_clear.se_cntl = (RADEON_FFACE_CULL_CW |
					 RADEON_BFACE_SOLID |
					 RADEON_FFACE_SOLID |
					 RADEON_FLAT_SHADE_VTX_LAST |
					 RADEON_DIFFUSE_SHADE_FLAT |
					 RADEON_ALPHA_SHADE_FLAT |
					 RADEON_SPECULAR_SHADE_FLAT |
					 RADEON_FOG_SHADE_FLAT |
					 RADEON_VTX_PIX_CENTER_OGL |
					 RADEON_ROUND_MODE_TRUNC |
					 RADEON_ROUND_PREC_8TH_PIX);

	DRM_GETSAREA();