From e15a24e154d6749ba0be3155ed802005c5e1ebf0 Mon Sep 17 00:00:00 2001 From: Gareth Hughes Date: Sat, 2 Dec 2000 06:14:18 +0000 Subject: Merged ati-4-1-1-branch into trunk. --- linux/Makefile.kernel | 3 +- linux/Makefile.linux | 6 +- linux/drm.h | 26 +- linux/r128_bufs.c | 20 +- linux/r128_cce.c | 1248 +++++++++++++++++++++++++++++++++++++ linux/r128_context.c | 18 +- linux/r128_dma.c | 909 --------------------------- linux/r128_drm.h | 275 +++++++-- linux/r128_drv.c | 221 +++---- linux/r128_drv.h | 490 +++++++++++---- linux/r128_state.c | 1630 +++++++++++++++++++++++++++++++++++++++++++++++++ 11 files changed, 3597 insertions(+), 1249 deletions(-) create mode 100644 linux/r128_cce.c delete mode 100644 linux/r128_dma.c create mode 100644 linux/r128_state.c (limited to 'linux') diff --git a/linux/Makefile.kernel b/linux/Makefile.kernel index b1b8d976..278ff264 100644 --- a/linux/Makefile.kernel +++ b/linux/Makefile.kernel @@ -38,7 +38,8 @@ endif gamma-objs := $(lib-objs) gamma_drv.o gamma_dma.o tdfx-objs := $(lib-objs) tdfx_drv.o tdfx_context.o -r128-objs := $(lib-objs) r128_drv.o r128_dma.o r128_context.o r128_bufs.o +r128-objs := $(lib-objs) r128_drv.o r128_cce.o r128_context.o r128_bufs.o\ + r128_state.o ffb-objs := $(lib-objs) ffb_drv.o ffb_context.o mga-objs := $(lib-objs) mga_drv.o mga_dma.o mga_context.o mga_bufs.o \ mga_state.o diff --git a/linux/Makefile.linux b/linux/Makefile.linux index adf19005..2d8abbb5 100644 --- a/linux/Makefile.linux +++ b/linux/Makefile.linux @@ -11,11 +11,11 @@ # 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 (including the next # paragraph) 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 @@ -60,7 +60,7 @@ GAMMAHEADERS= gamma_drv.h $(DRMHEADERS) TDFXOBJS= tdfx_drv.o tdfx_context.o TDFXHEADERS= tdfx_drv.h $(DRMHEADERS) -R128OBJS= r128_drv.o r128_dma.o r128_bufs.o r128_context.o +R128OBJS= r128_drv.o r128_cce.o r128_bufs.o r128_state.o r128_context.o R128HEADERS= r128_drv.h r128_drm.h $(DRMHEADERS) INC= /usr/include diff --git a/linux/drm.h b/linux/drm.h index 57032d6e..20691ccf 100644 --- a/linux/drm.h +++ b/linux/drm.h @@ -11,11 +11,11 @@ * 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 (including the next * paragraph) 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 @@ -23,7 +23,7 @@ * 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. - * + * * Authors: * Rickard E. (Rik) Faith * @@ -363,12 +363,20 @@ typedef struct drm_agp_info { #define DRM_IOCTL_I810_DOCOPY DRM_IO ( 0x48) /* Rage 128 specific ioctls */ -#define DRM_IOCTL_R128_INIT DRM_IOW( 0x40, drm_r128_init_t) -#define DRM_IOCTL_R128_RESET DRM_IO( 0x41) -#define DRM_IOCTL_R128_FLUSH DRM_IO( 0x42) -#define DRM_IOCTL_R128_IDLE DRM_IO( 0x43) -#define DRM_IOCTL_R128_PACKET DRM_IOW( 0x44, drm_r128_packet_t) -#define DRM_IOCTL_R128_VERTEX DRM_IOW( 0x45, drm_r128_vertex_t) +#define DRM_IOCTL_R128_INIT DRM_IOW( 0x40, drm_r128_init_t) +#define DRM_IOCTL_R128_CCE_START DRM_IO( 0x41) +#define DRM_IOCTL_R128_CCE_STOP DRM_IOW( 0x42, drm_r128_cce_stop_t) +#define DRM_IOCTL_R128_CCE_RESET DRM_IO( 0x43) +#define DRM_IOCTL_R128_CCE_IDLE DRM_IO( 0x44) +#define DRM_IOCTL_R128_RESET DRM_IO( 0x46) +#define DRM_IOCTL_R128_SWAP DRM_IO( 0x47) +#define DRM_IOCTL_R128_CLEAR DRM_IOW( 0x48, drm_r128_clear_t) +#define DRM_IOCTL_R128_VERTEX DRM_IOW( 0x49, drm_r128_vertex_t) +#define DRM_IOCTL_R128_INDICES DRM_IOW( 0x4a, drm_r128_indices_t) +#define DRM_IOCTL_R128_BLIT DRM_IOW( 0x4b, drm_r128_blit_t) +#define DRM_IOCTL_R128_DEPTH DRM_IOW( 0x4c, drm_r128_depth_t) +#define DRM_IOCTL_R128_STIPPLE DRM_IOW( 0x4d, drm_r128_stipple_t) +#define DRM_IOCTL_R128_PACKET DRM_IOWR(0x4e, drm_r128_packet_t) /* SiS specific ioctls */ #define SIS_IOCTL_FB_ALLOC DRM_IOWR( 0x44, drm_sis_mem_t) diff --git a/linux/r128_bufs.c b/linux/r128_bufs.c index 7e76441e..a060749e 100644 --- a/linux/r128_bufs.c +++ b/linux/r128_bufs.c @@ -11,11 +11,11 @@ * 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 (including the next * paragraph) 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 @@ -23,11 +23,11 @@ * 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. - * + * * Authors: Kevin E. Martin * Rickard E. (Rik) Faith * Jeff Hartmann - * + * */ #define __NO_VERSION__ @@ -94,7 +94,7 @@ int r128_addbufs_agp(struct inode *inode, struct file *filp, unsigned int cmd, } atomic_inc(&dev->buf_alloc); spin_unlock(&dev->count_lock); - + down(&dev->struct_sem); entry = &dma->bufs[order]; if (entry->buf_count) { @@ -102,7 +102,7 @@ int r128_addbufs_agp(struct inode *inode, struct file *filp, unsigned int cmd, atomic_dec(&dev->buf_alloc); return -ENOMEM; /* May only call once for each order */ } - + entry->buflist = drm_alloc(count * sizeof(*entry->buflist), DRM_MEM_BUFS); if (!entry->buflist) { @@ -111,7 +111,7 @@ int r128_addbufs_agp(struct inode *inode, struct file *filp, unsigned int cmd, return -ENOMEM; } memset(entry->buflist, 0, count * sizeof(*entry->buflist)); - + entry->buf_size = size; entry->page_order = page_order; offset = 0; @@ -243,16 +243,16 @@ int r128_mapbufs(struct inode *inode, struct file *filp, unsigned int cmd, if (dma->flags & _DRM_DMA_USE_AGP) { drm_map_t *map; - map = dev_priv->agp_vertbufs; + map = dev_priv->buffers; if (!map) { retcode = -EINVAL; goto done; } down(¤t->mm->mmap_sem); - virtual = do_mmap(filp, 0, map->size, + virtual = do_mmap(filp, 0, map->size, PROT_READ|PROT_WRITE, - MAP_SHARED, + MAP_SHARED, (unsigned long)map->offset); up(¤t->mm->mmap_sem); } else { diff --git a/linux/r128_cce.c b/linux/r128_cce.c new file mode 100644 index 00000000..0b1576a9 --- /dev/null +++ b/linux/r128_cce.c @@ -0,0 +1,1248 @@ +/* r128_cce.c -- ATI Rage 128 driver -*- linux-c -*- + * Created: Wed Apr 5 19:24:19 2000 by kevin@precisioninsight.com + * + * Copyright 2000 Precision Insight, Inc., Cedar Park, Texas. + * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California. + * All Rights Reserved. + * + * 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 (including the next + * paragraph) 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 + * PRECISION INSIGHT AND/OR ITS SUPPLIERS 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. + * + * Authors: + * Gareth Hughes + * + */ + +#define __NO_VERSION__ +#include "drmP.h" +#include "r128_drv.h" + +#include /* For task queue support */ +#include + + +/* FIXME: Temporary CCE packet buffer */ +u32 r128_cce_buffer[(1 << 14)] __attribute__ ((aligned (32))); + +/* CCE microcode (from ATI) */ +static u32 r128_cce_microcode[] = { + 0, 276838400, 0, 268449792, 2, 142, 2, 145, 0, 1076765731, 0, + 1617039951, 0, 774592877, 0, 1987540286, 0, 2307490946U, 0, + 599558925, 0, 589505315, 0, 596487092, 0, 589505315, 1, + 11544576, 1, 206848, 1, 311296, 1, 198656, 2, 912273422, 11, + 262144, 0, 0, 1, 33559837, 1, 7438, 1, 14809, 1, 6615, 12, 28, + 1, 6614, 12, 28, 2, 23, 11, 18874368, 0, 16790922, 1, 409600, 9, + 30, 1, 147854772, 16, 420483072, 3, 8192, 0, 10240, 1, 198656, + 1, 15630, 1, 51200, 10, 34858, 9, 42, 1, 33559823, 2, 10276, 1, + 15717, 1, 15718, 2, 43, 1, 15936948, 1, 570480831, 1, 14715071, + 12, 322123831, 1, 33953125, 12, 55, 1, 33559908, 1, 15718, 2, + 46, 4, 2099258, 1, 526336, 1, 442623, 4, 4194365, 1, 509952, 1, + 459007, 3, 0, 12, 92, 2, 46, 12, 176, 1, 15734, 1, 206848, 1, + 18432, 1, 133120, 1, 100670734, 1, 149504, 1, 165888, 1, + 15975928, 1, 1048576, 6, 3145806, 1, 15715, 16, 2150645232U, 2, + 268449859, 2, 10307, 12, 176, 1, 15734, 1, 15735, 1, 15630, 1, + 15631, 1, 5253120, 6, 3145810, 16, 2150645232U, 1, 15864, 2, 82, + 1, 343310, 1, 1064207, 2, 3145813, 1, 15728, 1, 7817, 1, 15729, + 3, 15730, 12, 92, 2, 98, 1, 16168, 1, 16167, 1, 16002, 1, 16008, + 1, 15974, 1, 15975, 1, 15990, 1, 15976, 1, 15977, 1, 15980, 0, + 15981, 1, 10240, 1, 5253120, 1, 15720, 1, 198656, 6, 110, 1, + 180224, 1, 103824738, 2, 112, 2, 3145839, 0, 536885440, 1, + 114880, 14, 125, 12, 206975, 1, 33559995, 12, 198784, 0, + 33570236, 1, 15803, 0, 15804, 3, 294912, 1, 294912, 3, 442370, + 1, 11544576, 0, 811612160, 1, 12593152, 1, 11536384, 1, + 14024704, 7, 310382726, 0, 10240, 1, 14796, 1, 14797, 1, 14793, + 1, 14794, 0, 14795, 1, 268679168, 1, 9437184, 1, 268449792, 1, + 198656, 1, 9452827, 1, 1075854602, 1, 1075854603, 1, 557056, 1, + 114880, 14, 159, 12, 198784, 1, 1109409213, 12, 198783, 1, + 1107312059, 12, 198784, 1, 1109409212, 2, 162, 1, 1075854781, 1, + 1073757627, 1, 1075854780, 1, 540672, 1, 10485760, 6, 3145894, + 16, 274741248, 9, 168, 3, 4194304, 3, 4209949, 0, 0, 0, 256, 14, + 174, 1, 114857, 1, 33560007, 12, 176, 0, 10240, 1, 114858, 1, + 33560018, 1, 114857, 3, 33560007, 1, 16008, 1, 114874, 1, + 33560360, 1, 114875, 1, 33560154, 0, 15963, 0, 256, 0, 4096, 1, + 409611, 9, 188, 0, 10240, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 +}; + + +#define DO_REMAP(_m) (_m)->handle = drm_ioremap((_m)->offset, (_m)->size) + +#define DO_REMAPFREE(_m) \ + do { \ + if ((_m)->handle && (_m)->size) \ + drm_ioremapfree((_m)->handle, (_m)->size); \ + } while (0) + +#define DO_FIND_MAP(_m, _o) \ + do { \ + int _i; \ + for (_i = 0; _i < dev->map_count; _i++) { \ + if (dev->maplist[_i]->offset == _o) { \ + _m = dev->maplist[_i]; \ + break; \ + } \ + } \ + } while (0) + + +int R128_READ_PLL(drm_device_t *dev, int addr) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + + R128_WRITE8(R128_CLOCK_CNTL_INDEX, addr & 0x1f); + return R128_READ(R128_CLOCK_CNTL_DATA); +} + + +static void r128_status( drm_r128_private_t *dev_priv ) +{ + printk( "GUI_STAT = 0x%08x\n", + (unsigned int)R128_READ( R128_GUI_STAT ) ); + printk( "PM4_STAT = 0x%08x\n", + (unsigned int)R128_READ( R128_PM4_STAT ) ); + printk( "PM4_BUFFER_DL_WPTR = 0x%08x\n", + (unsigned int)R128_READ( R128_PM4_BUFFER_DL_WPTR ) ); + printk( "PM4_BUFFER_DL_RPTR = 0x%08x\n", + (unsigned int)R128_READ( R128_PM4_BUFFER_DL_RPTR ) ); + printk( "PM4_MICRO_CNTL = 0x%08x\n", + (unsigned int)R128_READ( R128_PM4_MICRO_CNTL ) ); + printk( "PM4_BUFFER_CNTL = 0x%08x\n", + (unsigned int)R128_READ( R128_PM4_BUFFER_CNTL ) ); +} + + +/* ================================================================ + * Engine, FIFO control + */ + +static int r128_do_pixcache_flush( drm_r128_private_t *dev_priv ) +{ + u32 tmp; + int i; + + tmp = R128_READ( R128_PC_NGUI_CTLSTAT ) | R128_PC_FLUSH_ALL; + R128_WRITE( R128_PC_NGUI_CTLSTAT, tmp ); + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + if ( !(R128_READ( R128_PC_NGUI_CTLSTAT ) & R128_PC_BUSY) ) { + return 0; + } + udelay( 1 ); + } + + DRM_ERROR( "%s failed!\n", __FUNCTION__ ); + return -EBUSY; +} + +static int r128_do_wait_for_fifo( drm_r128_private_t *dev_priv, int entries ) +{ + int i; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + int slots = R128_READ( R128_GUI_STAT ) & R128_GUI_FIFOCNT_MASK; + if ( slots >= entries ) return 0; + udelay( 1 ); + } + + DRM_ERROR( "%s failed!\n", __FUNCTION__ ); + return -EBUSY; +} + +static int r128_do_wait_for_idle( drm_r128_private_t *dev_priv ) +{ + int i, ret; + + ret = r128_do_wait_for_fifo( dev_priv, 64 ); + if ( !ret ) return ret; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + if ( !(R128_READ( R128_GUI_STAT ) & R128_GUI_ACTIVE) ) { + r128_do_pixcache_flush( dev_priv ); + return 0; + } + udelay( 1 ); + } + + DRM_ERROR( "%s failed!\n", __FUNCTION__ ); + return -EBUSY; +} + + +/* ================================================================ + * CCE control, initialization + */ + +/* Load the microcode for the CCE */ +static void r128_cce_load_microcode( drm_r128_private_t *dev_priv ) +{ + int i; + + r128_do_wait_for_idle( dev_priv ); + + R128_WRITE( R128_PM4_MICROCODE_ADDR, 0 ); + for ( i = 0 ; i < 256 ; i++ ) { + R128_WRITE( R128_PM4_MICROCODE_DATAH, + r128_cce_microcode[i * 2] ); + R128_WRITE( R128_PM4_MICROCODE_DATAL, + r128_cce_microcode[i * 2 + 1] ); + } +} + +/* Flush any pending commands to the CCE. 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 r128_do_cce_flush( drm_r128_private_t *dev_priv ) +{ + u32 tmp; + + tmp = R128_READ( R128_PM4_BUFFER_DL_WPTR ) | R128_PM4_BUFFER_DL_DONE; + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, tmp ); +} + +/* Wait for the CCE to go idle. + */ +static int r128_do_cce_idle( drm_r128_private_t *dev_priv ) +{ + int i; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + if ( *dev_priv->ring.head == dev_priv->ring.tail ) { + int pm4stat = R128_READ( R128_PM4_STAT ); + if ( ( (pm4stat & R128_PM4_FIFOCNT_MASK) >= + dev_priv->cce_fifo_size ) && + !(pm4stat & (R128_PM4_BUSY | + R128_PM4_GUI_ACTIVE)) ) { + return r128_do_pixcache_flush( dev_priv ); + } + } + udelay( 1 ); + } + +#if 0 + DRM_ERROR( "failed!\n" ); + r128_status( dev_priv ); +#endif + return -EBUSY; +} + +/* Start the Concurrent Command Engine. + */ +static void r128_do_cce_start( drm_r128_private_t *dev_priv ) +{ + r128_do_wait_for_idle( dev_priv ); + + R128_WRITE( R128_PM4_BUFFER_CNTL, + dev_priv->cce_mode | dev_priv->ring.size_l2qw ); + R128_READ( R128_PM4_BUFFER_ADDR ); /* as per the sample code */ + R128_WRITE( R128_PM4_MICRO_CNTL, R128_PM4_MICRO_FREERUN ); + + dev_priv->cce_running = 1; +} + +/* Reset the Concurrent Command Engine. This will not flush any pending + * commangs, so you must wait for the CCE command stream to complete + * before calling this routine. + */ +static void r128_do_cce_reset( drm_r128_private_t *dev_priv ) +{ + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, 0 ); + R128_WRITE( R128_PM4_BUFFER_DL_RPTR, 0 ); + *dev_priv->ring.head = 0; + dev_priv->ring.tail = 0; +} + +/* Stop the Concurrent Command Engine. This will not flush any pending + * commangs, so you must flush the command stream and wait for the CCE + * to go idle before calling this routine. + */ +static void r128_do_cce_stop( drm_r128_private_t *dev_priv ) +{ + R128_WRITE( R128_PM4_MICRO_CNTL, 0 ); + R128_WRITE( R128_PM4_BUFFER_CNTL, R128_PM4_NONPM4 ); + + dev_priv->cce_running = 0; +} + +/* Reset the engine. This will stop the CCE if it is running. + */ +static int r128_do_engine_reset( drm_device_t *dev ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + u32 clock_cntl_index, mclk_cntl, gen_reset_cntl; + + r128_do_pixcache_flush( dev_priv ); + + clock_cntl_index = R128_READ( R128_CLOCK_CNTL_INDEX ); + mclk_cntl = R128_READ_PLL( dev, R128_MCLK_CNTL ); + + R128_WRITE_PLL( R128_MCLK_CNTL, + mclk_cntl | R128_FORCE_GCP | R128_FORCE_PIPE3D_CP ); + + gen_reset_cntl = R128_READ( R128_GEN_RESET_CNTL ); + + /* Taken from the sample code - do not change */ + R128_WRITE( R128_GEN_RESET_CNTL, + gen_reset_cntl | R128_SOFT_RESET_GUI ); + R128_READ( R128_GEN_RESET_CNTL ); + R128_WRITE( R128_GEN_RESET_CNTL, + gen_reset_cntl & ~R128_SOFT_RESET_GUI ); + R128_READ( R128_GEN_RESET_CNTL ); + + R128_WRITE_PLL( R128_MCLK_CNTL, mclk_cntl ); + R128_WRITE( R128_CLOCK_CNTL_INDEX, clock_cntl_index ); + R128_WRITE( R128_GEN_RESET_CNTL, gen_reset_cntl ); + + /* Reset the CCE ring */ + r128_do_cce_reset( dev_priv ); + + /* The CCE is no longer running after an engine reset */ + dev_priv->cce_running = 0; + + /* Reset any pending vertex, indirect buffers */ + r128_freelist_reset( dev ); + + return 0; +} + +static void r128_cce_init_ring_buffer( drm_device_t *dev ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + u32 ring_start; + u32 tmp; + + /* The manual (p. 2) says this address is in "VM space". This + * means it's an offset from the start of AGP space. + */ + ring_start = dev_priv->cce_ring->offset - dev->agp->base; + R128_WRITE( R128_PM4_BUFFER_OFFSET, ring_start | R128_AGP_OFFSET ); + + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, 0 ); + R128_WRITE( R128_PM4_BUFFER_DL_RPTR, 0 ); + + /* DL_RPTR_ADDR is a physical address in AGP space. */ + *dev_priv->ring.head = 0; + R128_WRITE( R128_PM4_BUFFER_DL_RPTR_ADDR, + dev_priv->ring_rptr->offset ); + + /* Set watermark control */ + R128_WRITE( R128_PM4_BUFFER_WM_CNTL, + ((R128_WATERMARK_L/4) << R128_WMA_SHIFT) + | ((R128_WATERMARK_M/4) << R128_WMB_SHIFT) + | ((R128_WATERMARK_N/4) << R128_WMC_SHIFT) + | ((R128_WATERMARK_K/64) << R128_WB_WM_SHIFT) ); + + /* Force read. Why? Because it's in the examples... */ + R128_READ( R128_PM4_BUFFER_ADDR ); + + /* Turn on bus mastering */ + tmp = R128_READ( R128_BUS_CNTL ) & ~R128_BUS_MASTER_DIS; + R128_WRITE( R128_BUS_CNTL, tmp ); +} + +static int r128_do_init_cce( drm_device_t *dev, drm_r128_init_t *init ) +{ + drm_r128_private_t *dev_priv; + int i; + + dev_priv = drm_alloc( sizeof(drm_r128_private_t), DRM_MEM_DRIVER ); + if ( dev_priv == NULL ) + return -ENOMEM; + dev->dev_private = (void *)dev_priv; + + memset( dev_priv, 0, sizeof(drm_r128_private_t) ); + + dev_priv->is_pci = init->is_pci; + + /* GH: We don't support PCI cards until PCI GART is implemented. + * Fail here so we can remove all checks for PCI cards around + * the CCE ring code. + */ + if ( dev_priv->is_pci ) { + drm_free( dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER ); + dev->dev_private = NULL; + return -EINVAL; + } + + dev_priv->usec_timeout = init->usec_timeout; + if ( dev_priv->usec_timeout < 1 || + dev_priv->usec_timeout > R128_MAX_USEC_TIMEOUT ) { + drm_free( dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER ); + dev->dev_private = NULL; + return -EINVAL; + } + + dev_priv->cce_mode = init->cce_mode; + dev_priv->cce_secure = init->cce_secure; + + /* GH: Simple idle check. + */ + atomic_set( &dev_priv->idle_count, 0 ); + + /* 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->cce_mode != R128_PM4_192BM ) && + ( init->cce_mode != R128_PM4_128BM_64INDBM ) && + ( init->cce_mode != R128_PM4_64BM_128INDBM ) && + ( init->cce_mode != R128_PM4_64BM_64VCBM_64INDBM ) ) { + drm_free( dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER ); + dev->dev_private = NULL; + return -EINVAL; + } + + switch ( init->cce_mode ) { + case R128_PM4_NONPM4: + dev_priv->cce_fifo_size = 0; + break; + case R128_PM4_192PIO: + case R128_PM4_192BM: + dev_priv->cce_fifo_size = 192; + break; + case R128_PM4_128PIO_64INDBM: + case R128_PM4_128BM_64INDBM: + dev_priv->cce_fifo_size = 128; + break; + case R128_PM4_64PIO_128INDBM: + case R128_PM4_64BM_128INDBM: + case R128_PM4_64PIO_64VCBM_64INDBM: + case R128_PM4_64BM_64VCBM_64INDBM: + case R128_PM4_64PIO_64VCPIO_64INDPIO: + dev_priv->cce_fifo_size = 64; + break; + } + + dev_priv->fb_bpp = init->fb_bpp; + 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; + + dev_priv->depth_bpp = init->depth_bpp; + dev_priv->depth_offset = init->depth_offset; + dev_priv->depth_pitch = init->depth_pitch; + dev_priv->span_offset = init->span_offset; + + dev_priv->front_pitch_offset_c = (((dev_priv->front_pitch/8) << 21) | + (dev_priv->front_offset >> 5)); + dev_priv->back_pitch_offset_c = (((dev_priv->back_pitch/8) << 21) | + (dev_priv->back_offset >> 5)); + dev_priv->depth_pitch_offset_c = (((dev_priv->depth_pitch/8) << 21) | + (dev_priv->depth_offset >> 5) | + R128_DST_TILE); + dev_priv->span_pitch_offset_c = (((dev_priv->depth_pitch/8) << 21) | + (dev_priv->span_offset >> 5)); + + /* FIXME: We want multiple shared areas, including one shared + * only by the X Server and kernel module. + */ + for ( i = 0 ; i < dev->map_count ; i++ ) { + if ( dev->maplist[i]->type == _DRM_SHM ) { + dev_priv->sarea = dev->maplist[i]; + break; + } + } + + DO_FIND_MAP( dev_priv->fb, init->fb_offset ); + DO_FIND_MAP( dev_priv->mmio, init->mmio_offset ); + DO_FIND_MAP( dev_priv->cce_ring, init->ring_offset ); + DO_FIND_MAP( dev_priv->ring_rptr, init->ring_rptr_offset ); + DO_FIND_MAP( dev_priv->buffers, init->buffers_offset ); + + if ( !dev_priv->is_pci ) { + DO_FIND_MAP( dev_priv->agp_textures, + init->agp_textures_offset ); + } + + dev_priv->sarea_priv = + (drm_r128_sarea_t *)((u8 *)dev_priv->sarea->handle + + init->sarea_priv_offset); + + DO_REMAP( dev_priv->cce_ring ); + DO_REMAP( dev_priv->ring_rptr ); + DO_REMAP( dev_priv->buffers ); +#if 0 + if ( !dev_priv->is_pci ) { + DO_REMAP( dev_priv->agp_textures ); + } +#endif + + dev_priv->ring.head = ((__volatile__ u32 *) + dev_priv->ring_rptr->handle); + + dev_priv->ring.start = (u32 *)dev_priv->cce_ring->handle; + dev_priv->ring.end = ((u32 *)dev_priv->cce_ring->handle + + init->ring_size / sizeof(u32)); + dev_priv->ring.size = init->ring_size; + dev_priv->ring.size_l2qw = drm_order( init->ring_size / 8 ); + + dev_priv->ring.tail_mask = + (dev_priv->ring.size / sizeof(u32)) - 1; + + dev_priv->sarea_priv->last_frame = 0; + R128_WRITE( R128_LAST_FRAME_REG, dev_priv->sarea_priv->last_frame ); + + dev_priv->sarea_priv->last_dispatch = 0; + R128_WRITE( R128_LAST_DISPATCH_REG, + dev_priv->sarea_priv->last_dispatch ); + + r128_cce_init_ring_buffer( dev ); + r128_cce_load_microcode( dev_priv ); + r128_do_engine_reset( dev ); + + return 0; +} + +static int r128_do_cleanup_cce( drm_device_t *dev ) +{ + if ( dev->dev_private ) { + drm_r128_private_t *dev_priv = dev->dev_private; + + DO_REMAPFREE( dev_priv->cce_ring ); + DO_REMAPFREE( dev_priv->ring_rptr ); + DO_REMAPFREE( dev_priv->buffers ); +#if 0 + if ( !dev_priv->is_pci ) { + DO_REMAPFREE( dev_priv->agp_textures ); + } +#endif + + drm_free( dev->dev_private, sizeof(drm_r128_private_t), + DRM_MEM_DRIVER ); + dev->dev_private = NULL; + } + + return 0; +} + +int r128_cce_init( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_init_t init; + + if ( copy_from_user( &init, (drm_r128_init_t *)arg, sizeof(init) ) ) + return -EFAULT; + + switch ( init.func ) { + case R128_INIT_CCE: + return r128_do_init_cce( dev, &init ); + case R128_CLEANUP_CCE: + return r128_do_cleanup_cce( dev ); + } + + return -EINVAL; +} + +int r128_cce_start( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + if ( dev_priv->cce_running || dev_priv->cce_mode == R128_PM4_NONPM4 ) { + DRM_DEBUG( "%s while CCE running\n", __FUNCTION__ ); + return 0; + } + + r128_do_cce_start( dev_priv ); + + return 0; +} + +/* Stop the CCE. The engine must have been idled before calling this + * routine. + */ +int r128_cce_stop( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_cce_stop_t stop; + int ret; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &stop, (drm_r128_init_t *)arg, sizeof(stop) ) ) + return -EFAULT; + + /* Flush any pending CCE commands. This ensures any outstanding + * commands are exectuted by the engine before we turn it off. + */ + if ( stop.flush ) { + r128_do_cce_flush( dev_priv ); + } + + /* If we fail to make the engine go idle, we return an error + * code so that the DRM ioctl wrapper can try again. + */ + if ( stop.idle ) { + ret = r128_do_cce_idle( dev_priv ); + if ( ret < 0 ) return ret; + } + + /* Finally, we can turn off the CCE. If the engine isn't idle, + * we will get some dropped triangles as they won't be fully + * rendered before the CCE is shut down. + */ + r128_do_cce_stop( dev_priv ); + + /* Reset the engine */ + r128_do_engine_reset( dev ); + + return 0; +} + +/* Just reset the CCE ring. Called as part of an X Server engine reset. + */ +int r128_cce_reset( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + if ( !dev_priv ) { + DRM_DEBUG( "%s called before init done\n", __FUNCTION__ ); + return -EINVAL; + } + + r128_do_cce_reset( dev_priv ); + + /* The CCE is no longer running after an engine reset */ + dev_priv->cce_running = 0; + + return 0; +} + +int r128_cce_idle( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( dev_priv->cce_running ) { + r128_do_cce_flush( dev_priv ); + } + + return r128_do_cce_idle( dev_priv ); +} + +int r128_engine_reset( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + return r128_do_engine_reset( dev ); +} + + +/* ================================================================ + * Freelist management + */ +#define R128_BUFFER_USED 0xffffffff +#define R128_BUFFER_FREE 0 + +static int r128_freelist_init( drm_device_t *dev ) +{ + drm_device_dma_t *dma = dev->dma; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_buf_t *buf; + drm_r128_buf_priv_t *buf_priv; + drm_r128_freelist_t *entry; + int i; + + dev_priv->head = drm_alloc( sizeof(drm_r128_freelist_t), + DRM_MEM_DRIVER ); + if ( dev_priv->head == NULL ) + return -ENOMEM; + + memset( dev_priv->head, 0, sizeof(drm_r128_freelist_t) ); + dev_priv->head->age = R128_BUFFER_USED; + + for ( i = 0 ; i < dma->buf_count ; i++ ) { + buf = dma->buflist[i]; + buf_priv = buf->dev_private; + + entry = drm_alloc( sizeof(drm_r128_freelist_t), + DRM_MEM_DRIVER ); + if ( !entry ) return -ENOMEM; + + entry->age = R128_BUFFER_FREE; + entry->buf = buf; + entry->prev = dev_priv->head; + entry->next = dev_priv->head->next; + if ( !entry->next ) + dev_priv->tail = entry; + + buf_priv->discard = 0; + buf_priv->dispatched = 0; + buf_priv->list_entry = entry; + + dev_priv->head->next = entry; + + if ( dev_priv->head->next ) + dev_priv->head->next->prev = entry; + } + + return 0; + +} + +drm_buf_t *r128_freelist_get( drm_device_t *dev ) +{ + drm_device_dma_t *dma = dev->dma; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_buf_priv_t *buf_priv; + drm_buf_t *buf; + int i, t; + + /* FIXME: Optimize -- use freelist code */ + + for ( i = 0 ; i < dma->buf_count ; i++ ) { + buf = dma->buflist[i]; + buf_priv = buf->dev_private; + if ( buf->pid == 0 ) + return buf; + } + + for ( t = 0 ; t < dev_priv->usec_timeout ; t++ ) { + u32 done_age = R128_READ( R128_LAST_DISPATCH_REG ); + + for ( i = 0 ; i < dma->buf_count ; i++ ) { + buf = dma->buflist[i]; + buf_priv = buf->dev_private; + if ( buf->pending && buf_priv->age <= done_age ) { + /* The buffer has been processed, so it + * can now be used. + */ + buf->pending = 0; + return buf; + } + } + udelay( 1 ); + } + + DRM_ERROR( "returning NULL!\n" ); + return NULL; +} + +void r128_freelist_reset( drm_device_t *dev ) +{ + drm_device_dma_t *dma = dev->dma; + int i; + + for ( i = 0 ; i < dma->buf_count ; i++ ) { + drm_buf_t *buf = dma->buflist[i]; + drm_r128_buf_priv_t *buf_priv = buf->dev_private; + buf_priv->age = 0; + } +} + + +/* ================================================================ + * CCE packet submission + */ + +int r128_wait_ring( drm_r128_private_t *dev_priv, int n ) +{ + drm_r128_ring_buffer_t *ring = &dev_priv->ring; + int i; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + ring->space = *ring->head - ring->tail; + if ( ring->space <= 0 ) + ring->space += ring->size; + + if ( ring->space >= n ) + return 0; + + udelay( 1 ); + } + + return -EBUSY; +} + +void r128_update_ring_snapshot( drm_r128_private_t *dev_priv ) +{ + drm_r128_ring_buffer_t *ring = &dev_priv->ring; + + ring->space = *ring->head - ring->tail; +#if R128_PERFORMANCE_BOXES + if ( ring->space == 0 ) + atomic_inc( &dev_priv->idle_count ); +#endif + if ( ring->space <= 0 ) + ring->space += ring->size; +} + +static int r128_verify_command( drm_r128_private_t *dev_priv, + u32 cmd, int *size ) +{ + int writing = 1; + + *size = 0; + + switch ( cmd & R128_CCE_PACKET_MASK ) { + case R128_CCE_PACKET0: + if ( (cmd & R128_CCE_PACKET0_REG_MASK) <= (0x1004 >> 2) && + (cmd & R128_CCE_PACKET0_REG_MASK) != + (R128_PM4_VC_FPU_SETUP >> 2) ) { + writing = 0; + } + *size = ((cmd & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; + break; + + case R128_CCE_PACKET1: + if ( (cmd & R128_CCE_PACKET1_REG0_MASK) <= (0x1004 >> 2) && + (cmd & R128_CCE_PACKET1_REG0_MASK) != + (R128_PM4_VC_FPU_SETUP >> 2) ) { + writing = 0; + } + if ( (cmd & R128_CCE_PACKET1_REG1_MASK) <= (0x1004 << 9) && + (cmd & R128_CCE_PACKET1_REG1_MASK) != + (R128_PM4_VC_FPU_SETUP << 9) ) { + writing = 0; + } + *size = 3; + break; + + case R128_CCE_PACKET2: + break; + + case R128_CCE_PACKET3: + *size = ((cmd & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; + break; + + } + + return writing; +} + +static int r128_submit_packet_ring_secure( drm_r128_private_t *dev_priv, + u32 *commands, int *count ) +{ +#if 0 + int write = dev_priv->sarea_priv->ring_write; + int *write_ptr = dev_priv->ring_start + write; + int c = *count; + u32 tmp = 0; + int psize = 0; + int writing = 1; + int timeout; + + while ( c > 0 ) { + tmp = *commands++; + if ( !psize ) { + writing = r128_verify_command( dev_priv, tmp, &psize ); + } + psize--; + + if ( writing ) { + write++; + *write_ptr++ = tmp; + } + if ( write >= dev_priv->ring_entries ) { + write = 0; + write_ptr = dev_priv->ring_start; + } + timeout = 0; + while ( write == *dev_priv->ring_read_ptr ) { + R128_READ( R128_PM4_BUFFER_DL_RPTR ); + if ( timeout++ >= dev_priv->usec_timeout ) + return -EBUSY; + udelay( 1 ); + } + c--; + } + + if ( write < 32 ) { + memcpy( dev_priv->ring_end, + dev_priv->ring_start, + write * sizeof(u32) ); + } + + /* Make sure WC cache has been flushed */ + r128_flush_write_combine(); + + dev_priv->sarea_priv->ring_write = write; + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, write ); + + *count = 0; +#endif + return 0; +} + +static int r128_submit_packet_ring_insecure( drm_r128_private_t *dev_priv, + u32 *commands, int *count ) +{ +#if 0 + int write = dev_priv->sarea_priv->ring_write; + int *write_ptr = dev_priv->ring_start + write; + int c = *count; + int timeout; + + while ( c > 0 ) { + write++; + *write_ptr++ = *commands++; + if ( write >= dev_priv->ring_entries ) { + write = 0; + write_ptr = dev_priv->ring_start; + } + + timeout = 0; + while ( write == *dev_priv->ring_read_ptr ) { + R128_READ( R128_PM4_BUFFER_DL_RPTR ); + if ( timeout++ >= dev_priv->usec_timeout ) + return -EBUSY; + udelay( 1 ); + } + c--; + } + + if ( write < 32 ) { + memcpy( dev_priv->ring_end, + dev_priv->ring_start, + write * sizeof(u32) ); + } + + /* Make sure WC cache has been flushed */ + r128_flush_write_combine(); + + dev_priv->sarea_priv->ring_write = write; + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, write ); + + *count = 0; +#endif + return 0; +} + +/* Internal packet submission routine. This uses the insecure versions + * of the packet submission functions, and thus should only be used for + * packets generated inside the kernel module. + */ +int r128_do_submit_packet( drm_r128_private_t *dev_priv, + u32 *buffer, int count ) +{ + int c = count; + int ret = 0; + +#if 0 + int left = 0; + + if ( c >= dev_priv->ring_entries ) { + c = dev_priv->ring_entries - 1; + left = count - c; + } + + /* Since this is only used by the kernel we can use the + * insecure ring buffer submit packet routine. + */ + ret = r128_submit_packet_ring_insecure( dev_priv, buffer, &c ); + c += left; +#endif + + return ( ret < 0 ) ? ret : c; +} + +/* External packet submission routine. This uses the secure versions + * by default, and can thus submit packets received from user space. + */ +int r128_cce_packet( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_packet_t packet; + u32 *buffer; + int c; + int size; + int ret = 0; + +#if 0 + /* GH: Disable packet submission for now. + */ + return -EINVAL; +#endif + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "r128_submit_packet called without lock held\n" ); + return -EINVAL; + } + + if ( copy_from_user( &packet, (drm_r128_packet_t *)arg, + sizeof(packet) ) ) + return -EFAULT; + +#if 0 + c = packet.count; + size = c * sizeof(*buffer); + + { + int left = 0; + + if ( c >= dev_priv->ring_entries ) { + c = dev_priv->ring_entries - 1; + size = c * sizeof(*buffer); + left = packet.count - c; + } + + buffer = kmalloc( size, 0 ); + if ( buffer == NULL) + return -ENOMEM; + if ( copy_from_user( buffer, packet.buffer, size ) ) + return -EFAULT; + + if ( dev_priv->cce_secure ) { + ret = r128_submit_packet_ring_secure( dev_priv, + buffer, &c ); + } else { + ret = r128_submit_packet_ring_insecure( dev_priv, + buffer, &c ); + } + c += left; + } + + kfree( buffer ); +#else + c = 0; +#endif + + packet.count = c; + if ( copy_to_user( (drm_r128_packet_t *)arg, &packet, + sizeof(packet) ) ) + return -EFAULT; + + if ( ret ) { + return ret; + } else if ( c > 0 ) { + return -EAGAIN; + } + return 0; +} + +#if 0 +static int r128_send_vertbufs( drm_device_t *dev, drm_r128_vertex_t *v ) +{ + drm_device_dma_t *dma = dev->dma; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_buf_priv_t *buf_priv; + drm_buf_t *buf; + int i, ret; + RING_LOCALS; + + /* Make sure we have valid data */ + for (i = 0; i < v->send_count; i++) { + int idx = v->send_indices[i]; + + if (idx < 0 || idx >= dma->buf_count) { + DRM_ERROR("Index %d (of %d max)\n", + idx, dma->buf_count - 1); + return -EINVAL; + } + buf = dma->buflist[idx]; + if (buf->pid != current->pid) { + DRM_ERROR("Process %d using buffer owned by %d\n", + current->pid, buf->pid); + return -EINVAL; + } + if (buf->pending) { + DRM_ERROR("Sending pending buffer:" + " buffer %d, offset %d\n", + v->send_indices[i], i); + return -EINVAL; + } + } + + /* Wait for idle, if we've wrapped to make sure that all pending + buffers have been processed */ + if (dev_priv->submit_age == R128_MAX_VBUF_AGE) { + if ((ret = r128_do_cce_idle(dev)) < 0) return ret; + dev_priv->submit_age = 0; + r128_freelist_reset(dev); + } + + /* Make sure WC cache has been flushed (if in PIO mode) */ + if (!dev_priv->cce_is_bm_mode) r128_flush_write_combine(); + + /* FIXME: Add support for sending vertex buffer to the CCE here + instead of in client code. The v->prim holds the primitive + type that should be drawn. Loop over the list buffers in + send_indices[] and submit a packet for each VB. + + This will require us to loop over the clip rects here as + well, which implies that we extend the kernel driver to allow + cliprects to be stored here. Note that the cliprects could + possibly come from the X server instead of the client, but + this will require additional changes to the DRI to allow for + this optimization. */ + + /* Submit a CCE packet that writes submit_age to R128_VB_AGE_REG */ +#if 0 + cce_buffer[0] = R128CCE0(R128_CCE_PACKET0, R128_VB_AGE_REG, 0); + cce_buffer[1] = dev_priv->submit_age; + + if ((ret = r128_do_submit_packet(dev, cce_buffer, 2)) < 0) { + /* Until we add support for sending VBs to the CCE in + this routine, we can recover from this error. After + we add that support, we won't be able to easily + recover, so we will probably have to implement + another mechanism for handling timeouts from packets + submitted directly by the kernel. */ + return ret; + } +#else + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_VB_AGE_REG, 0 ) ); + OUT_RING( dev_priv->submit_age ); + + ADVANCE_RING(); +#endif + /* Now that the submit packet request has succeeded, we can mark + the buffers as pending */ + for (i = 0; i < v->send_count; i++) { + buf = dma->buflist[v->send_indices[i]]; + buf->pending = 1; + + buf_priv = buf->dev_private; + buf_priv->age = dev_priv->submit_age; + } + + dev_priv->submit_age++; + + return 0; +} +#endif + + + + +static int r128_cce_get_buffers( drm_device_t *dev, drm_dma_t *d ) +{ + int i; + drm_buf_t *buf; + + for ( i = d->granted_count ; i < d->request_count ; i++ ) { + buf = r128_freelist_get( dev ); + if ( !buf ) return -EAGAIN; + + buf->pid = current->pid; + + if ( copy_to_user( &d->request_indices[i], &buf->idx, + sizeof(buf->idx) ) ) + return -EFAULT; + if ( copy_to_user( &d->request_sizes[i], &buf->total, + sizeof(buf->total) ) ) + return -EFAULT; + + d->granted_count++; + } + return 0; +} + +int r128_cce_buffers( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_device_dma_t *dma = dev->dma; + int ret = 0; + drm_dma_t d; + + if ( copy_from_user( &d, (drm_dma_t *) arg, sizeof(d) ) ) + return -EFAULT; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + /* Please don't send us buffers. + */ + if ( d.send_count != 0 ) { + DRM_ERROR( "Process %d trying to send %d buffers via drmDMA\n", + current->pid, d.send_count ); + return -EINVAL; + } + + /* We'll send you buffers. + */ + if ( d.request_count < 0 || d.request_count > dma->buf_count ) { + DRM_ERROR( "Process %d trying to get %d buffers (of %d max)\n", + current->pid, d.request_count, dma->buf_count ); + return -EINVAL; + } + + d.granted_count = 0; + + if ( d.request_count ) { + ret = r128_cce_get_buffers( dev, &d ); + } + + if ( copy_to_user( (drm_dma_t *) arg, &d, sizeof(d) ) ) + return -EFAULT; + + return ret; +} diff --git a/linux/r128_context.c b/linux/r128_context.c index 9cadadba..0741e774 100644 --- a/linux/r128_context.c +++ b/linux/r128_context.c @@ -11,11 +11,11 @@ * 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 (including the next * paragraph) 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 @@ -23,7 +23,7 @@ * 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. - * + * * Author: Rickard E. (Rik) Faith * */ @@ -53,21 +53,21 @@ int r128_context_switch(drm_device_t *dev, int old, int new) #if DRM_DMA_HISTOGRAM dev->ctx_start = get_cycles(); #endif - + DRM_DEBUG("Context switch from %d to %d\n", old, new); if (new == dev->last_context) { clear_bit(0, &dev->context_flag); return 0; } - + if (drm_flags & DRM_FLAG_NOCTX) { r128_context_switch_complete(dev, new); } else { sprintf(buf, "C %d %d\n", old, new); drm_write_string(dev, buf); } - + return 0; } @@ -75,7 +75,7 @@ int r128_context_switch_complete(drm_device_t *dev, int new) { dev->last_context = new; /* PRE/POST: This is the _only_ writer. */ dev->last_switch = jiffies; - + if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) { DRM_ERROR("Lock isn't held after context switch\n"); } @@ -86,11 +86,11 @@ int r128_context_switch_complete(drm_device_t *dev, int new) #if DRM_DMA_HISTOGRAM atomic_inc(&dev->histo.ctx[drm_histogram_slot(get_cycles() - dev->ctx_start)]); - + #endif clear_bit(0, &dev->context_flag); wake_up(&dev->context_wait); - + return 0; } diff --git a/linux/r128_dma.c b/linux/r128_dma.c deleted file mode 100644 index bcba6782..00000000 --- a/linux/r128_dma.c +++ /dev/null @@ -1,909 +0,0 @@ -/* r128_drv.c -- ATI Rage 128 driver -*- linux-c -*- - * Created: Wed Apr 5 19:24:19 2000 by kevin@precisioninsight.com - * - * Copyright 2000 Precision Insight, Inc., Cedar Park, Texas. - * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California. - * All Rights Reserved. - * - * 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 (including the next - * paragraph) 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 - * PRECISION INSIGHT AND/OR ITS SUPPLIERS 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. - * - * Authors: Kevin E. Martin - * - */ - -#define __NO_VERSION__ -#include "drmP.h" -#include "r128_drv.h" - -#include /* For task queue support */ -#include - - - -#define DO_REMAP(_m) (_m)->handle = drm_ioremap((_m)->offset, (_m)->size) - -#define DO_REMAPFREE(_m) \ - do { \ - if ((_m)->handle && (_m)->size) \ - drm_ioremapfree((_m)->handle, (_m)->size); \ - } while (0) - -#define DO_FIND_MAP(_m, _o) \ - do { \ - int _i; \ - for (_i = 0; _i < dev->map_count; _i++) { \ - if (dev->maplist[_i]->offset == _o) { \ - _m = dev->maplist[_i]; \ - break; \ - } \ - } \ - } while (0) - - -#define R128_MAX_VBUF_AGE 0x10000000 -#define R128_VB_AGE_REG R128_GUI_SCRATCH_REG0 - -int R128_READ_PLL(drm_device_t *dev, int addr) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - - R128_WRITE8(R128_CLOCK_CNTL_INDEX, addr & 0x1f); - return R128_READ(R128_CLOCK_CNTL_DATA); -} - -#define r128_flush_write_combine() mb() - - -static void r128_status(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - - printk("GUI_STAT = 0x%08x\n", - (unsigned int)R128_READ(R128_GUI_STAT)); - printk("PM4_STAT = 0x%08x\n", - (unsigned int)R128_READ(R128_PM4_STAT)); - printk("PM4_BUFFER_DL_WPTR = 0x%08x\n", - (unsigned int)R128_READ(R128_PM4_BUFFER_DL_WPTR)); - printk("PM4_BUFFER_DL_RPTR = 0x%08x\n", - (unsigned int)R128_READ(R128_PM4_BUFFER_DL_RPTR)); -} - -static int r128_do_cleanup_cce(drm_device_t *dev) -{ - if (dev->dev_private) { - drm_r128_private_t *dev_priv = dev->dev_private; - - if (!dev_priv->is_pci) { - DO_REMAPFREE(dev_priv->agp_ring); - DO_REMAPFREE(dev_priv->agp_read_ptr); - DO_REMAPFREE(dev_priv->agp_vertbufs); - DO_REMAPFREE(dev_priv->agp_indbufs); - DO_REMAPFREE(dev_priv->agp_textures); - } - - drm_free(dev->dev_private, sizeof(drm_r128_private_t), - DRM_MEM_DRIVER); - dev->dev_private = NULL; - } - - return 0; -} - -static int r128_do_init_cce(drm_device_t *dev, drm_r128_init_t *init) -{ - drm_r128_private_t *dev_priv; - int i; - - dev_priv = drm_alloc(sizeof(drm_r128_private_t), DRM_MEM_DRIVER); - if (dev_priv == NULL) return -ENOMEM; - dev->dev_private = (void *)dev_priv; - - memset(dev_priv, 0, sizeof(drm_r128_private_t)); - - dev_priv->is_pci = init->is_pci; - - dev_priv->usec_timeout = init->usec_timeout; - if (dev_priv->usec_timeout < 1 || - dev_priv->usec_timeout > R128_MAX_USEC_TIMEOUT) { - drm_free(dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER); - dev->dev_private = NULL; - return -EINVAL; - } - - dev_priv->cce_mode = init->cce_mode; - dev_priv->cce_fifo_size = init->cce_fifo_size; - dev_priv->cce_is_bm_mode = - ((init->cce_mode == R128_PM4_192BM) || - (init->cce_mode == R128_PM4_128BM_64INDBM) || - (init->cce_mode == R128_PM4_64BM_128INDBM) || - (init->cce_mode == R128_PM4_64BM_64VCBM_64INDBM)); - dev_priv->cce_secure = init->cce_secure; - - if (dev_priv->cce_is_bm_mode && dev_priv->is_pci) { - drm_free(dev_priv, sizeof(*dev_priv), DRM_MEM_DRIVER); - dev->dev_private = NULL; - return -EINVAL; - } - - for (i = 0; i < dev->map_count; i++) { - if (dev->maplist[i]->type == _DRM_SHM) { - dev_priv->sarea = dev->maplist[i]; - break; - } - } - - DO_FIND_MAP(dev_priv->fb, init->fb_offset); - if (!dev_priv->is_pci) { - DO_FIND_MAP(dev_priv->agp_ring, init->agp_ring_offset); - DO_FIND_MAP(dev_priv->agp_read_ptr, init->agp_read_ptr_offset); - DO_FIND_MAP(dev_priv->agp_vertbufs, init->agp_vertbufs_offset); - DO_FIND_MAP(dev_priv->agp_indbufs, init->agp_indbufs_offset); - DO_FIND_MAP(dev_priv->agp_textures, init->agp_textures_offset); - } - DO_FIND_MAP(dev_priv->mmio, init->mmio_offset); - - dev_priv->sarea_priv = - (drm_r128_sarea_t *)((u8 *)dev_priv->sarea->handle + - init->sarea_priv_offset); - - if (!dev_priv->is_pci) { - DO_REMAP(dev_priv->agp_ring); - DO_REMAP(dev_priv->agp_read_ptr); - DO_REMAP(dev_priv->agp_vertbufs); -#if 0 - DO_REMAP(dev_priv->agp_indirectbufs); - DO_REMAP(dev_priv->agp_textures); -#endif - - dev_priv->ring_size = init->ring_size; - dev_priv->ring_sizel2qw = drm_order(init->ring_size/8); - dev_priv->ring_entries = init->ring_size/sizeof(u32); - dev_priv->ring_read_ptr = ((__volatile__ u32 *) - dev_priv->agp_read_ptr->handle); - dev_priv->ring_start = (u32 *)dev_priv->agp_ring->handle; - dev_priv->ring_end = ((u32 *)dev_priv->agp_ring->handle - + dev_priv->ring_entries); - } - - dev_priv->submit_age = 0; - R128_WRITE(R128_VB_AGE_REG, dev_priv->submit_age); - - return 0; -} - -int r128_init_cce(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - drm_r128_init_t init; - - if (copy_from_user(&init, (drm_r128_init_t *)arg, sizeof(init))) - return -EFAULT; - - switch (init.func) { - case R128_INIT_CCE: - return r128_do_init_cce(dev, &init); - case R128_CLEANUP_CCE: - return r128_do_cleanup_cce(dev); - } - - return -EINVAL; -} - -static void r128_mark_vertbufs_done(drm_device_t *dev) -{ - drm_device_dma_t *dma = dev->dma; - int i; - - for (i = 0; i < dma->buf_count; i++) { - drm_buf_t *buf = dma->buflist[i]; - drm_r128_buf_priv_t *buf_priv = buf->dev_private; - buf_priv->age = 0; - } -} - -static int r128_do_pixcache_flush(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - u32 tmp; - int i; - - tmp = R128_READ(R128_PC_NGUI_CTLSTAT) | R128_PC_FLUSH_ALL; - R128_WRITE(R128_PC_NGUI_CTLSTAT, tmp); - - for (i = 0; i < dev_priv->usec_timeout; i++) { - if (!(R128_READ(R128_PC_NGUI_CTLSTAT) & R128_PC_BUSY)) - return 0; - udelay(1); - } - - return -EBUSY; -} - -static int r128_do_wait_for_fifo(drm_device_t *dev, int entries) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int i; - - for (i = 0; i < dev_priv->usec_timeout; i++) { - int slots = R128_READ(R128_GUI_STAT) & R128_GUI_FIFOCNT_MASK; - if (slots >= entries) return 0; - udelay(1); - } - return -EBUSY; -} - -static int r128_do_wait_for_idle(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int i, ret; - - if (!(ret = r128_do_wait_for_fifo(dev, 64))) return ret; - - for (i = 0; i < dev_priv->usec_timeout; i++) { - if (!(R128_READ(R128_GUI_STAT) & R128_GUI_ACTIVE)) { - (void)r128_do_pixcache_flush(dev); - return 0; - } - udelay(1); - } - return -EBUSY; -} - -int r128_do_engine_reset(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - u32 clock_cntl_index, mclk_cntl, gen_reset_cntl; - - (void)r128_do_pixcache_flush(dev); - - clock_cntl_index = R128_READ(R128_CLOCK_CNTL_INDEX); - mclk_cntl = R128_READ_PLL(dev, R128_MCLK_CNTL); - - R128_WRITE_PLL(R128_MCLK_CNTL, - mclk_cntl | R128_FORCE_GCP | R128_FORCE_PIPE3D_CP); - - gen_reset_cntl = R128_READ(R128_GEN_RESET_CNTL); - - R128_WRITE(R128_GEN_RESET_CNTL, gen_reset_cntl | R128_SOFT_RESET_GUI); - (void)R128_READ(R128_GEN_RESET_CNTL); - R128_WRITE(R128_GEN_RESET_CNTL, gen_reset_cntl & ~R128_SOFT_RESET_GUI); - (void)R128_READ(R128_GEN_RESET_CNTL); - - R128_WRITE_PLL(R128_MCLK_CNTL, mclk_cntl); - R128_WRITE(R128_CLOCK_CNTL_INDEX, clock_cntl_index); - R128_WRITE(R128_GEN_RESET_CNTL, gen_reset_cntl); - - /* For CCE ring buffer only */ - if (dev_priv->cce_is_bm_mode) { - R128_WRITE(R128_PM4_BUFFER_DL_WPTR, 0); - R128_WRITE(R128_PM4_BUFFER_DL_RPTR, 0); - *dev_priv->ring_read_ptr = 0; - dev_priv->sarea_priv->ring_write = 0; - } - - /* Reset the CCE mode */ - (void)r128_do_wait_for_idle(dev); - R128_WRITE(R128_PM4_BUFFER_CNTL, - dev_priv->cce_mode | dev_priv->ring_sizel2qw); - (void)R128_READ(R128_PM4_BUFFER_ADDR); /* as per the sample code */ - R128_WRITE(R128_PM4_MICRO_CNTL, R128_PM4_MICRO_FREERUN); - - r128_mark_vertbufs_done(dev); - return 0; -} - -int r128_eng_reset(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - - if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) || - dev->lock.pid != current->pid) { - DRM_ERROR("r128_eng_reset called without holding the lock\n"); - return -EINVAL; - } - - return r128_do_engine_reset(dev); -} - -static int r128_do_engine_flush(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - u32 tmp; - - tmp = R128_READ(R128_PM4_BUFFER_DL_WPTR); - R128_WRITE(R128_PM4_BUFFER_DL_WPTR, tmp | R128_PM4_BUFFER_DL_DONE); - - return 0; -} - -int r128_eng_flush(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - - if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) || - dev->lock.pid != current->pid) { - DRM_ERROR("r128_eng_flush called without holding the lock\n"); - return -EINVAL; - } - - return r128_do_engine_flush(dev); -} - -static int r128_do_cce_wait_for_fifo(drm_device_t *dev, int entries) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int i; - - for (i = 0; i < dev_priv->usec_timeout; i++) { - int slots = R128_READ(R128_PM4_STAT) & R128_PM4_FIFOCNT_MASK; - if (slots >= entries) return 0; - udelay(1); - } - return -EBUSY; -} - -int r128_do_cce_wait_for_idle(drm_device_t *dev) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int i; - - if (dev_priv->cce_is_bm_mode) { - for (i = 0; i < dev_priv->usec_timeout; i++) { - if (*dev_priv->ring_read_ptr == dev_priv->sarea_priv->ring_write) { - int pm4stat = R128_READ(R128_PM4_STAT); - if ((pm4stat & R128_PM4_FIFOCNT_MASK) >= dev_priv->cce_fifo_size && - !(pm4stat & (R128_PM4_BUSY | R128_PM4_GUI_ACTIVE))) { - return r128_do_pixcache_flush(dev); - } - } - udelay(1); - } - return -EBUSY; - } else { - int ret = r128_do_cce_wait_for_fifo(dev, dev_priv->cce_fifo_size); - if (ret < 0) return ret; - - for (i = 0; i < dev_priv->usec_timeout; i++) { - int pm4stat = R128_READ(R128_PM4_STAT); - if (!(pm4stat & (R128_PM4_BUSY | R128_PM4_GUI_ACTIVE))) { - return r128_do_pixcache_flush(dev); - } - udelay(1); - } - return -EBUSY; - } -} - -int r128_cce_idle(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - - if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) || - dev->lock.pid != current->pid) { - DRM_ERROR("r128_wait_idle called without holding the lock\n"); - return -EINVAL; - } - - return r128_do_cce_wait_for_idle(dev); -} - -static int r128_submit_packets_ring_secure(drm_device_t *dev, - u32 *commands, int *count) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int write = dev_priv->sarea_priv->ring_write; - int *write_ptr = dev_priv->ring_start + write; - int c = *count; - u32 tmp = 0; - int psize = 0; - int writing = 1; - int timeout; - - while (c > 0) { - tmp = *commands++; - if (!psize) { - writing = 1; - - if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET0) { - if ((tmp & R128_CCE_PACKET0_REG_MASK) <= (0x1004 >> 2)) { - if ((tmp & R128_CCE_PACKET0_REG_MASK) != - (R128_PM4_VC_FPU_SETUP >> 2)) { - writing = 0; - } - } - psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; - } else if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET1) { - if ((tmp & R128_CCE_PACKET1_REG0_MASK) <= (0x1004 >> 2)) { - if ((tmp & R128_CCE_PACKET1_REG0_MASK) != - (R128_PM4_VC_FPU_SETUP >> 2)) { - writing = 0; - } - } else if ((tmp & R128_CCE_PACKET1_REG1_MASK) <= - (0x1004 << 9)) { - if ((tmp & R128_CCE_PACKET1_REG1_MASK) != - (R128_PM4_VC_FPU_SETUP << 9)) { - writing = 0; - } - } - psize = 3; - } else { - psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; - } - } - psize--; - - if (writing) { - write++; - *write_ptr++ = tmp; - } - if (write >= dev_priv->ring_entries) { - write = 0; - write_ptr = dev_priv->ring_start; - } - timeout = 0; - while (write == *dev_priv->ring_read_ptr) { - (void)R128_READ(R128_PM4_BUFFER_DL_RPTR); - if (timeout++ >= dev_priv->usec_timeout) - return -EBUSY; - udelay(1); - } - c--; - } - - if (write < 32) - memcpy(dev_priv->ring_end, - dev_priv->ring_start, - write * sizeof(u32)); - - /* Make sure WC cache has been flushed */ - r128_flush_write_combine(); - - dev_priv->sarea_priv->ring_write = write; - R128_WRITE(R128_PM4_BUFFER_DL_WPTR, write); - - *count = 0; - - return 0; -} - -static int r128_submit_packets_pio_secure(drm_device_t *dev, - u32 *commands, int *count) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - u32 tmp = 0; - int psize = 0; - int writing = 1; - int addr = R128_PM4_FIFO_DATA_EVEN; - int ret; - - while (*count > 0) { - tmp = *commands++; - if (!psize) { - writing = 1; - - if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET0) { - if ((tmp & R128_CCE_PACKET0_REG_MASK) <= (0x1004 >> 2)) { - if ((tmp & R128_CCE_PACKET0_REG_MASK) != - (R128_PM4_VC_FPU_SETUP >> 2)) { - writing = 0; - } - } - psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; - } else if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET1) { - if ((tmp & R128_CCE_PACKET1_REG0_MASK) <= (0x1004 >> 2)) { - if ((tmp & R128_CCE_PACKET1_REG0_MASK) != - (R128_PM4_VC_FPU_SETUP >> 2)) { - writing = 0; - } - } else if ((tmp & R128_CCE_PACKET1_REG1_MASK) <= - (0x1004 << 9)) { - if ((tmp & R128_CCE_PACKET1_REG1_MASK) != - (R128_PM4_VC_FPU_SETUP << 9)) { - writing = 0; - } - } - psize = 3; - } else { - psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2; - } - } - psize--; - - if (writing) { - if ((ret = r128_do_cce_wait_for_fifo(dev, 1)) < 0) - return ret; - R128_WRITE(addr, tmp); - addr ^= 0x0004; - } - - *count -= 1; - } - - if (addr == R128_PM4_FIFO_DATA_ODD) { - if ((ret = r128_do_cce_wait_for_fifo(dev, 1)) < 0) return ret; - R128_WRITE(addr, R128_CCE_PACKET2); - } - - return 0; -} - -static int r128_submit_packets_ring(drm_device_t *dev, - u32 *commands, int *count) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int write = dev_priv->sarea_priv->ring_write; - int *write_ptr = dev_priv->ring_start + write; - int c = *count; - int timeout; - - while (c > 0) { - write++; - *write_ptr++ = *commands++; - if (write >= dev_priv->ring_entries) { - write = 0; - write_ptr = dev_priv->ring_start; - } - timeout = 0; - while (write == *dev_priv->ring_read_ptr) { - (void)R128_READ(R128_PM4_BUFFER_DL_RPTR); - if (timeout++ >= dev_priv->usec_timeout) - return -EBUSY; - udelay(1); - } - c--; - } - - if (write < 32) - memcpy(dev_priv->ring_end, - dev_priv->ring_start, - write * sizeof(u32)); - - /* Make sure WC cache has been flushed */ - r128_flush_write_combine(); - - dev_priv->sarea_priv->ring_write = write; - R128_WRITE(R128_PM4_BUFFER_DL_WPTR, write); - - *count = 0; - - return 0; -} - -static int r128_submit_packets_pio(drm_device_t *dev, - u32 *commands, int *count) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int ret; - - while (*count > 1) { - if ((ret = r128_do_cce_wait_for_fifo(dev, 2)) < 0) return ret; - R128_WRITE(R128_PM4_FIFO_DATA_EVEN, *commands++); - R128_WRITE(R128_PM4_FIFO_DATA_ODD, *commands++); - *count -= 2; - } - - if (*count) { - if ((ret = r128_do_cce_wait_for_fifo(dev, 2)) < 0) return ret; - R128_WRITE(R128_PM4_FIFO_DATA_EVEN, *commands++); - R128_WRITE(R128_PM4_FIFO_DATA_ODD, R128_CCE_PACKET2); - *count = 0; - } - - return 0; -} - -static int r128_do_submit_packets(drm_device_t *dev, u32 *buffer, int count) -{ - drm_r128_private_t *dev_priv = dev->dev_private; - int c = count; - int ret; - - if (dev_priv->cce_is_bm_mode) { - int left = 0; - - if (c >= dev_priv->ring_entries) { - c = dev_priv->ring_entries-1; - left = count - c; - } - - /* Since this is only used by the kernel we can use the - insecure ring buffer submit packet routine */ - ret = r128_submit_packets_ring(dev, buffer, &c); - - c += left; - } else { - /* Since this is only used by the kernel we can use the - insecure PIO submit packet routine */ - ret = r128_submit_packets_pio(dev, buffer, &c); - } - - if (ret < 0) return ret; - else return c; -} - -int r128_submit_pkt(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - drm_r128_private_t *dev_priv = dev->dev_private; - drm_r128_packet_t packet; - u32 *buffer; - int c; - int size; - int ret = 0; - - if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) || - dev->lock.pid != current->pid) { - DRM_ERROR("r128_submit_pkt called without holding the lock\n"); - return -EINVAL; - } - - if (copy_from_user(&packet, (drm_r128_packet_t *)arg, sizeof(packet))) - return -EFAULT; - - c = packet.count; - size = c * sizeof(*buffer); - - if (dev_priv->cce_is_bm_mode) { - int left = 0; - - if (c >= dev_priv->ring_entries) { - c = dev_priv->ring_entries-1; - size = c * sizeof(*buffer); - left = packet.count - c; - } - - if ((buffer = kmalloc(size, 0)) == NULL) return -ENOMEM; - if (copy_from_user(buffer, packet.buffer, size)) - return -EFAULT; - - if (dev_priv->cce_secure) - ret = r128_submit_packets_ring_secure(dev, buffer, &c); - else - ret = r128_submit_packets_ring(dev, buffer, &c); - - c += left; - } else { - if ((buffer = kmalloc(size, 0)) == NULL) return -ENOMEM; - if (copy_from_user(buffer, packet.buffer, size)) - return -EFAULT; - - if (dev_priv->cce_secure) - ret = r128_submit_packets_pio_secure(dev, buffer, &c); - else - ret = r128_submit_packets_pio(dev, buffer, &c); - } - - kfree(buffer); - - packet.count = c; - if (copy_to_user((drm_r128_packet_t *)arg, &packet, sizeof(packet))) - return -EFAULT; - - if (ret) return ret; - else if (c > 0) return -EAGAIN; - - return 0; -} - -static int r128_send_vertbufs(drm_device_t *dev, drm_r128_vertex_t *v) -{ - drm_device_dma_t *dma = dev->dma; - drm_r128_private_t *dev_priv = dev->dev_private; - drm_r128_buf_priv_t *buf_priv; - drm_buf_t *buf; - int i, ret; - u32 cce[2]; - - /* Make sure we have valid data */ - for (i = 0; i < v->send_count; i++) { - int idx = v->send_indices[i]; - - if (idx < 0 || idx >= dma->buf_count) { - DRM_ERROR("Index %d (of %d max)\n", - idx, dma->buf_count - 1); - return -EINVAL; - } - buf = dma->buflist[idx]; - if (buf->pid != current->pid) { - DRM_ERROR("Process %d using buffer owned by %d\n", - current->pid, buf->pid); - return -EINVAL; - } - if (buf->pending) { - DRM_ERROR("Sending pending buffer:" - " buffer %d, offset %d\n", - v->send_indices[i], i); - return -EINVAL; - } - } - - /* Wait for idle, if we've wrapped to make sure that all pending - buffers have been processed */ - if (dev_priv->submit_age == R128_MAX_VBUF_AGE) { - if ((ret = r128_do_cce_wait_for_idle(dev)) < 0) return ret; - dev_priv->submit_age = 0; - r128_mark_vertbufs_done(dev); - } - - /* Make sure WC cache has been flushed (if in PIO mode) */ - if (!dev_priv->cce_is_bm_mode) r128_flush_write_combine(); - - /* FIXME: Add support for sending vertex buffer to the CCE here - instead of in client code. The v->prim holds the primitive - type that should be drawn. Loop over the list buffers in - send_indices[] and submit a packet for each VB. - - This will require us to loop over the clip rects here as - well, which implies that we extend the kernel driver to allow - cliprects to be stored here. Note that the cliprects could - possibly come from the X server instead of the client, but - this will require additional changes to the DRI to allow for - this optimization. */ - - /* Submit a CCE packet that writes submit_age to R128_VB_AGE_REG */ - cce[0] = R128CCE0(R128_CCE_PACKET0, R128_VB_AGE_REG, 0); - cce[1] = dev_priv->submit_age; - if ((ret = r128_do_submit_packets(dev, cce, 2)) < 0) { - /* Until we add support for sending VBs to the CCE in - this routine, we can recover from this error. After - we add that support, we won't be able to easily - recover, so we will probably have to implement - another mechanism for handling timeouts from packets - submitted directly by the kernel. */ - return ret; - } - - /* Now that the submit packet request has succeeded, we can mark - the buffers as pending */ - for (i = 0; i < v->send_count; i++) { - buf = dma->buflist[v->send_indices[i]]; - buf->pending = 1; - - buf_priv = buf->dev_private; - buf_priv->age = dev_priv->submit_age; - } - - dev_priv->submit_age++; - - return 0; -} - -static drm_buf_t *r128_freelist_get(drm_device_t *dev) -{ - drm_device_dma_t *dma = dev->dma; - drm_r128_private_t *dev_priv = dev->dev_private; - drm_r128_buf_priv_t *buf_priv; - drm_buf_t *buf; - int i, t; - - /* FIXME: Optimize -- use freelist code */ - - for (i = 0; i < dma->buf_count; i++) { - buf = dma->buflist[i]; - buf_priv = buf->dev_private; - if (buf->pid == 0) return buf; - } - - for (t = 0; t < dev_priv->usec_timeout; t++) { - u32 done_age = R128_READ(R128_VB_AGE_REG); - - for (i = 0; i < dma->buf_count; i++) { - buf = dma->buflist[i]; - buf_priv = buf->dev_private; - if (buf->pending && buf_priv->age <= done_age) { - /* The buffer has been processed, so it - can now be used */ - buf->pending = 0; - return buf; - } - } - udelay(1); - } - - r128_status(dev); - return NULL; -} - - -static int r128_get_vertbufs(drm_device_t *dev, drm_r128_vertex_t *v) -{ - drm_buf_t *buf; - int i; - - for (i = v->granted_count; i < v->request_count; i++) { - buf = r128_freelist_get(dev); - if (!buf) break; - buf->pid = current->pid; - if (copy_to_user(&v->request_indices[i], - &buf->idx, - sizeof(buf->idx)) || - copy_to_user(&v->request_sizes[i], - &buf->total, - sizeof(buf->total))) - return -EFAULT; - ++v->granted_count; - } - return 0; -} - -int r128_vertex_buf(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) -{ - drm_file_t *priv = filp->private_data; - drm_device_t *dev = priv->dev; - drm_r128_private_t *dev_priv = dev->dev_private; - drm_device_dma_t *dma = dev->dma; - int retcode = 0; - drm_r128_vertex_t v; - - if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) || - dev->lock.pid != current->pid) { - DRM_ERROR("r128_vertex_buf called without holding the lock\n"); - return -EINVAL; - } - - if (!dev_priv || dev_priv->is_pci) { - DRM_ERROR("r128_vertex_buf called with a PCI card\n"); - return -EINVAL; - } - - if (copy_from_user(&v, (drm_r128_vertex_t *)arg, sizeof(v))) - return -EFAULT; - DRM_DEBUG("%d: %d send, %d req\n", - current->pid, v.send_count, v.request_count); - - if (v.send_count < 0 || v.send_count > dma->buf_count) { - DRM_ERROR("Process %d trying to send %d buffers (of %d max)\n", - current->pid, v.send_count, dma->buf_count); - return -EINVAL; - } - if (v.request_count < 0 || v.request_count > dma->buf_count) { - DRM_ERROR("Process %d trying to get %d buffers (of %d max)\n", - current->pid, v.request_count, dma->buf_count); - return -EINVAL; - } - - if (v.send_count) { - retcode = r128_send_vertbufs(dev, &v); - } - - v.granted_count = 0; - - if (!retcode && v.request_count) { - retcode = r128_get_vertbufs(dev, &v); - } - - DRM_DEBUG("%d returning, granted = %d\n", - current->pid, v.granted_count); - if (copy_to_user((drm_r128_vertex_t *)arg, &v, sizeof(v))) - return -EFAULT; - - return retcode; -} diff --git a/linux/r128_drm.h b/linux/r128_drm.h index 8a4842cf..b81d2fca 100644 --- a/linux/r128_drm.h +++ b/linux/r128_drm.h @@ -11,11 +11,11 @@ * 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 (including the next * paragraph) 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 @@ -24,88 +24,251 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * Authors: Kevin E. Martin + * Authors: + * Kevin E. Martin + * Gareth Hughes * */ #ifndef _R128_DRM_H_ #define _R128_DRM_H_ +/* WARNING: If you change any of these defines, make sure to change the + * defines in the X server file (r128_sarea.h) + */ +#ifndef __R128_SAREA_DEFINES__ +#define __R128_SAREA_DEFINES__ + +/* What needs to be changed for the current vertex buffer? + */ +#define R128_UPLOAD_CONTEXT 0x001 +#define R128_UPLOAD_SETUP 0x002 +#define R128_UPLOAD_TEX0 0x004 +#define R128_UPLOAD_TEX1 0x008 +#define R128_UPLOAD_TEX0IMAGES 0x010 +#define R128_UPLOAD_TEX1IMAGES 0x020 +#define R128_UPLOAD_CORE 0x040 +#define R128_UPLOAD_MASKS 0x080 +#define R128_UPLOAD_WINDOW 0x100 +#define R128_UPLOAD_CLIPRECTS 0x200 /* handled client-side */ +#define R128_REQUIRE_QUIESCENCE 0x400 +#define R128_UPLOAD_ALL 0x7ff + +#define R128_FRONT 0x1 +#define R128_BACK 0x2 +#define R128_DEPTH 0x4 + +/* Primitive types + */ +#define R128_POINTS 0x1 +#define R128_LINES 0x2 +#define R128_LINE_STRIP 0x3 +#define R128_TRIANGLES 0x4 +#define R128_TRIANGLE_FAN 0x5 +#define R128_TRIANGLE_STRIP 0x6 + +/* Vertex/indirect buffer size + */ +#if 1 +#define R128_BUFFER_SIZE 16384 +#else +#define R128_BUFFER_SIZE (128 * 1024) +#endif + +/* Byte offsets for indirect buffer data + */ +#define R128_INDEX_PRIM_OFFSET 20 +#define R128_HOSTDATA_BLIT_OFFSET 32 + +/* 2048x2048 @ 32bpp texture requires this many indirect buffers + */ +#define R128_MAX_BLIT_BUFFERS ((2048 * 2048 * 4) / R128_BUFFER_SIZE) + +/* Keep these small for testing. + */ +#define R128_NR_SAREA_CLIPRECTS 12 + +/* There are 2 heaps (local/AGP). Each region within a heap is a + * minimum of 64k, and there are at most 64 of them per heap. + */ +#define R128_LOCAL_TEX_HEAP 0 +#define R128_AGP_TEX_HEAP 1 +#define R128_NR_TEX_HEAPS 2 +#define R128_NR_TEX_REGIONS 64 +#define R128_LOG_TEX_GRANULARITY 16 + +#define R128_NR_CONTEXT_REGS 12 +#define R128_TEX_MAXLEVELS 11 + +#endif /* __R128_SAREA_DEFINES__ */ + +typedef struct { + /* Context state - can be written in one large chunk */ + unsigned int dst_pitch_offset_c; + unsigned int dp_gui_master_cntl_c; + unsigned int sc_top_left_c; + unsigned int sc_bottom_right_c; + unsigned int z_offset_c; + unsigned int z_pitch_c; + unsigned int z_sten_cntl_c; + unsigned int tex_cntl_c; + unsigned int misc_3d_state_cntl_reg; + unsigned int texture_clr_cmp_clr_c; + unsigned int texture_clr_cmp_msk_c; + unsigned int fog_color_c; + + /* Texture state */ + unsigned int tex_size_pitch_c; + unsigned int constant_color_c; + + /* Setup state */ + unsigned int pm4_vc_fpu_setup; + unsigned int setup_cntl; + + /* Mask state */ + unsigned int dp_write_mask; + unsigned int sten_ref_mask_c; + unsigned int plane_3d_mask_c; + + /* Window state */ + unsigned int window_xy_offset; + + /* Core state */ + unsigned int scale_3d_cntl; +} drm_r128_context_regs_t; + +/* Setup registers for each texture unit */ +typedef struct { + unsigned int tex_cntl; + unsigned int tex_combine_cntl; + unsigned int tex_size_pitch; + unsigned int tex_offset[R128_TEX_MAXLEVELS]; + unsigned int tex_border_color; +} drm_r128_texture_regs_t; + + +typedef struct drm_tex_region { + unsigned char next, prev; + unsigned char in_use; + int age; +} drm_tex_region_t; + +typedef struct drm_r128_sarea { + /* The channel for communication of state information to the kernel + * on firing a vertex buffer. + */ + drm_r128_context_regs_t context_state; + drm_r128_texture_regs_t tex_state[R128_NR_TEX_HEAPS]; + unsigned int dirty; + unsigned int vertsize; + unsigned int vc_format; + + /* The current cliprects, or a subset thereof. + */ + drm_clip_rect_t boxes[R128_NR_SAREA_CLIPRECTS]; + unsigned int nbox; + + /* Counters for client-side throttling of rendering clients. + */ + unsigned int last_frame; + unsigned int last_dispatch; + + drm_tex_region_t tex_list[R128_NR_TEX_HEAPS][R128_NR_TEX_REGIONS+1]; + int tex_age[R128_NR_TEX_HEAPS]; + int ctx_owner; +} drm_r128_sarea_t; + + /* WARNING: If you change any of these defines, make sure to change the * defines in the Xserver file (xf86drmR128.h) */ typedef struct drm_r128_init { - enum { + enum { R128_INIT_CCE = 0x01, R128_CLEANUP_CCE = 0x02 } func; int sarea_priv_offset; int is_pci; int cce_mode; - int cce_fifo_size; int cce_secure; int ring_size; int usec_timeout; - int fb_offset; - int agp_ring_offset; - int agp_read_ptr_offset; - int agp_vertbufs_offset; - int agp_indbufs_offset; - int agp_textures_offset; - int mmio_offset; + unsigned int fb_bpp; + unsigned int front_offset, front_pitch; + unsigned int back_offset, back_pitch; + unsigned int depth_bpp; + unsigned int depth_offset, depth_pitch; + unsigned int span_offset; + + unsigned int fb_offset; + unsigned int mmio_offset; + unsigned int ring_offset; + unsigned int ring_rptr_offset; + unsigned int buffers_offset; + unsigned int agp_textures_offset; } drm_r128_init_t; -typedef struct drm_r128_packet { - unsigned int *buffer; - int count; - int flags; -} drm_r128_packet_t; +typedef struct drm_r128_cce_stop { + int flush; + int idle; +} drm_r128_cce_stop_t; -typedef enum drm_r128_prim { - _DRM_R128_PRIM_NONE = 0x0001, - _DRM_R128_PRIM_POINT = 0x0002, - _DRM_R128_PRIM_LINE = 0x0004, - _DRM_R128_PRIM_POLY_LINE = 0x0008, - _DRM_R128_PRIM_TRI_LIST = 0x0010, - _DRM_R128_PRIM_TRI_FAN = 0x0020, - _DRM_R128_PRIM_TRI_STRIP = 0x0040, - _DRM_R128_PRIM_TRI_TYPE2 = 0x0080 -} drm_r128_prim_t; +typedef struct drm_r128_clear { + unsigned int flags; + int x, y, w, h; + unsigned int clear_color; + unsigned int clear_depth; + unsigned int color_mask; + unsigned int depth_mask; +} drm_r128_clear_t; typedef struct drm_r128_vertex { - /* Indices here refer to the offset into - buflist in drm_buf_get_t. */ - int send_count; /* Number of buffers to send */ - int *send_indices; /* List of handles to buffers */ - int *send_sizes; /* Lengths of data to send */ - drm_r128_prim_t prim; /* Primitive type */ - int request_count; /* Number of buffers requested */ - int *request_indices; /* Buffer information */ - int *request_sizes; - int granted_count; /* Number of buffers granted */ + int prim; + int idx; /* Index of vertex buffer */ + int count; /* Number of vertices in buffer */ + int discard; /* Client finished with buffer? */ } drm_r128_vertex_t; -/* WARNING: If you change any of these defines, make sure to change the - * defines in the Xserver file (r128_sarea.h) - */ -#define R128_LOCAL_TEX_HEAP 0 -#define R128_AGP_TEX_HEAP 1 -#define R128_NR_TEX_HEAPS 2 -#define R128_NR_TEX_REGIONS 64 -#define R128_LOG_TEX_GRANULARITY 16 +typedef struct drm_r128_indices { + int prim; + int idx; + int start; + int end; + int discard; /* Client finished with buffer? */ +} drm_r128_indices_t; -typedef struct drm_tex_region { - unsigned char next, prev; - unsigned char in_use; - int age; -} drm_tex_region_t; +typedef struct drm_r128_blit { + int idx; + int pitch; + int offset; + int format; + unsigned short x, y; + unsigned short width, height; +} drm_r128_blit_t; -typedef struct drm_r128_sarea { - drm_tex_region_t tex_list[R128_NR_TEX_HEAPS][R128_NR_TEX_REGIONS+1]; - int tex_age[R128_NR_TEX_HEAPS]; - int ctx_owner; - int ring_write; -} drm_r128_sarea_t; +typedef struct drm_r128_depth { + enum { + R128_WRITE_SPAN = 0x01, + R128_WRITE_PIXELS = 0x02, + R128_READ_SPAN = 0x03, + R128_READ_PIXELS = 0x04 + } func; + int n; + int *x; + int *y; + unsigned int *buffer; + unsigned char *mask; +} drm_r128_depth_t; + +typedef struct drm_r128_stipple { + unsigned int *mask; +} drm_r128_stipple_t; + +typedef struct drm_r128_packet { + unsigned int *buffer; + int count; + int flags; +} drm_r128_packet_t; #endif diff --git a/linux/r128_drv.c b/linux/r128_drv.c index 969ada93..89b68696 100644 --- a/linux/r128_drv.c +++ b/linux/r128_drv.c @@ -24,8 +24,10 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * Authors: Rickard E. (Rik) Faith - * Kevin E. Martin + * Authors: + * Rickard E. (Rik) Faith + * Kevin E. Martin + * Gareth Hughes * */ @@ -33,15 +35,15 @@ #include "drmP.h" #include "r128_drv.h" -#define R128_NAME "r128" -#define R128_DESC "ATI Rage 128" -#define R128_DATE "20000928" -#define R128_MAJOR 1 -#define R128_MINOR 0 -#define R128_PATCHLEVEL 0 +#define R128_NAME "r128" +#define R128_DESC "ATI Rage 128" +#define R128_DATE "20001201" +#define R128_MAJOR 2 +#define R128_MINOR 1 +#define R128_PATCHLEVEL 0 -static drm_device_t r128_device; -drm_ctx_t r128_res_ctx; +static drm_device_t r128_device; +drm_ctx_t r128_res_ctx; static struct file_operations r128_fops = { #if LINUX_VERSION_CODE >= 0x020400 @@ -65,52 +67,61 @@ static struct miscdevice r128_misc = { }; static drm_ioctl_desc_t r128_ioctls[] = { - [DRM_IOCTL_NR(DRM_IOCTL_VERSION)] = { r128_version, 0, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_GET_UNIQUE)] = { drm_getunique, 0, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_GET_MAGIC)] = { drm_getmagic, 0, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_IRQ_BUSID)] = { drm_irq_busid, 0, 1 }, - - [DRM_IOCTL_NR(DRM_IOCTL_SET_UNIQUE)] = { drm_setunique, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_BLOCK)] = { drm_block, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_UNBLOCK)] = { drm_unblock, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AUTH_MAGIC)] = { drm_authmagic, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_ADD_MAP)] = { drm_addmap, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_ADD_BUFS)] = { r128_addbufs, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_MARK_BUFS)] = { drm_markbufs, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_INFO_BUFS)] = { drm_infobufs, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_MAP_BUFS)] = { r128_mapbufs, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_FREE_BUFS)] = { drm_freebufs, 1, 0 }, - - [DRM_IOCTL_NR(DRM_IOCTL_ADD_CTX)] = { r128_addctx, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_RM_CTX)] = { r128_rmctx, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_MOD_CTX)] = { r128_modctx, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_GET_CTX)] = { r128_getctx, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_SWITCH_CTX)] = { r128_switchctx, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_NEW_CTX)] = { r128_newctx, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_RES_CTX)] = { r128_resctx, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_ADD_DRAW)] = { drm_adddraw, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_RM_DRAW)] = { drm_rmdraw, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_LOCK)] = { r128_lock, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_UNLOCK)] = { r128_unlock, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_FINISH)] = { drm_finish, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_VERSION)] = { r128_version, 0, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_GET_UNIQUE)] = { drm_getunique, 0, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_GET_MAGIC)] = { drm_getmagic, 0, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_IRQ_BUSID)] = { drm_irq_busid, 0, 1 }, + + [DRM_IOCTL_NR(DRM_IOCTL_SET_UNIQUE)] = { drm_setunique, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_BLOCK)] = { drm_block, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_UNBLOCK)] = { drm_unblock, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AUTH_MAGIC)] = { drm_authmagic, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_ADD_MAP)] = { drm_addmap, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_ADD_BUFS)] = { r128_addbufs, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_MARK_BUFS)] = { drm_markbufs, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_INFO_BUFS)] = { drm_infobufs, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_MAP_BUFS)] = { r128_mapbufs, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_FREE_BUFS)] = { drm_freebufs, 1, 0 }, + + [DRM_IOCTL_NR(DRM_IOCTL_ADD_CTX)] = { r128_addctx, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_RM_CTX)] = { r128_rmctx, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_MOD_CTX)] = { r128_modctx, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_GET_CTX)] = { r128_getctx, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_SWITCH_CTX)] = { r128_switchctx, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_NEW_CTX)] = { r128_newctx, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_RES_CTX)] = { r128_resctx, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_ADD_DRAW)] = { drm_adddraw, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_RM_DRAW)] = { drm_rmdraw, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_DMA)] = { r128_cce_buffers, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_LOCK)] = { r128_lock, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_UNLOCK)] = { r128_unlock, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_FINISH)] = { drm_finish, 1, 0 }, #if defined(CONFIG_AGP) || defined(CONFIG_AGP_MODULE) - [DRM_IOCTL_NR(DRM_IOCTL_AGP_ACQUIRE)] = { drm_agp_acquire, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_RELEASE)] = { drm_agp_release, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_ENABLE)] = { drm_agp_enable, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_INFO)] = { drm_agp_info, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_ALLOC)] = { drm_agp_alloc, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_FREE)] = { drm_agp_free, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_BIND)] = { drm_agp_bind, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_AGP_UNBIND)] = { drm_agp_unbind, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_ACQUIRE)] = { drm_agp_acquire, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_RELEASE)] = { drm_agp_release, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_ENABLE)] = { drm_agp_enable, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_INFO)] = { drm_agp_info, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_ALLOC)] = { drm_agp_alloc, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_FREE)] = { drm_agp_free, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_BIND)] = { drm_agp_bind, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_AGP_UNBIND)] = { drm_agp_unbind, 1, 1 }, #endif - [DRM_IOCTL_NR(DRM_IOCTL_R128_INIT)] = { r128_init_cce, 1, 1 }, - [DRM_IOCTL_NR(DRM_IOCTL_R128_RESET)] = { r128_eng_reset, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_R128_FLUSH)] = { r128_eng_flush, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_R128_PACKET)] = { r128_submit_pkt, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_R128_IDLE)] = { r128_cce_idle, 1, 0 }, - [DRM_IOCTL_NR(DRM_IOCTL_R128_VERTEX)] = { r128_vertex_buf, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_INIT)] = { r128_cce_init, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_CCE_START)] = { r128_cce_start, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_CCE_STOP)] = { r128_cce_stop, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_CCE_RESET)] = { r128_cce_reset, 1, 1 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_CCE_IDLE)] = { r128_cce_idle, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_RESET)] = { r128_engine_reset, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_SWAP)] = { r128_cce_swap, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_CLEAR)] = { r128_cce_clear, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_VERTEX)] = { r128_cce_vertex, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_INDICES)] = { r128_cce_indices, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_BLIT)] = { r128_cce_blit, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_DEPTH)] = { r128_cce_depth, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_STIPPLE)] = { r128_cce_stipple, 1, 0 }, + [DRM_IOCTL_NR(DRM_IOCTL_R128_PACKET)] = { r128_cce_packet, 1, 0 }, }; #define R128_IOCTL_COUNT DRM_ARRAY_SIZE(r128_ioctls) @@ -349,12 +360,12 @@ static int __init r128_init(void) #if defined(CONFIG_AGP) || defined(CONFIG_AGP_MODULE) dev->agp = drm_agp_init(); - if (dev->agp == NULL) { - DRM_ERROR("Cannot initialize agpgart module.\n"); - drm_proc_cleanup(); - misc_deregister(&r128_misc); - r128_takedown(dev); - return -ENOMEM; + if (dev->agp == NULL) { + DRM_ERROR("Cannot initialize agpgart module.\n"); + drm_proc_cleanup(); + misc_deregister(&r128_misc); + r128_takedown(dev); + return -ENOMEM; } #ifdef CONFIG_MTRR @@ -413,8 +424,8 @@ module_init(r128_init); module_exit(r128_cleanup); -int r128_version(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) +int r128_version(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) { drm_version_t version; int len; @@ -424,13 +435,13 @@ int r128_version(struct inode *inode, struct file *filp, unsigned int cmd, sizeof(version))) return -EFAULT; -#define DRM_COPY(name,value) \ - len = strlen(value); \ - if (len > name##_len) len = name##_len; \ - name##_len = strlen(value); \ - if (len && name) { \ - if (copy_to_user(name, value, len)) \ - return -EFAULT; \ +#define DRM_COPY(name,value) \ + len = strlen(value); \ + if (len > name##_len) len = name##_len; \ + name##_len = strlen(value); \ + if (len && name) { \ + if (copy_to_user(name, value, len)) \ + return -EFAULT; \ } version.version_major = R128_MAJOR; @@ -506,9 +517,8 @@ int r128_release(struct inode *inode, struct file *filp) } /* r128_ioctl is called whenever a process performs an ioctl on /dev/drm. */ - -int r128_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) +int r128_ioctl(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) { int nr = DRM_IOCTL_NR(cmd); drm_file_t *priv = filp->private_data; @@ -534,19 +544,25 @@ int r128_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, DRM_DEBUG("no function\n"); retcode = -EINVAL; } else if ((ioctl->root_only && !capable(CAP_SYS_ADMIN)) - || (ioctl->auth_needed && !priv->authenticated)) { + || (ioctl->auth_needed && !priv->authenticated)) { retcode = -EACCES; } else { retcode = (func)(inode, filp, cmd, arg); } } +#if 0 + if ( retcode ) { + DRM_INFO( "%s 0x%x ret = %d\n", __FUNCTION__, nr, retcode ); + } +#endif + atomic_dec(&dev->ioctl_count); return retcode; } -int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) +int r128_lock(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) { drm_file_t *priv = filp->private_data; drm_device_t *dev = priv->dev; @@ -572,33 +588,10 @@ int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, lock.context, current->pid, dev->lock.hw_lock->lock, lock.flags); -#if 0 - /* dev->queue_count == 0 right now for - r128. FIXME? */ - if (lock.context < 0 || lock.context >= dev->queue_count) + if (lock.context < 0) return -EINVAL; -#endif if (!ret) { -#if 0 - if (_DRM_LOCKING_CONTEXT(dev->lock.hw_lock->lock) - != lock.context) { - long j = jiffies - dev->lock.lock_time; - - if (lock.context == r128_res_ctx.handle && - j >= 0 && j < DRM_LOCK_SLICE) { - /* Can't take lock if we just had it and - there is contention. */ - DRM_DEBUG("%d (pid %d) delayed j=%d dev=%d jiffies=%d\n", - lock.context, current->pid, j, - dev->lock.lock_time, jiffies); - current->state = TASK_INTERRUPTIBLE; - current->policy |= SCHED_YIELD; - schedule_timeout(DRM_LOCK_SLICE-j); - DRM_DEBUG("jiffies=%d\n", jiffies); - } - } -#endif add_wait_queue(&dev->lock.lock_queue, &entry); for (;;) { current->state = TASK_INTERRUPTIBLE; @@ -617,9 +610,6 @@ int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, /* Contention */ atomic_inc(&dev->total_sleeps); -#if 1 - current->policy |= SCHED_YIELD; -#endif schedule(); if (signal_pending(current)) { ret = -ERESTARTSYS; @@ -630,32 +620,6 @@ int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, remove_wait_queue(&dev->lock.lock_queue, &entry); } -#if 0 - if (!ret && dev->last_context != lock.context && - lock.context != r128_res_ctx.handle && - dev->last_context != r128_res_ctx.handle) { - add_wait_queue(&dev->context_wait, &entry); - current->state = TASK_INTERRUPTIBLE; - /* PRE: dev->last_context != lock.context */ - r128_context_switch(dev, dev->last_context, lock.context); - /* POST: we will wait for the context - switch and will dispatch on a later call - when dev->last_context == lock.context - NOTE WE HOLD THE LOCK THROUGHOUT THIS - TIME! */ - current->policy |= SCHED_YIELD; - schedule(); - current->state = TASK_RUNNING; - remove_wait_queue(&dev->context_wait, &entry); - if (signal_pending(current)) { - ret = -EINTR; - } else if (dev->last_context != lock.context) { - DRM_ERROR("Context mismatch: %d %d\n", - dev->last_context, lock.context); - } - } -#endif - if (!ret) { sigemptyset(&dev->sigmask); sigaddset(&dev->sigmask, SIGSTOP); @@ -670,6 +634,7 @@ int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, } if (lock.flags & _DRM_LOCK_QUIESCENT) { /* Make hardware quiescent */ + DRM_DEBUG( "not quiescent!\n" ); #if 0 r128_quiescent(dev); #endif @@ -692,8 +657,8 @@ int r128_lock(struct inode *inode, struct file *filp, unsigned int cmd, } -int r128_unlock(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) +int r128_unlock(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) { drm_file_t *priv = filp->private_data; drm_device_t *dev = priv->dev; diff --git a/linux/r128_drv.h b/linux/r128_drv.h index 63b98c72..cffd0800 100644 --- a/linux/r128_drv.h +++ b/linux/r128_drv.h @@ -24,75 +24,136 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * Authors: Rickard E. (Rik) Faith - * Kevin E. Martin + * Authors: + * Rickard E. (Rik) Faith + * Kevin E. Martin + * Gareth Hughes * */ -#ifndef _R128_DRV_H_ -#define _R128_DRV_H_ +#ifndef __R128_DRV_H__ +#define __R128_DRV_H__ -typedef struct drm_r128_private { - int is_pci; - - int cce_mode; - int cce_fifo_size; - int cce_is_bm_mode; - int cce_secure; - - drm_r128_sarea_t *sarea_priv; - - __volatile__ u32 *ring_read_ptr; +typedef struct drm_r128_freelist { + unsigned int age; + drm_buf_t *buf; + struct drm_r128_freelist *next; + struct drm_r128_freelist *prev; +} drm_r128_freelist_t; - u32 *ring_start; - u32 *ring_end; - int ring_size; - int ring_sizel2qw; - int ring_entries; +typedef struct drm_r128_ring_buffer { + u32 *start; + u32 *end; + int size; + int size_l2qw; - int submit_age; + volatile u32 *head; + u32 tail; + u32 tail_mask; + int space; +} drm_r128_ring_buffer_t; - int usec_timeout; +typedef struct drm_r128_private { + drm_r128_ring_buffer_t ring; + drm_r128_sarea_t *sarea_priv; - drm_map_t *sarea; - drm_map_t *fb; - drm_map_t *agp_ring; - drm_map_t *agp_read_ptr; - drm_map_t *agp_vertbufs; - drm_map_t *agp_indbufs; - drm_map_t *agp_textures; - drm_map_t *mmio; + int cce_mode; + int cce_fifo_size; + int cce_secure; + int cce_running; + + drm_r128_freelist_t *head; + drm_r128_freelist_t *tail; + + int usec_timeout; + int is_pci; + + atomic_t idle_count; + + unsigned int fb_bpp; + unsigned int front_offset; + unsigned int front_pitch; + unsigned int back_offset; + unsigned int back_pitch; + + unsigned int depth_bpp; + unsigned int depth_offset; + unsigned int depth_pitch; + unsigned int span_offset; + + u32 front_pitch_offset_c; + u32 back_pitch_offset_c; + u32 depth_pitch_offset_c; + u32 span_pitch_offset_c; + + drm_map_t *sarea; + drm_map_t *fb; + drm_map_t *mmio; + drm_map_t *cce_ring; + drm_map_t *ring_rptr; + drm_map_t *buffers; + drm_map_t *agp_textures; } drm_r128_private_t; typedef struct drm_r128_buf_priv { - u32 age; + u32 age; + int prim; + int discard; + int dispatched; + drm_r128_freelist_t *list_entry; } drm_r128_buf_priv_t; /* r128_drv.c */ -extern int r128_version(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_open(struct inode *inode, struct file *filp); -extern int r128_release(struct inode *inode, struct file *filp); -extern int r128_ioctl(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_lock(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_unlock(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); - - /* r128_dma.c */ -extern int r128_init_cce(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_eng_reset(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_eng_flush(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_submit_pkt(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_cce_idle(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); -extern int r128_vertex_buf(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); +extern int r128_version( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_open( struct inode *inode, struct file *filp ); +extern int r128_release( struct inode *inode, struct file *filp ); +extern int r128_ioctl( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_lock( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_unlock( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); + + /* r128_cce.c */ +extern int r128_cce_init( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_start( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_stop( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_reset( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_idle( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_engine_reset( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_packet( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_buffers( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); + +extern void r128_freelist_reset( drm_device_t *dev ); +extern drm_buf_t *r128_freelist_get( drm_device_t *dev ); + +extern int r128_wait_ring( drm_r128_private_t *dev_priv, int n ); +extern void r128_update_ring_snapshot( drm_r128_private_t *dev_priv ); + + /* r128_state.c */ +extern int r128_cce_clear( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_swap( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_vertex( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_indices( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_blit( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_depth( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); +extern int r128_cce_stipple( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ); /* r128_bufs.c */ extern int r128_addbufs(struct inode *inode, struct file *filp, @@ -124,77 +185,204 @@ extern int r128_context_switch_complete(drm_device_t *dev, int new); * for Rage 128 kernel driver. */ -#define R128_PC_NGUI_CTLSTAT 0x0184 -# define R128_PC_FLUSH_ALL 0x00ff -# define R128_PC_BUSY (1 << 31) - -#define R128_CLOCK_CNTL_INDEX 0x0008 -#define R128_CLOCK_CNTL_DATA 0x000c -# define R128_PLL_WR_EN (1 << 7) - -#define R128_MCLK_CNTL 0x000f -# define R128_FORCE_GCP (1 << 16) -# define R128_FORCE_PIPE3D_CP (1 << 17) -# define R128_FORCE_RCP (1 << 18) - -#define R128_GEN_RESET_CNTL 0x00f0 -# define R128_SOFT_RESET_GUI (1 << 0) - -#define R128_PM4_BUFFER_CNTL 0x0704 -# define R128_PM4_NONPM4 (0 << 28) -# define R128_PM4_192PIO (1 << 28) -# define R128_PM4_192BM (2 << 28) -# define R128_PM4_128PIO_64INDBM (3 << 28) -# define R128_PM4_128BM_64INDBM (4 << 28) -# define R128_PM4_64PIO_128INDBM (5 << 28) -# define R128_PM4_64BM_128INDBM (6 << 28) -# define R128_PM4_64PIO_64VCBM_64INDBM (7 << 28) -# define R128_PM4_64BM_64VCBM_64INDBM (8 << 28) -# define R128_PM4_64PIO_64VCPIO_64INDPIO (15 << 28) - - -#define R128_PM4_BUFFER_DL_RPTR 0x0710 -#define R128_PM4_BUFFER_DL_WPTR 0x0714 -# define R128_PM4_BUFFER_DL_DONE (1 << 31) - -#define R128_PM4_VC_FPU_SETUP 0x071c - -#define R128_PM4_STAT 0x07b8 -# define R128_PM4_FIFOCNT_MASK 0x0fff -# define R128_PM4_BUSY (1 << 16) -# define R128_PM4_GUI_ACTIVE (1 << 31) - -#define R128_PM4_BUFFER_ADDR 0x07f0 -#define R128_PM4_MICRO_CNTL 0x07fc -# define R128_PM4_MICRO_FREERUN (1 << 30) - -#define R128_PM4_FIFO_DATA_EVEN 0x1000 -#define R128_PM4_FIFO_DATA_ODD 0x1004 - -#define R128_GUI_SCRATCH_REG0 0x15e0 -#define R128_GUI_SCRATCH_REG1 0x15e4 -#define R128_GUI_SCRATCH_REG2 0x15e8 -#define R128_GUI_SCRATCH_REG3 0x15ec -#define R128_GUI_SCRATCH_REG4 0x15f0 -#define R128_GUI_SCRATCH_REG5 0x15f4 - -#define R128_GUI_STAT 0x1740 -# define R128_GUI_FIFOCNT_MASK 0x0fff -# define R128_GUI_ACTIVE (1 << 31) - +#define R128_AUX_SC_CNTL 0x1660 +# define R128_AUX1_SC_EN (1 << 0) +# define R128_AUX1_SC_MODE_OR (0 << 1) +# define R128_AUX1_SC_MODE_NAND (1 << 1) +# define R128_AUX2_SC_EN (1 << 2) +# define R128_AUX2_SC_MODE_OR (0 << 3) +# define R128_AUX2_SC_MODE_NAND (1 << 3) +# define R128_AUX3_SC_EN (1 << 4) +# define R128_AUX3_SC_MODE_OR (0 << 5) +# define R128_AUX3_SC_MODE_NAND (1 << 5) +#define R128_AUX1_SC_LEFT 0x1664 +#define R128_AUX1_SC_RIGHT 0x1668 +#define R128_AUX1_SC_TOP 0x166c +#define R128_AUX1_SC_BOTTOM 0x1670 +#define R128_AUX2_SC_LEFT 0x1674 +#define R128_AUX2_SC_RIGHT 0x1678 +#define R128_AUX2_SC_TOP 0x167c +#define R128_AUX2_SC_BOTTOM 0x1680 +#define R128_AUX3_SC_LEFT 0x1684 +#define R128_AUX3_SC_RIGHT 0x1688 +#define R128_AUX3_SC_TOP 0x168c +#define R128_AUX3_SC_BOTTOM 0x1690 + +#define R128_BRUSH_DATA0 0x1480 +#define R128_BUS_CNTL 0x0030 +# define R128_BUS_MASTER_DIS (1 << 6) + +#define R128_CLOCK_CNTL_INDEX 0x0008 +#define R128_CLOCK_CNTL_DATA 0x000c +# define R128_PLL_WR_EN (1 << 7) + +#define R128_CONSTANT_COLOR_C 0x1d34 + +#define R128_DP_GUI_MASTER_CNTL 0x146c +# define R128_GMC_SRC_PITCH_OFFSET_CNTL (1 << 0) +# define R128_GMC_DST_PITCH_OFFSET_CNTL (1 << 1) +# define R128_GMC_BRUSH_SOLID_COLOR (13 << 4) +# define R128_GMC_BRUSH_NONE (15 << 4) +# define R128_GMC_DST_16BPP (4 << 8) +# define R128_GMC_DST_24BPP (5 << 8) +# define R128_GMC_DST_32BPP (6 << 8) +# define R128_GMC_DST_DATATYPE_SHIFT 8 +# define R128_GMC_SRC_DATATYPE_COLOR (3 << 12) +# define R128_DP_SRC_SOURCE_MEMORY (2 << 24) +# define R128_DP_SRC_SOURCE_HOST_DATA (3 << 24) +# define R128_GMC_CLR_CMP_CNTL_DIS (1 << 28) +# define R128_GMC_AUX_CLIP_DIS (1 << 29) +# define R128_GMC_WR_MSK_DIS (1 << 30) +# define R128_ROP3_S 0x00cc0000 +# define R128_ROP3_P 0x00f00000 +#define R128_DP_WRITE_MASK 0x16cc +#define R128_DST_PITCH_OFFSET_C 0x1c80 +# define R128_DST_TILE (1 << 31) + +#define R128_GEN_RESET_CNTL 0x00f0 +# define R128_SOFT_RESET_GUI (1 << 0) + +#define R128_GUI_SCRATCH_REG0 0x15e0 +#define R128_GUI_SCRATCH_REG1 0x15e4 +#define R128_GUI_SCRATCH_REG2 0x15e8 +#define R128_GUI_SCRATCH_REG3 0x15ec +#define R128_GUI_SCRATCH_REG4 0x15f0 +#define R128_GUI_SCRATCH_REG5 0x15f4 + +#define R128_GUI_STAT 0x1740 +# define R128_GUI_FIFOCNT_MASK 0x0fff +# define R128_GUI_ACTIVE (1 << 31) + +#define R128_MCLK_CNTL 0x000f +# define R128_FORCE_GCP (1 << 16) +# define R128_FORCE_PIPE3D_CP (1 << 17) +# define R128_FORCE_RCP (1 << 18) + +#define R128_PC_GUI_CTLSTAT 0x1748 +#define R128_PC_NGUI_CTLSTAT 0x0184 +# define R128_PC_FLUSH_GUI (3 << 0) +# define R128_PC_RI_GUI (1 << 2) +# define R128_PC_FLUSH_ALL 0x00ff +# define R128_PC_BUSY (1 << 31) + +#define R128_PRIM_TEX_CNTL_C 0x1cb0 + +#define R128_SCALE_3D_CNTL 0x1a00 +#define R128_SEC_TEX_CNTL_C 0x1d00 +#define R128_SEC_TEXTURE_BORDER_COLOR_C 0x1d3c +#define R128_SETUP_CNTL 0x1bc4 +#define R128_STEN_REF_MASK_C 0x1d40 + +#define R128_TEX_CNTL_C 0x1c9c +# define R128_TEX_CACHE_FLUSH (1 << 23) + +#define R128_WINDOW_XY_OFFSET 0x1bcc + + +/* CCE registers + */ +#define R128_PM4_BUFFER_OFFSET 0x0700 +#define R128_PM4_BUFFER_CNTL 0x0704 +# define R128_PM4_MASK (15 << 28) +# define R128_PM4_NONPM4 (0 << 28) +# define R128_PM4_192PIO (1 << 28) +# define R128_PM4_192BM (2 << 28) +# define R128_PM4_128PIO_64INDBM (3 << 28) +# define R128_PM4_128BM_64INDBM (4 << 28) +# define R128_PM4_64PIO_128INDBM (5 << 28) +# define R128_PM4_64BM_128INDBM (6 << 28) +# define R128_PM4_64PIO_64VCBM_64INDBM (7 << 28) +# define R128_PM4_64BM_64VCBM_64INDBM (8 << 28) +# define R128_PM4_64PIO_64VCPIO_64INDPIO (15 << 28) + +#define R128_PM4_BUFFER_WM_CNTL 0x0708 +# define R128_WMA_SHIFT 0 +# define R128_WMB_SHIFT 8 +# define R128_WMC_SHIFT 16 +# define R128_WB_WM_SHIFT 24 + +#define R128_PM4_BUFFER_DL_RPTR_ADDR 0x070c +#define R128_PM4_BUFFER_DL_RPTR 0x0710 +#define R128_PM4_BUFFER_DL_WPTR 0x0714 +# define R128_PM4_BUFFER_DL_DONE (1 << 31) + +#define R128_PM4_VC_FPU_SETUP 0x071c + +#define R128_PM4_IW_INDOFF 0x0738 +#define R128_PM4_IW_INDSIZE 0x073c + +#define R128_PM4_STAT 0x07b8 +# define R128_PM4_FIFOCNT_MASK 0x0fff +# define R128_PM4_BUSY (1 << 16) +# define R128_PM4_GUI_ACTIVE (1 << 31) + +#define R128_PM4_MICROCODE_ADDR 0x07d4 +#define R128_PM4_MICROCODE_RADDR 0x07d8 +#define R128_PM4_MICROCODE_DATAH 0x07dc +#define R128_PM4_MICROCODE_DATAL 0x07e0 + +#define R128_PM4_BUFFER_ADDR 0x07f0 +#define R128_PM4_MICRO_CNTL 0x07fc +# define R128_PM4_MICRO_FREERUN (1 << 30) + +#define R128_PM4_FIFO_DATA_EVEN 0x1000 +#define R128_PM4_FIFO_DATA_ODD 0x1004 + + +/* CCE command packets + */ +#define R128_CCE_PACKET0 0x00000000 +#define R128_CCE_PACKET1 0x40000000 +#define R128_CCE_PACKET2 0x80000000 +#define R128_CCE_PACKET3 0xC0000000 +# define R128_CNTL_HOSTDATA_BLT 0x00009400 +# define R128_CNTL_PAINT_MULTI 0x00009A00 +# define R128_CNTL_BITBLT_MULTI 0x00009B00 +# define R128_3D_RNDR_GEN_INDX_PRIM 0x00002300 + +#define R128_CCE_PACKET_MASK 0xC0000000 +#define R128_CCE_PACKET_COUNT_MASK 0x3fff0000 +#define R128_CCE_PACKET0_REG_MASK 0x000007ff +#define R128_CCE_PACKET1_REG0_MASK 0x000007ff +#define R128_CCE_PACKET1_REG1_MASK 0x003ff800 + +#define R128_CCE_VC_CNTL_PRIM_TYPE_NONE 0x00000000 +#define R128_CCE_VC_CNTL_PRIM_TYPE_POINT 0x00000001 +#define R128_CCE_VC_CNTL_PRIM_TYPE_LINE 0x00000002 +#define R128_CCE_VC_CNTL_PRIM_TYPE_POLY_LINE 0x00000003 +#define R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST 0x00000004 +#define R128_CCE_VC_CNTL_PRIM_TYPE_TRI_FAN 0x00000005 +#define R128_CCE_VC_CNTL_PRIM_TYPE_TRI_STRIP 0x00000006 +#define R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2 0x00000007 +#define R128_CCE_VC_CNTL_PRIM_WALK_IND 0x00000010 +#define R128_CCE_VC_CNTL_PRIM_WALK_LIST 0x00000020 +#define R128_CCE_VC_CNTL_PRIM_WALK_RING 0x00000030 +#define R128_CCE_VC_CNTL_NUM_SHIFT 16 + +#define R128_DATATYPE_CI8 2 +#define R128_DATATYPE_ARGB1555 3 +#define R128_DATATYPE_RGB565 4 +#define R128_DATATYPE_RGB888 5 +#define R128_DATATYPE_ARGB8888 6 +#define R128_DATATYPE_RGB332 7 +#define R128_DATATYPE_RGB8 9 +#define R128_DATATYPE_ARGB4444 15 + +/* Constants */ +#define R128_AGP_OFFSET 0x02000000 + +#define R128_WATERMARK_L 16 +#define R128_WATERMARK_M 8 +#define R128_WATERMARK_N 8 +#define R128_WATERMARK_K 128 -/* CCE command packets */ -#define R128_CCE_PACKET0 0x00000000 -#define R128_CCE_PACKET1 0x40000000 -#define R128_CCE_PACKET2 0x80000000 -# define R128_CCE_PACKET_MASK 0xC0000000 -# define R128_CCE_PACKET_COUNT_MASK 0x3fff0000 -# define R128_CCE_PACKET0_REG_MASK 0x000007ff -# define R128_CCE_PACKET1_REG0_MASK 0x000007ff -# define R128_CCE_PACKET1_REG1_MASK 0x003ff800 +#define R128_MAX_USEC_TIMEOUT 100000 /* 100 ms */ +#define R128_LAST_FRAME_REG R128_GUI_SCRATCH_REG0 +#define R128_LAST_DISPATCH_REG R128_GUI_SCRATCH_REG1 +#define R128_MAX_VB_AGE 0xffffffff -#define R128_MAX_USEC_TIMEOUT 100000 /* 100 ms */ +#define R128_MAX_VB_VERTS (0xffff) #define R128_BASE(reg) ((u32)(dev_priv->mmio->handle)) @@ -221,4 +409,58 @@ extern int R128_READ_PLL(drm_device_t *dev, int addr); #define R128CCE2(p) ((p)) #define R128CCE3(p,n) ((p) | ((n) << 16)) -#endif + + + +#define CCE_PACKET0( reg, n ) (R128_CCE_PACKET0 | \ + ((n) << 16) | ((reg) >> 2)) +#define CCE_PACKET1( reg0, reg1 ) (R128_CCE_PACKET1 | \ + (((reg1) >> 2) << 11) | ((reg0) >> 2)) +#define CCE_PACKET2() (R128_CCE_PACKET2) +#define CCE_PACKET3( pkt, n ) (R128_CCE_PACKET3 | \ + (pkt) | ((n) << 16)) + + +#define r128_flush_write_combine() mb() + + +#define R128_VERBOSE 0 + +#define RING_LOCALS int write; unsigned int tail_mask; volatile u32 *ring; + +#define BEGIN_RING( n ) do { \ + if ( R128_VERBOSE ) { \ + DRM_INFO( "BEGIN_RING( %d ) in %s\n", \ + n, __FUNCTION__ ); \ + } \ + if ( dev_priv->ring.space < n * sizeof(u32) ) { \ + r128_wait_ring( dev_priv, n * sizeof(u32) ); \ + } \ + dev_priv->ring.space -= n * sizeof(u32); \ + ring = dev_priv->ring.start; \ + write = dev_priv->ring.tail; \ + tail_mask = dev_priv->ring.tail_mask; \ +} while (0) + +#define ADVANCE_RING() do { \ + if ( R128_VERBOSE ) { \ + DRM_INFO( "ADVANCE_RING() tail=0x%06x wr=0x%06x\n", \ + write, dev_priv->ring.tail ); \ + } \ + r128_flush_write_combine(); \ + dev_priv->ring.tail = write; \ + R128_WRITE( R128_PM4_BUFFER_DL_WPTR, write ); \ +} while (0) + +#define OUT_RING( x ) do { \ + if ( R128_VERBOSE ) { \ + DRM_INFO( " OUT_RING( 0x%08x ) at 0x%x\n", \ + (unsigned int)(x), write ); \ + } \ + ring[write++] = x; \ + write &= tail_mask; \ +} while (0) + +#define R128_PERFORMANCE_BOXES 0 + +#endif /* __R128_DRV_H__ */ diff --git a/linux/r128_state.c b/linux/r128_state.c new file mode 100644 index 00000000..faad25a2 --- /dev/null +++ b/linux/r128_state.c @@ -0,0 +1,1630 @@ +/* r128_state.c -- State support for r128 -*- linux-c -*- + * Created: Thu Jan 27 02:53:43 2000 by gareth@valinux.com + * + * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California. + * All Rights Reserved. + * + * 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 (including the next + * paragraph) 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 + * PRECISION INSIGHT AND/OR ITS SUPPLIERS 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. + * + * Authors: + * Gareth Hughes + * + */ + +#define __NO_VERSION__ +#include "drmP.h" +#include "r128_drv.h" +#include "drm.h" + + +/* ================================================================ + * CCE hardware state programming functions + */ + +static void r128_emit_clip_rects( drm_r128_private_t *dev_priv, + drm_clip_rect_t *boxes, int count ) +{ + u32 aux_sc_cntl = 0x00000000; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 17 ); + + if ( count >= 1 ) { + OUT_RING( CCE_PACKET0( R128_AUX1_SC_LEFT, 3 ) ); + OUT_RING( boxes[0].x1 ); + OUT_RING( boxes[0].x2 - 1 ); + OUT_RING( boxes[0].y1 ); + OUT_RING( boxes[0].y2 - 1 ); + + aux_sc_cntl |= (R128_AUX1_SC_EN | R128_AUX1_SC_MODE_OR); + } + if ( count >= 2 ) { + OUT_RING( CCE_PACKET0( R128_AUX2_SC_LEFT, 3 ) ); + OUT_RING( boxes[1].x1 ); + OUT_RING( boxes[1].x2 - 1 ); + OUT_RING( boxes[1].y1 ); + OUT_RING( boxes[1].y2 - 1 ); + + aux_sc_cntl |= (R128_AUX2_SC_EN | R128_AUX2_SC_MODE_OR); + } + if ( count >= 3 ) { + OUT_RING( CCE_PACKET0( R128_AUX3_SC_LEFT, 3 ) ); + OUT_RING( boxes[2].x1 ); + OUT_RING( boxes[2].x2 - 1 ); + OUT_RING( boxes[2].y1 ); + OUT_RING( boxes[2].y2 - 1 ); + + aux_sc_cntl |= (R128_AUX3_SC_EN | R128_AUX3_SC_MODE_OR); + } + + OUT_RING( CCE_PACKET0( R128_AUX_SC_CNTL, 0 ) ); + OUT_RING( aux_sc_cntl ); + + ADVANCE_RING(); +} + +static inline void r128_emit_core( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_SCALE_3D_CNTL, 0 ) ); + OUT_RING( ctx->scale_3d_cntl ); + + ADVANCE_RING(); +} + +static inline void r128_emit_context( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 13 ); + + OUT_RING( CCE_PACKET0( R128_DST_PITCH_OFFSET_C, 11 ) ); + OUT_RING( ctx->dst_pitch_offset_c ); + OUT_RING( ctx->dp_gui_master_cntl_c ); + OUT_RING( ctx->sc_top_left_c ); + OUT_RING( ctx->sc_bottom_right_c ); + OUT_RING( ctx->z_offset_c ); + OUT_RING( ctx->z_pitch_c ); + OUT_RING( ctx->z_sten_cntl_c ); + OUT_RING( ctx->tex_cntl_c ); + OUT_RING( ctx->misc_3d_state_cntl_reg ); + OUT_RING( ctx->texture_clr_cmp_clr_c ); + OUT_RING( ctx->texture_clr_cmp_msk_c ); + OUT_RING( ctx->fog_color_c ); + + ADVANCE_RING(); +} + +static inline void r128_emit_setup( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 3 ); + + OUT_RING( CCE_PACKET1( R128_SETUP_CNTL, R128_PM4_VC_FPU_SETUP ) ); + OUT_RING( ctx->setup_cntl ); + OUT_RING( ctx->pm4_vc_fpu_setup ); + + ADVANCE_RING(); +} + +static inline void r128_emit_masks( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 5 ); + + OUT_RING( CCE_PACKET0( R128_DP_WRITE_MASK, 0 ) ); + OUT_RING( ctx->dp_write_mask ); + + OUT_RING( CCE_PACKET0( R128_STEN_REF_MASK_C, 1 ) ); + OUT_RING( ctx->sten_ref_mask_c ); + OUT_RING( ctx->plane_3d_mask_c ); + + ADVANCE_RING(); +} + +static inline void r128_emit_window( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_WINDOW_XY_OFFSET, 0 ) ); + OUT_RING( ctx->window_xy_offset ); + + ADVANCE_RING(); +} + +static inline void r128_emit_tex0( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_context_regs_t *ctx = &sarea_priv->context_state; + drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[0]; + int i; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 7 + R128_TEX_MAXLEVELS ); + + OUT_RING( CCE_PACKET0( R128_PRIM_TEX_CNTL_C, + 2 + R128_TEX_MAXLEVELS ) ); + OUT_RING( tex->tex_cntl ); + OUT_RING( tex->tex_combine_cntl ); + OUT_RING( ctx->tex_size_pitch_c ); + for ( i = 0 ; i < R128_TEX_MAXLEVELS ; i++ ) { + OUT_RING( tex->tex_offset[i] ); + } + + OUT_RING( CCE_PACKET0( R128_CONSTANT_COLOR_C, 1 ) ); + OUT_RING( ctx->constant_color_c ); + OUT_RING( tex->tex_border_color ); + + ADVANCE_RING(); +} + +static inline void r128_emit_tex1( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[1]; + int i; + RING_LOCALS; + DRM_DEBUG( " %s\n", __FUNCTION__ ); + + BEGIN_RING( 5 + R128_TEX_MAXLEVELS ); + + OUT_RING( CCE_PACKET0( R128_SEC_TEX_CNTL_C, + 1 + R128_TEX_MAXLEVELS ) ); + OUT_RING( tex->tex_cntl ); + OUT_RING( tex->tex_combine_cntl ); + for ( i = 0 ; i < R128_TEX_MAXLEVELS ; i++ ) { + OUT_RING( tex->tex_offset[i] ); + } + + OUT_RING( CCE_PACKET0( R128_SEC_TEXTURE_BORDER_COLOR_C, 0 ) ); + OUT_RING( tex->tex_border_color ); + + ADVANCE_RING(); +} + +static inline void r128_emit_state( drm_r128_private_t *dev_priv ) +{ + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + unsigned int dirty = sarea_priv->dirty; + + DRM_DEBUG( "%s: dirty=0x%08x\n", __FUNCTION__, dirty ); + + if ( dirty & R128_UPLOAD_CORE ) { + r128_emit_core( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_CORE; + } + + if ( dirty & R128_UPLOAD_CONTEXT ) { + r128_emit_context( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_CONTEXT; + } + + if ( dirty & R128_UPLOAD_SETUP ) { + r128_emit_setup( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_SETUP; + } + + if ( dirty & R128_UPLOAD_MASKS ) { + r128_emit_masks( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_MASKS; + } + + if ( dirty & R128_UPLOAD_WINDOW ) { + r128_emit_window( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_WINDOW; + } + + if ( dirty & R128_UPLOAD_TEX0 ) { + r128_emit_tex0( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_TEX0; + } + + if ( dirty & R128_UPLOAD_TEX1 ) { + r128_emit_tex1( dev_priv ); + sarea_priv->dirty &= ~R128_UPLOAD_TEX1; + } + + /* Turn off the texture cache flushing */ + sarea_priv->context_state.tex_cntl_c &= ~R128_TEX_CACHE_FLUSH; + + sarea_priv->dirty &= ~R128_REQUIRE_QUIESCENCE; +} + + +#if R128_PERFORMANCE_BOXES +/* ================================================================ + * Performance monitoring functions + */ + +static void r128_clear_box( drm_r128_private_t *dev_priv, + int x, int y, int w, int h, + int r, int g, int b ) +{ + u32 pitch, offset; + u32 fb_bpp, color; + RING_LOCALS; + + switch ( dev_priv->fb_bpp ) { + case 16: + fb_bpp = R128_GMC_DST_16BPP; + color = (((r & 0xf8) << 8) | + ((g & 0xfc) << 3) | + ((b & 0xf8) >> 3)); + break; + case 24: + fb_bpp = R128_GMC_DST_24BPP; + color = ((r << 16) | (g << 8) | b); + break; + case 32: + default: + fb_bpp = R128_GMC_DST_32BPP; + color = (((0xff) << 24) | (r << 16) | (g << 8) | b); + break; + } + + offset = dev_priv->back_offset; + pitch = dev_priv->back_pitch >> 3; + + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | fb_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS ); + + OUT_RING( (pitch << 21) | (offset >> 5) ); + OUT_RING( color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); +} + +static void r128_cce_performance_boxes( drm_r128_private_t *dev_priv ) +{ + if ( atomic_read( &dev_priv->idle_count ) == 0 ) { + r128_clear_box( dev_priv, 64, 4, 8, 8, 0, 255, 0 ); + } else { + atomic_set( &dev_priv->idle_count, 0 ); + } +} + +#endif + + +/* ================================================================ + * CCE command dispatch functions + */ + +static void r128_print_dirty( const char *msg, unsigned int flags ) +{ + DRM_INFO( "%s: (0x%x) %s%s%s%s%s%s%s%s%s\n", + msg, + flags, + (flags & R128_UPLOAD_CORE) ? "core, " : "", + (flags & R128_UPLOAD_CONTEXT) ? "context, " : "", + (flags & R128_UPLOAD_SETUP) ? "setup, " : "", + (flags & R128_UPLOAD_TEX0) ? "tex0, " : "", + (flags & R128_UPLOAD_TEX1) ? "tex1, " : "", + (flags & R128_UPLOAD_MASKS) ? "masks, " : "", + (flags & R128_UPLOAD_WINDOW) ? "window, " : "", + (flags & R128_UPLOAD_CLIPRECTS) ? "cliprects, " : "", + (flags & R128_REQUIRE_QUIESCENCE) ? "quiescence, " : "" ); +} + +static void r128_cce_dispatch_clear( drm_device_t *dev, + unsigned int flags, + int cx, int cy, int cw, int ch, + unsigned int clear_color, + unsigned int clear_depth, + unsigned int color_mask, + unsigned int depth_mask ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + int nbox = sarea_priv->nbox; + drm_clip_rect_t *pbox = sarea_priv->boxes; + u32 fb_bpp, depth_bpp; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + switch ( dev_priv->fb_bpp ) { + case 16: + fb_bpp = R128_GMC_DST_16BPP; + break; + case 24: + fb_bpp = R128_GMC_DST_24BPP; + break; + case 32: + default: + fb_bpp = R128_GMC_DST_32BPP; + break; + } + switch ( dev_priv->depth_bpp ) { + case 16: + depth_bpp = R128_GMC_DST_16BPP; + break; + case 24: + depth_bpp = R128_GMC_DST_32BPP; + break; + case 32: + depth_bpp = R128_GMC_DST_32BPP; + break; + default: + return; + } + + for ( i = 0 ; i < nbox ; i++ ) { + int x = pbox[i].x1; + int y = pbox[i].y1; + int w = pbox[i].x2 - x; + int h = pbox[i].y2 - y; + + DRM_DEBUG( "dispatch clear %d,%d-%d,%d flags 0x%x\n", + pbox[i].x1, pbox[i].y1, pbox[i].x2, + pbox[i].y2, flags ); + + if ( flags & (R128_FRONT | R128_BACK) ) { + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_DP_WRITE_MASK, 0 ) ); + OUT_RING( color_mask ); + + ADVANCE_RING(); + } + + if ( flags & R128_FRONT ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | fb_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS ); + + OUT_RING( dev_priv->front_pitch_offset_c ); + OUT_RING( clear_color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); + } + + if ( flags & R128_BACK ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | fb_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS ); + + OUT_RING( dev_priv->back_pitch_offset_c ); + OUT_RING( clear_color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); + } + + if ( flags & R128_DEPTH ) { + BEGIN_RING( 8 ); + + OUT_RING( CCE_PACKET0( R128_DP_WRITE_MASK, 0 ) ); + OUT_RING( depth_mask ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( clear_depth ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); + } + } +} + +static void r128_cce_dispatch_swap( drm_device_t *dev ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + int nbox = sarea_priv->nbox; + drm_clip_rect_t *pbox = sarea_priv->boxes; + u32 fb_bpp; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + +#if R128_PERFORMANCE_BOXES + /* Do some trivial performance monitoring... + */ + r128_cce_performance_boxes( dev_priv ); +#endif + + switch ( dev_priv->fb_bpp ) { + case 16: + fb_bpp = R128_GMC_DST_16BPP; + break; + case 24: + fb_bpp = R128_GMC_DST_24BPP; + break; + case 32: + default: + fb_bpp = R128_GMC_DST_32BPP; + break; + } + + for ( i = 0 ; i < nbox ; i++ ) { + int x = pbox[i].x1; + int y = pbox[i].y1; + int w = pbox[i].x2 - x; + int h = pbox[i].y2 - y; + + BEGIN_RING( 7 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_BITBLT_MULTI, 5 ) ); + OUT_RING( R128_GMC_SRC_PITCH_OFFSET_CNTL + | R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_NONE + | fb_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_S + | R128_DP_SRC_SOURCE_MEMORY + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS + | R128_GMC_WR_MSK_DIS ); + +#if 1 + OUT_RING( dev_priv->back_pitch_offset_c ); + OUT_RING( dev_priv->front_pitch_offset_c ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); +#else + OUT_RING( dev_priv->depth_pitch_offset_c /*& ~R128_DST_TILE*/ ); + OUT_RING( dev_priv->front_pitch_offset_c ); + + OUT_RING( (0 << 16) | 0 ); + OUT_RING( (0 << 16) | 0 ); + OUT_RING( (800 << 16) | 600 ); +#endif + + ADVANCE_RING(); + } + + /* Increment the frame counter. The client-side 3D driver must + * throttle the framerate by waiting for this value before + * performing the swapbuffer ioctl. + */ + dev_priv->sarea_priv->last_frame++; + + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_LAST_FRAME_REG, 0 ) ); + OUT_RING( dev_priv->sarea_priv->last_frame ); + + ADVANCE_RING(); +} + +static void r128_cce_dispatch_vertex( drm_device_t *dev, + drm_buf_t *buf ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_buf_priv_t *buf_priv = buf->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + int format = sarea_priv->vc_format; + int offset = dev_priv->buffers->offset + buf->offset - dev->agp->base; + int size = buf->used; + int prim = buf_priv->prim; + int i = 0; + RING_LOCALS; + DRM_DEBUG( "%s: buf=%d nbox=%d\n", + __FUNCTION__, buf->idx, sarea_priv->nbox ); + + r128_update_ring_snapshot( dev_priv ); + + if ( 0 ) + r128_print_dirty( "dispatch_vertex", sarea_priv->dirty ); + + if ( buf->used ) { + buf_priv->dispatched = 1; + + if ( sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS ) { + r128_emit_state( dev_priv ); + } + + do { + /* Emit the next set of up to three cliprects */ + if ( i < sarea_priv->nbox ) { + r128_emit_clip_rects( dev_priv, + &sarea_priv->boxes[i], + sarea_priv->nbox - i ); + } + + /* Emit the vertex buffer rendering commands */ + BEGIN_RING( 5 ); + + OUT_RING( CCE_PACKET3( R128_3D_RNDR_GEN_INDX_PRIM, 3 ) ); + OUT_RING( offset ); + OUT_RING( size ); + OUT_RING( format ); + OUT_RING( prim | R128_CCE_VC_CNTL_PRIM_WALK_LIST | + (size << R128_CCE_VC_CNTL_NUM_SHIFT) ); + + ADVANCE_RING(); + + i += 3; + } while ( i < sarea_priv->nbox ); + } + + if ( buf_priv->discard ) { + buf_priv->age = dev_priv->sarea_priv->last_dispatch; + + /* Emit the vertex buffer age */ + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_LAST_DISPATCH_REG, 0 ) ); + OUT_RING( buf_priv->age ); + + ADVANCE_RING(); + + buf->pending = 1; + buf->used = 0; + /* FIXME: Check dispatched field */ + buf_priv->dispatched = 0; + } + + dev_priv->sarea_priv->last_dispatch++; + +#if 0 + if ( dev_priv->submit_age == R128_MAX_VB_AGE ) { + ret = r128_do_cce_idle( dev_priv ); + if ( ret < 0 ) return ret; + dev_priv->submit_age = 0; + r128_freelist_reset( dev ); + } +#endif + + sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS; + sarea_priv->nbox = 0; +} + + + + +static void r128_cce_dispatch_indirect( drm_device_t *dev, + drm_buf_t *buf, + int start, int end ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_buf_priv_t *buf_priv = buf->dev_private; + RING_LOCALS; + DRM_DEBUG( "indirect: buf=%d s=0x%x e=0x%x\n", + buf->idx, start, end ); + + r128_update_ring_snapshot( dev_priv ); + + if ( start != end ) { + int offset = (dev_priv->buffers->offset - dev->agp->base + + buf->offset + start); + int dwords = (end - start + 3) / sizeof(u32); + + /* Indirect buffer data must be an even number of + * dwords, so if we've been given an odd number we must + * pad the data with a Type-2 CCE packet. + */ + if ( dwords & 1 ) { + u32 *data = (u32 *) + ((char *)dev_priv->buffers->handle + + buf->offset + start); + data[dwords++] = R128_CCE_PACKET2; + } + + buf_priv->dispatched = 1; + + /* Fire off the indirect buffer */ + BEGIN_RING( 3 ); + + OUT_RING( CCE_PACKET0( R128_PM4_IW_INDOFF, 1 ) ); + OUT_RING( offset ); + OUT_RING( dwords ); + + ADVANCE_RING(); + } + + if ( buf_priv->discard ) { + buf_priv->age = dev_priv->sarea_priv->last_dispatch; + + /* Emit the indirect buffer age */ + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_LAST_DISPATCH_REG, 0 ) ); + OUT_RING( buf_priv->age ); + + ADVANCE_RING(); + + buf->pending = 1; + buf->used = 0; + /* FIXME: Check dispatched field */ + buf_priv->dispatched = 0; + } + + dev_priv->sarea_priv->last_dispatch++; + +#if 0 + if ( dev_priv->submit_age == R128_MAX_VB_AGE ) { + ret = r128_do_cce_idle( dev_priv ); + if ( ret < 0 ) return ret; + dev_priv->submit_age = 0; + r128_freelist_reset( dev ); + } +#endif +} + +static void r128_cce_dispatch_indices( drm_device_t *dev, + drm_buf_t *buf, + int start, int end, + int count ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_buf_priv_t *buf_priv = buf->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + int format = sarea_priv->vc_format; + int offset = dev_priv->buffers->offset - dev->agp->base; + int prim = buf_priv->prim; + u32 *data; + int dwords; + int i = 0; + RING_LOCALS; + DRM_DEBUG( "indices: s=%d e=%d c=%d\n", start, end, count ); + + r128_update_ring_snapshot( dev_priv ); + + if ( 0 ) + r128_print_dirty( "dispatch_indices", sarea_priv->dirty ); + + if ( start != end ) { + buf_priv->dispatched = 1; + + if ( sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS ) { + r128_emit_state( dev_priv ); + } + + dwords = (end - start + 3) / sizeof(u32); + + data = (u32 *)((char *)dev_priv->buffers->handle + + buf->offset + start); + + data[0] = CCE_PACKET3( R128_3D_RNDR_GEN_INDX_PRIM, dwords-2 ); + + data[1] = offset; + data[2] = R128_MAX_VB_VERTS; + data[3] = format; + data[4] = (prim | R128_CCE_VC_CNTL_PRIM_WALK_IND | + (count << 16)); + + if ( count & 0x1 ) { + data[dwords-1] &= 0x0000ffff; + } + + do { + /* Emit the next set of up to three cliprects */ + if ( i < sarea_priv->nbox ) { + r128_emit_clip_rects( dev_priv, + &sarea_priv->boxes[i], + sarea_priv->nbox - i ); + } + + r128_cce_dispatch_indirect( dev, buf, start, end ); + + i += 3; + } while ( i < sarea_priv->nbox ); + } + + if ( buf_priv->discard ) { + buf_priv->age = dev_priv->sarea_priv->last_dispatch; + + /* Emit the vertex buffer age */ + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_LAST_DISPATCH_REG, 0 ) ); + OUT_RING( buf_priv->age ); + + ADVANCE_RING(); + + buf->pending = 1; + /* FIXME: Check dispatched field */ + buf_priv->dispatched = 0; + } + + dev_priv->sarea_priv->last_dispatch++; + +#if 0 + if ( dev_priv->submit_age == R128_MAX_VB_AGE ) { + ret = r128_do_cce_idle( dev_priv ); + if ( ret < 0 ) return ret; + dev_priv->submit_age = 0; + r128_freelist_reset( dev ); + } +#endif + + sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS; + sarea_priv->nbox = 0; +} + +static int r128_cce_dispatch_blit( drm_device_t *dev, + drm_r128_blit_t *blit ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_r128_buf_priv_t *buf_priv; + u32 *data; + int dword_shift, dwords; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + /* The compiler won't optimize away a division by a variable, + * even if the only legal values are powers of two. Thus, we'll + * use a shift instead. + */ + switch ( blit->format ) { + case R128_DATATYPE_ARGB1555: + case R128_DATATYPE_RGB565: + case R128_DATATYPE_ARGB4444: + dword_shift = 1; + break; + case R128_DATATYPE_ARGB8888: + dword_shift = 0; + break; + default: + DRM_ERROR( "invalid blit format %d\n", blit->format ); + return -EINVAL; + } + + /* Flush the pixel cache, and mark the contents as Read Invalid. + * This ensures no pixel data gets mixed up with the texture + * data from the host data blit, otherwise part of the texture + * image may be corrupted. + */ + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_PC_GUI_CTLSTAT, 0 ) ); + OUT_RING( R128_PC_RI_GUI | R128_PC_FLUSH_GUI ); + + ADVANCE_RING(); + + /* Dispatch the indirect buffer. + */ + buf = dma->buflist[blit->idx]; + buf_priv = buf->dev_private; + + if ( buf->pid != current->pid ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + current->pid, buf->pid ); + return -EINVAL; + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", blit->idx ); + return -EINVAL; + } + + buf_priv->discard = 1; + + dwords = (blit->width * blit->height) >> dword_shift; + + data = (u32 *)((char *)dev_priv->buffers->handle + buf->offset); + + data[0] = CCE_PACKET3( R128_CNTL_HOSTDATA_BLT, dwords + 6 ); + data[1] = ( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_NONE + | (blit->format << 8) + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_S + | R128_DP_SRC_SOURCE_HOST_DATA + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_AUX_CLIP_DIS + | R128_GMC_WR_MSK_DIS ); + + data[2] = (blit->pitch << 21) | (blit->offset >> 5); + data[3] = 0xffffffff; + data[4] = 0xffffffff; + data[5] = (blit->y << 16) | blit->x; + data[6] = (blit->height << 16) | blit->width; + data[7] = dwords; + + buf->used = (dwords + 8) * sizeof(u32); + + r128_cce_dispatch_indirect( dev, buf, 0, buf->used ); + + /* Flush the pixel cache after the blit completes. This ensures + * the texture data is written out to memory before rendering + * continues. + */ + BEGIN_RING( 2 ); + + OUT_RING( CCE_PACKET0( R128_PC_GUI_CTLSTAT, 0 ) ); + OUT_RING( R128_PC_FLUSH_GUI ); + + ADVANCE_RING(); + + return 0; +} + + +/* ================================================================ + * Tiled depth buffer management + */ + +static int r128_cce_dispatch_write_span( drm_device_t *dev, + drm_r128_depth_t *depth ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + int count, x, y; + u32 *buffer; + u8 *mask; + u32 depth_bpp; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + switch ( dev_priv->depth_bpp ) { + case 16: + depth_bpp = R128_GMC_DST_16BPP; + break; + case 24: + depth_bpp = R128_GMC_DST_32BPP; + break; + case 32: + depth_bpp = R128_GMC_DST_32BPP; + break; + default: + return -EINVAL; + } + + count = depth->n; + if ( copy_from_user( &x, depth->x, sizeof(x) ) ) { + return -EFAULT; + } + if ( copy_from_user( &y, depth->y, sizeof(y) ) ) { + return -EFAULT; + } + + buffer = kmalloc( depth->n * sizeof(u32), 0 ); + if ( buffer == NULL ) + return -ENOMEM; + if ( copy_from_user( buffer, depth->buffer, + depth->n * sizeof(u32) ) ) { + kfree( buffer ); + return -EFAULT; + } + + if ( depth->mask ) { + mask = kmalloc( depth->n * sizeof(u8), 0 ); + if ( mask == NULL ) { + kfree( buffer ); + return -ENOMEM; + } + if ( copy_from_user( mask, depth->mask, + depth->n * sizeof(u8) ) ) { + kfree( buffer ); + kfree( mask ); + return -EFAULT; + } + + for ( i = 0 ; i < count ; i++, x++ ) { + if ( mask[i] ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, + 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( buffer[i] ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (1 << 16) | 1 ); + + ADVANCE_RING(); + } + } + + kfree( mask ); + } else { + for ( i = 0 ; i < count ; i++, x++ ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( buffer[i] ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (1 << 16) | 1 ); + + ADVANCE_RING(); + } + } + + kfree( buffer ); + + return 0; +} + +static int r128_cce_dispatch_write_pixels( drm_device_t *dev, + drm_r128_depth_t *depth ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + int count, *x, *y; + u32 *buffer; + u8 *mask; + u32 depth_bpp; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + switch ( dev_priv->depth_bpp ) { + case 16: + depth_bpp = R128_GMC_DST_16BPP; + break; + case 24: + depth_bpp = R128_GMC_DST_32BPP; + break; + case 32: + depth_bpp = R128_GMC_DST_32BPP; + break; + default: + return -EINVAL; + } + + count = depth->n; + + x = kmalloc( count * sizeof(*x), 0 ); + if ( x == NULL ) { + return -ENOMEM; + } + y = kmalloc( count * sizeof(*y), 0 ); + if ( y == NULL ) { + kfree( x ); + return -ENOMEM; + } + if ( copy_from_user( x, depth->x, count * sizeof(int) ) ) { + kfree( x ); + kfree( y ); + return -EFAULT; + } + if ( copy_from_user( y, depth->y, count * sizeof(int) ) ) { + kfree( x ); + kfree( y ); + return -EFAULT; + } + + buffer = kmalloc( depth->n * sizeof(u32), 0 ); + if ( buffer == NULL ) { + kfree( x ); + kfree( y ); + return -ENOMEM; + } + if ( copy_from_user( buffer, depth->buffer, + depth->n * sizeof(u32) ) ) { + kfree( x ); + kfree( y ); + kfree( buffer ); + return -EFAULT; + } + + if ( depth->mask ) { + mask = kmalloc( depth->n * sizeof(u8), 0 ); + if ( mask == NULL ) { + kfree( x ); + kfree( y ); + kfree( buffer ); + return -ENOMEM; + } + if ( copy_from_user( mask, depth->mask, + depth->n * sizeof(u8) ) ) { + kfree( x ); + kfree( y ); + kfree( buffer ); + kfree( mask ); + return -EFAULT; + } + + for ( i = 0 ; i < count ; i++ ) { + if ( mask[i] ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, + 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( buffer[i] ); + + OUT_RING( (x[i] << 16) | y[i] ); + OUT_RING( (1 << 16) | 1 ); + + ADVANCE_RING(); + } + } + + kfree( mask ); + } else { + for ( i = 0 ; i < count ; i++ ) { + BEGIN_RING( 6 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_SOLID_COLOR + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_P + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( buffer[i] ); + + OUT_RING( (x[i] << 16) | y[i] ); + OUT_RING( (1 << 16) | 1 ); + + ADVANCE_RING(); + } + } + + kfree( x ); + kfree( y ); + kfree( buffer ); + + return 0; +} + +static int r128_cce_dispatch_read_span( drm_device_t *dev, + drm_r128_depth_t *depth ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + int count, x, y; + u32 depth_bpp; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + switch ( dev_priv->depth_bpp ) { + case 16: + depth_bpp = R128_GMC_DST_16BPP; + break; + case 24: + depth_bpp = R128_GMC_DST_32BPP; + break; + case 32: + depth_bpp = R128_GMC_DST_32BPP; + break; + default: + return -EINVAL; + } + + count = depth->n; + if ( copy_from_user( &x, depth->x, sizeof(x) ) ) { + return -EFAULT; + } + if ( copy_from_user( &y, depth->y, sizeof(y) ) ) { + return -EFAULT; + } + + BEGIN_RING( 7 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_BITBLT_MULTI, 5 ) ); + OUT_RING( R128_GMC_SRC_PITCH_OFFSET_CNTL + | R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_NONE + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_S + | R128_DP_SRC_SOURCE_MEMORY + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( dev_priv->span_pitch_offset_c ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (0 << 16) | 0 ); + OUT_RING( (count << 16) | 1 ); + + ADVANCE_RING(); + + return 0; +} + +static int r128_cce_dispatch_read_pixels( drm_device_t *dev, + drm_r128_depth_t *depth ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + int count, *x, *y; + u32 depth_bpp; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + switch ( dev_priv->depth_bpp ) { + case 16: + depth_bpp = R128_GMC_DST_16BPP; + break; + case 24: + depth_bpp = R128_GMC_DST_32BPP; + break; + case 32: + depth_bpp = R128_GMC_DST_32BPP; + break; + default: + return -EINVAL; + } + + count = depth->n; + if ( count > dev_priv->depth_pitch ) { + count = dev_priv->depth_pitch; + } + + x = kmalloc( count * sizeof(*x), 0 ); + if ( x == NULL ) { + return -ENOMEM; + } + y = kmalloc( count * sizeof(*y), 0 ); + if ( y == NULL ) { + kfree( x ); + return -ENOMEM; + } + if ( copy_from_user( x, depth->x, count * sizeof(int) ) ) { + kfree( x ); + kfree( y ); + return -EFAULT; + } + if ( copy_from_user( y, depth->y, count * sizeof(int) ) ) { + kfree( x ); + kfree( y ); + return -EFAULT; + } + + for ( i = 0 ; i < count ; i++ ) { + BEGIN_RING( 7 ); + + OUT_RING( CCE_PACKET3( R128_CNTL_BITBLT_MULTI, 5 ) ); + OUT_RING( R128_GMC_SRC_PITCH_OFFSET_CNTL + | R128_GMC_DST_PITCH_OFFSET_CNTL + | R128_GMC_BRUSH_NONE + | depth_bpp + | R128_GMC_SRC_DATATYPE_COLOR + | R128_ROP3_S + | R128_DP_SRC_SOURCE_MEMORY + | R128_GMC_CLR_CMP_CNTL_DIS + | R128_GMC_WR_MSK_DIS ); + + OUT_RING( dev_priv->depth_pitch_offset_c ); + OUT_RING( dev_priv->span_pitch_offset_c ); + + OUT_RING( (x[i] << 16) | y[i] ); + OUT_RING( (i << 16) | 0 ); + OUT_RING( (1 << 16) | 1 ); + + ADVANCE_RING(); + } + + kfree( x ); + kfree( y ); + + return 0; +} + + +/* ================================================================ + * Polygon stipple + */ + +static void r128_cce_dispatch_stipple( drm_device_t *dev, u32 *stipple ) +{ + drm_r128_private_t *dev_priv = dev->dev_private; + int i; + RING_LOCALS; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + r128_update_ring_snapshot( dev_priv ); + + BEGIN_RING( 33 ); + + OUT_RING( CCE_PACKET0( R128_BRUSH_DATA0, 31 ) ); + for ( i = 0 ; i < 32 ; i++ ) { + OUT_RING( stipple[i] ); + } + + ADVANCE_RING(); +} + + +/* ================================================================ + * IOCTL functions + */ + +int r128_cce_clear( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_r128_clear_t clear; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "r128_cce_clear called without lock held\n" ); + return -EINVAL; + } + + if ( copy_from_user( &clear, (drm_r128_clear_t *) arg, + sizeof(clear) ) ) + return -EFAULT; + + if ( sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS ) + sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS; + + r128_cce_dispatch_clear( dev, clear.flags, + clear.x, clear.y, clear.w, clear.h, + clear.clear_color, clear.clear_depth, + clear.color_mask, clear.depth_mask ); + + /* Make sure we restore the 3D state next time. + */ + dev_priv->sarea_priv->dirty |= R128_UPLOAD_CONTEXT | R128_UPLOAD_MASKS; + + return 0; +} + +int r128_cce_swap( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv; + DRM_DEBUG( "%s\n", __FUNCTION__ ); + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "r128_cce_swap called without lock held\n" ); + return -EINVAL; + } + + if ( sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS ) + sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS; + + r128_cce_dispatch_swap( dev ); + + /* Make sure we restore the 3D state next time. + */ + dev_priv->sarea_priv->dirty |= R128_UPLOAD_CONTEXT | R128_UPLOAD_MASKS; + + return 0; +} + +int r128_cce_vertex( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_r128_buf_priv_t *buf_priv; + drm_r128_vertex_t vertex; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + if ( !dev_priv || dev_priv->is_pci ) { + DRM_ERROR( "%s called with a PCI card\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &vertex, (drm_r128_vertex_t *)arg, + sizeof(vertex) ) ) + return -EFAULT; + + DRM_DEBUG( "%s: pid=%d index=%d count=%d discard=%d\n", + __FUNCTION__, current->pid, + vertex.idx, vertex.count, vertex.discard ); + + if ( vertex.idx < 0 || vertex.idx >= dma->buf_count ) { + DRM_ERROR( "buffer index %d (of %d max)\n", + vertex.idx, dma->buf_count - 1 ); + return -EINVAL; + } + if ( vertex.prim < 0 || + vertex.prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2 ) { + DRM_ERROR( "buffer prim %d\n", vertex.prim ); + return -EINVAL; + } + + buf = dma->buflist[vertex.idx]; + buf_priv = buf->dev_private; + + if ( buf->pid != current->pid ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + current->pid, buf->pid ); + return -EINVAL; + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", vertex.idx ); + return -EINVAL; + } + + buf->used = vertex.count; + buf_priv->prim = vertex.prim; + buf_priv->discard = vertex.discard; + + r128_cce_dispatch_vertex( dev, buf ); + + return 0; +} + +int r128_cce_indices( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_private_t *dev_priv = dev->dev_private; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_r128_buf_priv_t *buf_priv; + drm_r128_indices_t elts; + int count; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + if ( !dev_priv || dev_priv->is_pci ) { + DRM_ERROR( "%s called with a PCI card\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &elts, (drm_r128_indices_t *)arg, + sizeof(elts) ) ) + return -EFAULT; + + DRM_DEBUG( "%s: pid=%d buf=%d s=%d e=%d d=%d\n", + __FUNCTION__, current->pid, + elts.idx, elts.start, elts.end, elts.discard ); + + if ( elts.idx < 0 || elts.idx >= dma->buf_count ) { + DRM_ERROR( "buffer index %d (of %d max)\n", + elts.idx, dma->buf_count - 1 ); + return -EINVAL; + } + if ( elts.prim < 0 || + elts.prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2 ) { + DRM_ERROR( "buffer prim %d\n", elts.prim ); + return -EINVAL; + } + + buf = dma->buflist[elts.idx]; + buf_priv = buf->dev_private; + + if ( buf->pid != current->pid ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + current->pid, buf->pid ); + return -EINVAL; + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", elts.idx ); + return -EINVAL; + } + + count = (elts.end - elts.start) / sizeof(u16); + elts.start -= R128_INDEX_PRIM_OFFSET; + + if ( elts.start & 0x7 ) { + DRM_ERROR( "misaligned buffer 0x%x\n", elts.start ); + return -EINVAL; + } + if ( elts.start < buf->used ) { + DRM_ERROR( "no header 0x%x - 0x%x\n", elts.start, buf->used ); + return -EINVAL; + } + + buf->used = elts.end; + buf_priv->prim = elts.prim; + buf_priv->discard = elts.discard; + + r128_cce_dispatch_indices( dev, buf, elts.start, elts.end, count ); + + return 0; +} + +int r128_cce_blit( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_device_dma_t *dma = dev->dma; + drm_r128_blit_t blit; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &blit, (drm_r128_blit_t *)arg, + sizeof(blit) ) ) + return -EFAULT; + + DRM_DEBUG( "%s: pid=%d index=%d\n", + __FUNCTION__, current->pid, blit.idx ); + + if ( blit.idx < 0 || blit.idx >= dma->buf_count ) { + DRM_ERROR( "buffer index %d (of %d max)\n", + blit.idx, dma->buf_count - 1 ); + return -EINVAL; + } + + return r128_cce_dispatch_blit( dev, &blit ); +} + +int r128_cce_depth( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_depth_t depth; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &depth, (drm_r128_depth_t *)arg, + sizeof(depth) ) ) + return -EFAULT; + + switch ( depth.func ) { + case R128_WRITE_SPAN: + return r128_cce_dispatch_write_span( dev, &depth ); + case R128_WRITE_PIXELS: + return r128_cce_dispatch_write_pixels( dev, &depth ); + case R128_READ_SPAN: + return r128_cce_dispatch_read_span( dev, &depth ); + case R128_READ_PIXELS: + return r128_cce_dispatch_read_pixels( dev, &depth ); + } + + return -EINVAL; +} + +int r128_cce_stipple( struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg ) +{ + drm_file_t *priv = filp->private_data; + drm_device_t *dev = priv->dev; + drm_r128_stipple_t stipple; + u32 mask[32]; + + if ( !_DRM_LOCK_IS_HELD( dev->lock.hw_lock->lock ) || + dev->lock.pid != current->pid ) { + DRM_ERROR( "%s called without lock held\n", __FUNCTION__ ); + return -EINVAL; + } + + if ( copy_from_user( &stipple, (drm_r128_stipple_t *)arg, + sizeof(stipple) ) ) + return -EFAULT; + + if ( copy_from_user( &mask, stipple.mask, + 32 * sizeof(u32) ) ) + return -EFAULT; + + r128_cce_dispatch_stipple( dev, mask ); + + return 0; +} -- cgit v1.2.3