diff options
| author | Keith Whitwell <keith@tungstengraphics.com> | 2002-07-15 20:21:18 +0000 |
|---|---|---|
| committer | Keith Whitwell <keith@tungstengraphics.com> | 2002-07-15 20:21:18 +0000 |
| commit | eef49a73b541d69be5da9b289deff36f5bb96ab0 (patch) | |
| tree | 0cc0fa5559d7bf87c25d277ef63d816d0090ccf4 /shared-core | |
| parent | deb296f395bb16ef7cf4bfe48f6e6c2cbacea0bc (diff) | |
Merge trunk into r200-branchr200-0-1-20020715
Diffstat (limited to 'shared-core')
| -rw-r--r-- | shared-core/drm.h | 30 | ||||
| -rw-r--r-- | shared-core/radeon_cp.c | 1565 | ||||
| -rw-r--r-- | shared-core/radeon_drm.h | 512 | ||||
| -rw-r--r-- | shared-core/radeon_drv.h | 67 | ||||
| -rw-r--r-- | shared-core/radeon_state.c | 2074 |
5 files changed, 4237 insertions, 11 deletions
diff --git a/shared-core/drm.h b/shared-core/drm.h index 6ab295c4..376568e0 100644 --- a/shared-core/drm.h +++ b/shared-core/drm.h @@ -38,10 +38,27 @@ #if defined(__linux__) #include <linux/config.h> #include <asm/ioctl.h> /* For _IO* macros */ -#define DRM_IOCTL_NR(n) _IOC_NR(n) -#elif defined(__FreeBSD__) +#define DRM_IOCTL_NR(n) _IOC_NR(n) +#define DRM_IOC_VOID _IOC_NONE +#define DRM_IOC_READ _IOC_READ +#define DRM_IOC_WRITE _IOC_WRITE +#define DRM_IOC_READWRITE _IOC_READ|_IOC_WRITE +#define DRM_IOC(dir, group, nr, size) _IOC(dir, group, nr, size) +#elif defined(__FreeBSD__) || defined(__NetBSD__) +#if defined(__FreeBSD__) && defined(XFree86Server) +/* Prevent name collision when including sys/ioccom.h */ +#undef ioctl #include <sys/ioccom.h> -#define DRM_IOCTL_NR(n) ((n) & 0xff) +#define ioctl(a,b,c) xf86ioctl(a,b,c) +#else +#include <sys/ioccom.h> +#endif /* __FreeBSD__ && xf86ioctl */ +#define DRM_IOCTL_NR(n) ((n) & 0xff) +#define DRM_IOC_VOID IOC_VOID +#define DRM_IOC_READ IOC_OUT +#define DRM_IOC_WRITE IOC_IN +#define DRM_IOC_READWRITE IOC_INOUT +#define DRM_IOC(dir, group, nr, size) _IOC(dir, group, nr, size) #endif #define XFREE86_VERSION(major,minor,patch,snap) \ @@ -367,10 +384,9 @@ typedef struct drm_scatter_gather { #define DRM_IOCTL_BASE 'd' #define DRM_IO(nr) _IO(DRM_IOCTL_BASE,nr) -#define DRM_IOR(nr,size) _IOR(DRM_IOCTL_BASE,nr,size) -#define DRM_IOW(nr,size) _IOW(DRM_IOCTL_BASE,nr,size) -#define DRM_IOWR(nr,size) _IOWR(DRM_IOCTL_BASE,nr,size) - +#define DRM_IOR(nr,type) _IOR(DRM_IOCTL_BASE,nr,type) +#define DRM_IOW(nr,type) _IOW(DRM_IOCTL_BASE,nr,type) +#define DRM_IOWR(nr,type) _IOWR(DRM_IOCTL_BASE,nr,type) #define DRM_IOCTL_VERSION DRM_IOWR(0x00, drm_version_t) #define DRM_IOCTL_GET_UNIQUE DRM_IOWR(0x01, drm_unique_t) diff --git a/shared-core/radeon_cp.c b/shared-core/radeon_cp.c new file mode 100644 index 00000000..2a9b6a54 --- /dev/null +++ b/shared-core/radeon_cp.c @@ -0,0 +1,1565 @@ +/* radeon_cp.c -- CP support for Radeon -*- linux-c -*- + * + * Copyright 2000 Precision Insight, Inc., Cedar Park, Texas. + * Copyright 2000 VA Linux Systems, Inc., Fremont, 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 <martin@valinux.com> + * Gareth Hughes <gareth@valinux.com> + */ + +#include "radeon.h" +#include "drmP.h" +#include "drm.h" +#include "radeon_drm.h" +#include "radeon_drv.h" + +#define RADEON_FIFO_DEBUG 0 + +#if defined(__alpha__) || defined(__powerpc__) +# define PCIGART_ENABLED +#else +# undef PCIGART_ENABLED +#endif + + +/* CP microcode (from ATI) */ +static u32 R200_cp_microcode[][2] = { + { 0x21007000, 0000000000 }, + { 0x20007000, 0000000000 }, + { 0x000000ab, 0x00000004 }, + { 0x000000af, 0x00000004 }, + { 0x66544a49, 0000000000 }, + { 0x49494174, 0000000000 }, + { 0x54517d83, 0000000000 }, + { 0x498d8b64, 0000000000 }, + { 0x49494949, 0000000000 }, + { 0x49da493c, 0000000000 }, + { 0x49989898, 0000000000 }, + { 0xd34949d5, 0000000000 }, + { 0x9dc90e11, 0000000000 }, + { 0xce9b9b9b, 0000000000 }, + { 0x000f0000, 0x00000016 }, + { 0x352e232c, 0000000000 }, + { 0x00000013, 0x00000004 }, + { 0x000f0000, 0x00000016 }, + { 0x352e272c, 0000000000 }, + { 0x000f0001, 0x00000016 }, + { 0x3239362f, 0000000000 }, + { 0x000077ef, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x00000020, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00061000, 0x00000002 }, + { 0x00000020, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00061000, 0x00000002 }, + { 0x00000020, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00000016, 0x00000004 }, + { 0x0003802a, 0x00000002 }, + { 0x040067e0, 0x00000002 }, + { 0x00000016, 0x00000004 }, + { 0x000077e0, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x000037e1, 0x00000002 }, + { 0x040067e1, 0x00000006 }, + { 0x000077e0, 0x00000002 }, + { 0x000077e1, 0x00000002 }, + { 0x000077e1, 0x00000006 }, + { 0xffffffff, 0000000000 }, + { 0x10000000, 0000000000 }, + { 0x0003802a, 0x00000002 }, + { 0x040067e0, 0x00000006 }, + { 0x00007675, 0x00000002 }, + { 0x00007676, 0x00000002 }, + { 0x00007677, 0x00000002 }, + { 0x00007678, 0x00000006 }, + { 0x0003802b, 0x00000002 }, + { 0x04002676, 0x00000002 }, + { 0x00007677, 0x00000002 }, + { 0x00007678, 0x00000006 }, + { 0x0000002e, 0x00000018 }, + { 0x0000002e, 0x00000018 }, + { 0000000000, 0x00000006 }, + { 0x0000002f, 0x00000018 }, + { 0x0000002f, 0x00000018 }, + { 0000000000, 0x00000006 }, + { 0x01605000, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x00098000, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x64c0603d, 0x00000004 }, + { 0x00080000, 0x00000016 }, + { 0000000000, 0000000000 }, + { 0x0400251d, 0x00000002 }, + { 0x00007580, 0x00000002 }, + { 0x00067581, 0x00000002 }, + { 0x04002580, 0x00000002 }, + { 0x00067581, 0x00000002 }, + { 0x00000046, 0x00000004 }, + { 0x00005000, 0000000000 }, + { 0x00061000, 0x00000002 }, + { 0x0000750e, 0x00000002 }, + { 0x00019000, 0x00000002 }, + { 0x00011055, 0x00000014 }, + { 0x00000055, 0x00000012 }, + { 0x0400250f, 0x00000002 }, + { 0x0000504a, 0x00000004 }, + { 0x00007565, 0x00000002 }, + { 0x00007566, 0x00000002 }, + { 0x00000051, 0x00000004 }, + { 0x01e655b4, 0x00000002 }, + { 0x4401b0dc, 0x00000002 }, + { 0x01c110dc, 0x00000002 }, + { 0x2666705d, 0x00000018 }, + { 0x040c2565, 0x00000002 }, + { 0x0000005d, 0x00000018 }, + { 0x04002564, 0x00000002 }, + { 0x00007566, 0x00000002 }, + { 0x00000054, 0x00000004 }, + { 0x00401060, 0x00000008 }, + { 0x00101000, 0x00000002 }, + { 0x000d80ff, 0x00000002 }, + { 0x00800063, 0x00000008 }, + { 0x000f9000, 0x00000002 }, + { 0x000e00ff, 0x00000002 }, + { 0000000000, 0x00000006 }, + { 0x00000080, 0x00000018 }, + { 0x00000054, 0x00000004 }, + { 0x00007576, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x00009000, 0x00000002 }, + { 0x00041000, 0x00000002 }, + { 0x0c00350e, 0x00000002 }, + { 0x00049000, 0x00000002 }, + { 0x00051000, 0x00000002 }, + { 0x01e785f8, 0x00000002 }, + { 0x00200000, 0x00000002 }, + { 0x00600073, 0x0000000c }, + { 0x00007563, 0x00000002 }, + { 0x006075f0, 0x00000021 }, + { 0x20007068, 0x00000004 }, + { 0x00005068, 0x00000004 }, + { 0x00007576, 0x00000002 }, + { 0x00007577, 0x00000002 }, + { 0x0000750e, 0x00000002 }, + { 0x0000750f, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00600076, 0x0000000c }, + { 0x006075f0, 0x00000021 }, + { 0x000075f8, 0x00000002 }, + { 0x00000076, 0x00000004 }, + { 0x000a750e, 0x00000002 }, + { 0x0020750f, 0x00000002 }, + { 0x00600079, 0x00000004 }, + { 0x00007570, 0x00000002 }, + { 0x00007571, 0x00000002 }, + { 0x00007572, 0x00000006 }, + { 0x00005000, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00007568, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x00000084, 0x0000000c }, + { 0x00058000, 0x00000002 }, + { 0x0c607562, 0x00000002 }, + { 0x00000086, 0x00000004 }, + { 0x00600085, 0x00000004 }, + { 0x400070dd, 0000000000 }, + { 0x000380dd, 0x00000002 }, + { 0x00000093, 0x0000001c }, + { 0x00065095, 0x00000018 }, + { 0x040025bb, 0x00000002 }, + { 0x00061096, 0x00000018 }, + { 0x040075bc, 0000000000 }, + { 0x000075bb, 0x00000002 }, + { 0x000075bc, 0000000000 }, + { 0x00090000, 0x00000006 }, + { 0x00090000, 0x00000002 }, + { 0x000d8002, 0x00000006 }, + { 0x00005000, 0x00000002 }, + { 0x00007821, 0x00000002 }, + { 0x00007800, 0000000000 }, + { 0x00007821, 0x00000002 }, + { 0x00007800, 0000000000 }, + { 0x01665000, 0x00000002 }, + { 0x000a0000, 0x00000002 }, + { 0x000671cc, 0x00000002 }, + { 0x0286f1cd, 0x00000002 }, + { 0x000000a3, 0x00000010 }, + { 0x21007000, 0000000000 }, + { 0x000000aa, 0x0000001c }, + { 0x00065000, 0x00000002 }, + { 0x000a0000, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x000b0000, 0x00000002 }, + { 0x38067000, 0x00000002 }, + { 0x000a00a6, 0x00000004 }, + { 0x20007000, 0000000000 }, + { 0x01200000, 0x00000002 }, + { 0x20077000, 0x00000002 }, + { 0x01200000, 0x00000002 }, + { 0x20007000, 0000000000 }, + { 0x00061000, 0x00000002 }, + { 0x0120751b, 0x00000002 }, + { 0x8040750a, 0x00000002 }, + { 0x8040750b, 0x00000002 }, + { 0x00110000, 0x00000002 }, + { 0x000380dd, 0x00000002 }, + { 0x000000bd, 0x0000001c }, + { 0x00061096, 0x00000018 }, + { 0x844075bd, 0x00000002 }, + { 0x00061095, 0x00000018 }, + { 0x840075bb, 0x00000002 }, + { 0x00061096, 0x00000018 }, + { 0x844075bc, 0x00000002 }, + { 0x000000c0, 0x00000004 }, + { 0x804075bd, 0x00000002 }, + { 0x800075bb, 0x00000002 }, + { 0x804075bc, 0x00000002 }, + { 0x00108000, 0x00000002 }, + { 0x01400000, 0x00000002 }, + { 0x006000c4, 0x0000000c }, + { 0x20c07000, 0x00000020 }, + { 0x000000c6, 0x00000012 }, + { 0x00800000, 0x00000006 }, + { 0x0080751d, 0x00000006 }, + { 0x000025bb, 0x00000002 }, + { 0x000040c0, 0x00000004 }, + { 0x0000775c, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00661000, 0x00000002 }, + { 0x0460275d, 0x00000020 }, + { 0x00004000, 0000000000 }, + { 0x00007999, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00661000, 0x00000002 }, + { 0x0460299b, 0x00000020 }, + { 0x00004000, 0000000000 }, + { 0x01e00830, 0x00000002 }, + { 0x21007000, 0000000000 }, + { 0x00005000, 0x00000002 }, + { 0x00038042, 0x00000002 }, + { 0x040025e0, 0x00000002 }, + { 0x000075e1, 0000000000 }, + { 0x00000001, 0000000000 }, + { 0x000380d9, 0x00000002 }, + { 0x04007394, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, +}; + + +static u32 radeon_cp_microcode[][2] = { + { 0x21007000, 0000000000 }, + { 0x20007000, 0000000000 }, + { 0x000000b4, 0x00000004 }, + { 0x000000b8, 0x00000004 }, + { 0x6f5b4d4c, 0000000000 }, + { 0x4c4c427f, 0000000000 }, + { 0x5b568a92, 0000000000 }, + { 0x4ca09c6d, 0000000000 }, + { 0xad4c4c4c, 0000000000 }, + { 0x4ce1af3d, 0000000000 }, + { 0xd8afafaf, 0000000000 }, + { 0xd64c4cdc, 0000000000 }, + { 0x4cd10d10, 0000000000 }, + { 0x000f0000, 0x00000016 }, + { 0x362f242d, 0000000000 }, + { 0x00000012, 0x00000004 }, + { 0x000f0000, 0x00000016 }, + { 0x362f282d, 0000000000 }, + { 0x000380e7, 0x00000002 }, + { 0x04002c97, 0x00000002 }, + { 0x000f0001, 0x00000016 }, + { 0x333a3730, 0000000000 }, + { 0x000077ef, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x00000021, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00061000, 0x00000002 }, + { 0x00000021, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00061000, 0x00000002 }, + { 0x00000021, 0x0000001a }, + { 0x00004000, 0x0000001e }, + { 0x00000017, 0x00000004 }, + { 0x0003802b, 0x00000002 }, + { 0x040067e0, 0x00000002 }, + { 0x00000017, 0x00000004 }, + { 0x000077e0, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x000037e1, 0x00000002 }, + { 0x040067e1, 0x00000006 }, + { 0x000077e0, 0x00000002 }, + { 0x000077e1, 0x00000002 }, + { 0x000077e1, 0x00000006 }, + { 0xffffffff, 0000000000 }, + { 0x10000000, 0000000000 }, + { 0x0003802b, 0x00000002 }, + { 0x040067e0, 0x00000006 }, + { 0x00007675, 0x00000002 }, + { 0x00007676, 0x00000002 }, + { 0x00007677, 0x00000002 }, + { 0x00007678, 0x00000006 }, + { 0x0003802c, 0x00000002 }, + { 0x04002676, 0x00000002 }, + { 0x00007677, 0x00000002 }, + { 0x00007678, 0x00000006 }, + { 0x0000002f, 0x00000018 }, + { 0x0000002f, 0x00000018 }, + { 0000000000, 0x00000006 }, + { 0x00000030, 0x00000018 }, + { 0x00000030, 0x00000018 }, + { 0000000000, 0x00000006 }, + { 0x01605000, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x00098000, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x64c0603e, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00080000, 0x00000016 }, + { 0000000000, 0000000000 }, + { 0x0400251d, 0x00000002 }, + { 0x00007580, 0x00000002 }, + { 0x00067581, 0x00000002 }, + { 0x04002580, 0x00000002 }, + { 0x00067581, 0x00000002 }, + { 0x00000049, 0x00000004 }, + { 0x00005000, 0000000000 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x0000750e, 0x00000002 }, + { 0x00019000, 0x00000002 }, + { 0x00011055, 0x00000014 }, + { 0x00000055, 0x00000012 }, + { 0x0400250f, 0x00000002 }, + { 0x0000504f, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00007565, 0x00000002 }, + { 0x00007566, 0x00000002 }, + { 0x00000058, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x01e655b4, 0x00000002 }, + { 0x4401b0e4, 0x00000002 }, + { 0x01c110e4, 0x00000002 }, + { 0x26667066, 0x00000018 }, + { 0x040c2565, 0x00000002 }, + { 0x00000066, 0x00000018 }, + { 0x04002564, 0x00000002 }, + { 0x00007566, 0x00000002 }, + { 0x0000005d, 0x00000004 }, + { 0x00401069, 0x00000008 }, + { 0x00101000, 0x00000002 }, + { 0x000d80ff, 0x00000002 }, + { 0x0080006c, 0x00000008 }, + { 0x000f9000, 0x00000002 }, + { 0x000e00ff, 0x00000002 }, + { 0000000000, 0x00000006 }, + { 0x0000008f, 0x00000018 }, + { 0x0000005b, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00007576, 0x00000002 }, + { 0x00065000, 0x00000002 }, + { 0x00009000, 0x00000002 }, + { 0x00041000, 0x00000002 }, + { 0x0c00350e, 0x00000002 }, + { 0x00049000, 0x00000002 }, + { 0x00051000, 0x00000002 }, + { 0x01e785f8, 0x00000002 }, + { 0x00200000, 0x00000002 }, + { 0x0060007e, 0x0000000c }, + { 0x00007563, 0x00000002 }, + { 0x006075f0, 0x00000021 }, + { 0x20007073, 0x00000004 }, + { 0x00005073, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00007576, 0x00000002 }, + { 0x00007577, 0x00000002 }, + { 0x0000750e, 0x00000002 }, + { 0x0000750f, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00600083, 0x0000000c }, + { 0x006075f0, 0x00000021 }, + { 0x000075f8, 0x00000002 }, + { 0x00000083, 0x00000004 }, + { 0x000a750e, 0x00000002 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x0020750f, 0x00000002 }, + { 0x00600086, 0x00000004 }, + { 0x00007570, 0x00000002 }, + { 0x00007571, 0x00000002 }, + { 0x00007572, 0x00000006 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00005000, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00007568, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x00000095, 0x0000000c }, + { 0x00058000, 0x00000002 }, + { 0x0c607562, 0x00000002 }, + { 0x00000097, 0x00000004 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x00600096, 0x00000004 }, + { 0x400070e5, 0000000000 }, + { 0x000380e6, 0x00000002 }, + { 0x040025c5, 0x00000002 }, + { 0x000380e5, 0x00000002 }, + { 0x000000a8, 0x0000001c }, + { 0x000650aa, 0x00000018 }, + { 0x040025bb, 0x00000002 }, + { 0x000610ab, 0x00000018 }, + { 0x040075bc, 0000000000 }, + { 0x000075bb, 0x00000002 }, + { 0x000075bc, 0000000000 }, + { 0x00090000, 0x00000006 }, + { 0x00090000, 0x00000002 }, + { 0x000d8002, 0x00000006 }, + { 0x00007832, 0x00000002 }, + { 0x00005000, 0x00000002 }, + { 0x000380e7, 0x00000002 }, + { 0x04002c97, 0x00000002 }, + { 0x00007820, 0x00000002 }, + { 0x00007821, 0x00000002 }, + { 0x00007800, 0000000000 }, + { 0x01200000, 0x00000002 }, + { 0x20077000, 0x00000002 }, + { 0x01200000, 0x00000002 }, + { 0x20007000, 0x00000002 }, + { 0x00061000, 0x00000002 }, + { 0x0120751b, 0x00000002 }, + { 0x8040750a, 0x00000002 }, + { 0x8040750b, 0x00000002 }, + { 0x00110000, 0x00000002 }, + { 0x000380e5, 0x00000002 }, + { 0x000000c6, 0x0000001c }, + { 0x000610ab, 0x00000018 }, + { 0x844075bd, 0x00000002 }, + { 0x000610aa, 0x00000018 }, + { 0x840075bb, 0x00000002 }, + { 0x000610ab, 0x00000018 }, + { 0x844075bc, 0x00000002 }, + { 0x000000c9, 0x00000004 }, + { 0x804075bd, 0x00000002 }, + { 0x800075bb, 0x00000002 }, + { 0x804075bc, 0x00000002 }, + { 0x00108000, 0x00000002 }, + { 0x01400000, 0x00000002 }, + { 0x006000cd, 0x0000000c }, + { 0x20c07000, 0x00000020 }, + { 0x000000cf, 0x00000012 }, + { 0x00800000, 0x00000006 }, + { 0x0080751d, 0x00000006 }, + { 0000000000, 0000000000 }, + { 0x0000775c, 0x00000002 }, + { 0x00a05000, 0x00000002 }, + { 0x00661000, 0x00000002 }, + { 0x0460275d, 0x00000020 }, + { 0x00004000, 0000000000 }, + { 0x01e00830, 0x00000002 }, + { 0x21007000, 0000000000 }, + { 0x6464614d, 0000000000 }, + { 0x69687420, 0000000000 }, + { 0x00000073, 0000000000 }, + { 0000000000, 0000000000 }, + { 0x00005000, 0x00000002 }, + { 0x000380d0, 0x00000002 }, + { 0x040025e0, 0x00000002 }, + { 0x000075e1, 0000000000 }, + { 0x00000001, 0000000000 }, + { 0x000380e0, 0x00000002 }, + { 0x04002394, 0x00000002 }, + { 0x00005000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0x00000008, 0000000000 }, + { 0x00000004, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, + { 0000000000, 0000000000 }, +}; + + +int RADEON_READ_PLL(drm_device_t *dev, int addr) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + + RADEON_WRITE8(RADEON_CLOCK_CNTL_INDEX, addr & 0x1f); + return RADEON_READ(RADEON_CLOCK_CNTL_DATA); +} + +#if RADEON_FIFO_DEBUG +static void radeon_status( drm_radeon_private_t *dev_priv ) +{ + printk( "%s:\n", __FUNCTION__ ); + printk( "RBBM_STATUS = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_RBBM_STATUS ) ); + printk( "CP_RB_RTPR = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_CP_RB_RPTR ) ); + printk( "CP_RB_WTPR = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_CP_RB_WPTR ) ); + printk( "AIC_CNTL = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_AIC_CNTL ) ); + printk( "AIC_STAT = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_AIC_STAT ) ); + printk( "AIC_PT_BASE = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_AIC_PT_BASE ) ); + printk( "TLB_ADDR = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_AIC_TLB_ADDR ) ); + printk( "TLB_DATA = 0x%08x\n", + (unsigned int)RADEON_READ( RADEON_AIC_TLB_DATA ) ); +} +#endif + + +/* ================================================================ + * Engine, FIFO control + */ + +static int radeon_do_pixcache_flush( drm_radeon_private_t *dev_priv ) +{ + u32 tmp; + int i; + + tmp = RADEON_READ( RADEON_RB2D_DSTCACHE_CTLSTAT ); + tmp |= RADEON_RB2D_DC_FLUSH_ALL; + RADEON_WRITE( RADEON_RB2D_DSTCACHE_CTLSTAT, tmp ); + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + if ( !(RADEON_READ( RADEON_RB2D_DSTCACHE_CTLSTAT ) + & RADEON_RB2D_DC_BUSY) ) { + return 0; + } + DRM_UDELAY( 1 ); + } + +#if RADEON_FIFO_DEBUG + DRM_ERROR( "failed!\n" ); + radeon_status( dev_priv ); +#endif + return DRM_ERR(EBUSY); +} + +static int radeon_do_wait_for_fifo( drm_radeon_private_t *dev_priv, + int entries ) +{ + int i; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + int slots = ( RADEON_READ( RADEON_RBBM_STATUS ) + & RADEON_RBBM_FIFOCNT_MASK ); + if ( slots >= entries ) return 0; + DRM_UDELAY( 1 ); + } + +#if RADEON_FIFO_DEBUG + DRM_ERROR( "failed!\n" ); + radeon_status( dev_priv ); +#endif + return DRM_ERR(EBUSY); +} + +static int radeon_do_wait_for_idle( drm_radeon_private_t *dev_priv ) +{ + int i, ret; + + ret = radeon_do_wait_for_fifo( dev_priv, 64 ); + if ( ret ) return ret; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + if ( !(RADEON_READ( RADEON_RBBM_STATUS ) + & RADEON_RBBM_ACTIVE) ) { + radeon_do_pixcache_flush( dev_priv ); + return 0; + } + DRM_UDELAY( 1 ); + } + +#if RADEON_FIFO_DEBUG + DRM_ERROR( "failed!\n" ); + radeon_status( dev_priv ); +#endif + return DRM_ERR(EBUSY); +} + + +/* ================================================================ + * CP control, initialization + */ + +/* Load the microcode for the CP */ +static void radeon_cp_load_microcode( drm_radeon_private_t *dev_priv ) +{ + int i; + DRM_DEBUG( "\n" ); + + radeon_do_wait_for_idle( dev_priv ); + + RADEON_WRITE( RADEON_CP_ME_RAM_ADDR, 0 ); + + if (dev_priv->is_r200) + { + DRM_INFO("Loading R200 Microcode\n"); + for ( i = 0 ; i < 256 ; i++ ) + { + RADEON_WRITE( RADEON_CP_ME_RAM_DATAH, + R200_cp_microcode[i][1] ); + RADEON_WRITE( RADEON_CP_ME_RAM_DATAL, + R200_cp_microcode[i][0] ); + } + } + else + { + for ( i = 0 ; i < 256 ; i++ ) { + RADEON_WRITE( RADEON_CP_ME_RAM_DATAH, + radeon_cp_microcode[i][1] ); + RADEON_WRITE( RADEON_CP_ME_RAM_DATAL, + radeon_cp_microcode[i][0] ); + } + } +} + +/* Flush any pending commands to the CP. This should only be used just + * prior to a wait for idle, as it informs the engine that the command + * stream is ending. + */ +static void radeon_do_cp_flush( drm_radeon_private_t *dev_priv ) +{ + DRM_DEBUG( "\n" ); +#if 0 + u32 tmp; + + tmp = RADEON_READ( RADEON_CP_RB_WPTR ) | (1 << 31); + RADEON_WRITE( RADEON_CP_RB_WPTR, tmp ); +#endif +} + +/* Wait for the CP to go idle. + */ +int radeon_do_cp_idle( drm_radeon_private_t *dev_priv ) +{ + RING_LOCALS; + DRM_DEBUG( "\n" ); + + BEGIN_RING( 6 ); + + RADEON_PURGE_CACHE(); + RADEON_PURGE_ZCACHE(); + RADEON_WAIT_UNTIL_IDLE(); + + ADVANCE_RING(); + COMMIT_RING(); + + return radeon_do_wait_for_idle( dev_priv ); +} + +/* Start the Command Processor. + */ +static void radeon_do_cp_start( drm_radeon_private_t *dev_priv ) +{ + RING_LOCALS; + DRM_DEBUG( "\n" ); + + radeon_do_wait_for_idle( dev_priv ); + + RADEON_WRITE( RADEON_CP_CSQ_CNTL, dev_priv->cp_mode ); + + dev_priv->cp_running = 1; + + BEGIN_RING( 6 ); + + RADEON_PURGE_CACHE(); + RADEON_PURGE_ZCACHE(); + RADEON_WAIT_UNTIL_IDLE(); + + ADVANCE_RING(); + COMMIT_RING(); +} + +/* Reset the Command Processor. This will not flush any pending + * commands, so you must wait for the CP command stream to complete + * before calling this routine. + */ +static void radeon_do_cp_reset( drm_radeon_private_t *dev_priv ) +{ + u32 cur_read_ptr; + DRM_DEBUG( "\n" ); + + cur_read_ptr = RADEON_READ( RADEON_CP_RB_RPTR ); + RADEON_WRITE( RADEON_CP_RB_WPTR, cur_read_ptr ); + *dev_priv->ring.head = cur_read_ptr; + dev_priv->ring.tail = cur_read_ptr; +} + +/* Stop the Command Processor. This will not flush any pending + * commands, so you must flush the command stream and wait for the CP + * to go idle before calling this routine. + */ +static void radeon_do_cp_stop( drm_radeon_private_t *dev_priv ) +{ + DRM_DEBUG( "\n" ); + + RADEON_WRITE( RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIDIS_INDDIS ); + + dev_priv->cp_running = 0; +} + +/* Reset the engine. This will stop the CP if it is running. + */ +static int radeon_do_engine_reset( drm_device_t *dev ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + u32 clock_cntl_index, mclk_cntl, rbbm_soft_reset; + DRM_DEBUG( "\n" ); + + radeon_do_pixcache_flush( dev_priv ); + + clock_cntl_index = RADEON_READ( RADEON_CLOCK_CNTL_INDEX ); + mclk_cntl = RADEON_READ_PLL( dev, RADEON_MCLK_CNTL ); + + RADEON_WRITE_PLL( RADEON_MCLK_CNTL, ( mclk_cntl | + RADEON_FORCEON_MCLKA | + RADEON_FORCEON_MCLKB | + RADEON_FORCEON_YCLKA | + RADEON_FORCEON_YCLKB | + RADEON_FORCEON_MC | + RADEON_FORCEON_AIC ) ); + + rbbm_soft_reset = RADEON_READ( RADEON_RBBM_SOFT_RESET ); + + RADEON_WRITE( RADEON_RBBM_SOFT_RESET, ( rbbm_soft_reset | + RADEON_SOFT_RESET_CP | + RADEON_SOFT_RESET_HI | + RADEON_SOFT_RESET_SE | + RADEON_SOFT_RESET_RE | + RADEON_SOFT_RESET_PP | + RADEON_SOFT_RESET_E2 | + RADEON_SOFT_RESET_RB ) ); + RADEON_READ( RADEON_RBBM_SOFT_RESET ); + RADEON_WRITE( RADEON_RBBM_SOFT_RESET, ( rbbm_soft_reset & + ~( RADEON_SOFT_RESET_CP | + RADEON_SOFT_RESET_HI | + RADEON_SOFT_RESET_SE | + RADEON_SOFT_RESET_RE | + RADEON_SOFT_RESET_PP | + RADEON_SOFT_RESET_E2 | + RADEON_SOFT_RESET_RB ) ) ); + RADEON_READ( RADEON_RBBM_SOFT_RESET ); + + + RADEON_WRITE_PLL( RADEON_MCLK_CNTL, mclk_cntl ); + RADEON_WRITE( RADEON_CLOCK_CNTL_INDEX, clock_cntl_index ); + RADEON_WRITE( RADEON_RBBM_SOFT_RESET, rbbm_soft_reset ); + + /* Reset the CP ring */ + radeon_do_cp_reset( dev_priv ); + + /* The CP is no longer running after an engine reset */ + dev_priv->cp_running = 0; + + /* Reset any pending vertex, indirect buffers */ + radeon_freelist_reset( dev ); + + return 0; +} + +static void radeon_cp_init_ring_buffer( drm_device_t *dev, + drm_radeon_private_t *dev_priv ) +{ + u32 ring_start, cur_read_ptr; + u32 tmp; + + /* Initialize the memory controller */ + RADEON_WRITE( RADEON_MC_FB_LOCATION, + (dev_priv->agp_vm_start - 1) & 0xffff0000 ); + + if ( !dev_priv->is_pci ) { + RADEON_WRITE( RADEON_MC_AGP_LOCATION, + (((dev_priv->agp_vm_start - 1 + + dev_priv->agp_size) & 0xffff0000) | + (dev_priv->agp_vm_start >> 16)) ); + } + +#if __REALLY_HAVE_AGP + if ( !dev_priv->is_pci ) + ring_start = (dev_priv->cp_ring->offset + - dev->agp->base + + dev_priv->agp_vm_start); + else +#endif + ring_start = (dev_priv->cp_ring->offset + - dev->sg->handle + + dev_priv->agp_vm_start); + + RADEON_WRITE( RADEON_CP_RB_BASE, ring_start ); + + /* Set the write pointer delay */ + RADEON_WRITE( RADEON_CP_RB_WPTR_DELAY, 0 ); + + /* Initialize the ring buffer's read and write pointers */ + cur_read_ptr = RADEON_READ( RADEON_CP_RB_RPTR ); + RADEON_WRITE( RADEON_CP_RB_WPTR, cur_read_ptr ); + *dev_priv->ring.head = cur_read_ptr; + dev_priv->ring.tail = cur_read_ptr; + + if ( !dev_priv->is_pci ) { + RADEON_WRITE( RADEON_CP_RB_RPTR_ADDR, + dev_priv->ring_rptr->offset ); + } else { + drm_sg_mem_t *entry = dev->sg; + unsigned long tmp_ofs, page_ofs; + + tmp_ofs = dev_priv->ring_rptr->offset - dev->sg->handle; + page_ofs = tmp_ofs >> PAGE_SHIFT; + + RADEON_WRITE( RADEON_CP_RB_RPTR_ADDR, + entry->busaddr[page_ofs]); + DRM_DEBUG( "ring rptr: offset=0x%08x handle=0x%08lx\n", + entry->busaddr[page_ofs], + entry->handle + tmp_ofs ); + } + + /* Initialize the scratch register pointer. This will cause + * the scratch register values to be written out to memory + * whenever they are updated. + * + * We simply put this behind the ring read pointer, this works + * with PCI GART as well as (whatever kind of) AGP GART + */ + RADEON_WRITE( RADEON_SCRATCH_ADDR, RADEON_READ( RADEON_CP_RB_RPTR_ADDR ) + + RADEON_SCRATCH_REG_OFFSET ); + + dev_priv->scratch = ((__volatile__ u32 *) + dev_priv->ring.head + + (RADEON_SCRATCH_REG_OFFSET / sizeof(u32))); + + RADEON_WRITE( RADEON_SCRATCH_UMSK, 0x7 ); + + dev_priv->sarea_priv->last_frame = dev_priv->scratch[0] = 0; + RADEON_WRITE( RADEON_LAST_FRAME_REG, + dev_priv->sarea_priv->last_frame ); + + dev_priv->sarea_priv->last_dispatch = dev_priv->scratch[1] = 0; + RADEON_WRITE( RADEON_LAST_DISPATCH_REG, + dev_priv->sarea_priv->last_dispatch ); + + dev_priv->sarea_priv->last_clear = dev_priv->scratch[2] = 0; + RADEON_WRITE( RADEON_LAST_CLEAR_REG, + dev_priv->sarea_priv->last_clear ); + + /* Set ring buffer size */ +#ifdef __BIG_ENDIAN + RADEON_WRITE( RADEON_CP_RB_CNTL, dev_priv->ring.size_l2qw | RADEON_BUF_SWAP_32BIT ); +#else + RADEON_WRITE( RADEON_CP_RB_CNTL, dev_priv->ring.size_l2qw ); +#endif + + radeon_do_wait_for_idle( dev_priv ); + + /* Turn on bus mastering */ + tmp = RADEON_READ( RADEON_BUS_CNTL ) & ~RADEON_BUS_MASTER_DIS; + RADEON_WRITE( RADEON_BUS_CNTL, tmp ); + + /* Sync everything up */ + RADEON_WRITE( RADEON_ISYNC_CNTL, + (RADEON_ISYNC_ANY2D_IDLE3D | + RADEON_ISYNC_ANY3D_IDLE2D | + RADEON_ISYNC_WAIT_IDLEGUI | + RADEON_ISYNC_CPSCRATCH_IDLEGUI) ); +} + +static int radeon_do_init_cp( drm_device_t *dev, drm_radeon_init_t *init ) +{ + drm_radeon_private_t *dev_priv; + u32 tmp; + DRM_DEBUG( "\n" ); + + dev_priv = DRM(alloc)( sizeof(drm_radeon_private_t), DRM_MEM_DRIVER ); + if ( dev_priv == NULL ) + return DRM_ERR(ENOMEM); + + memset( dev_priv, 0, sizeof(drm_radeon_private_t) ); + + dev_priv->is_pci = init->is_pci; + +#if !defined(PCIGART_ENABLED) + /* PCI support is not 100% working, so we disable it here. + */ + if ( dev_priv->is_pci ) { + DRM_ERROR( "PCI GART not yet supported for Radeon!\n" ); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } +#endif + + if ( dev_priv->is_pci && !dev->sg ) { + DRM_ERROR( "PCI GART memory not allocated!\n" ); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + + dev_priv->usec_timeout = init->usec_timeout; + if ( dev_priv->usec_timeout < 1 || + dev_priv->usec_timeout > RADEON_MAX_USEC_TIMEOUT ) { + DRM_DEBUG( "TIMEOUT problem!\n" ); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + + dev_priv->is_r200 = (init->func == RADEON_INIT_R200_CP); + dev_priv->cp_mode = init->cp_mode; + + /* 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->cp_mode != RADEON_CSQ_PRIBM_INDDIS ) && + ( init->cp_mode != RADEON_CSQ_PRIBM_INDBM ) ) { + DRM_DEBUG( "BAD cp_mode (%x)!\n", init->cp_mode ); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + + switch ( init->fb_bpp ) { + case 16: + dev_priv->color_fmt = RADEON_COLOR_FORMAT_RGB565; + break; + case 32: + default: + dev_priv->color_fmt = RADEON_COLOR_FORMAT_ARGB8888; + break; + } + dev_priv->front_offset = init->front_offset; + dev_priv->front_pitch = init->front_pitch; + dev_priv->back_offset = init->back_offset; + dev_priv->back_pitch = init->back_pitch; + + switch ( init->depth_bpp ) { + case 16: + dev_priv->depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; + break; + case 32: + default: + dev_priv->depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; + break; + } + dev_priv->depth_offset = init->depth_offset; + dev_priv->depth_pitch = init->depth_pitch; + + dev_priv->front_pitch_offset = (((dev_priv->front_pitch/64) << 22) | + (dev_priv->front_offset >> 10)); + dev_priv->back_pitch_offset = (((dev_priv->back_pitch/64) << 22) | + (dev_priv->back_offset >> 10)); + dev_priv->depth_pitch_offset = (((dev_priv->depth_pitch/64) << 22) | + (dev_priv->depth_offset >> 10)); + + /* Hardware state for depth clears. Remove this if/when we no + * longer clear the depth buffer with a 3D rectangle. Hard-code + * all values to prevent unwanted 3D state from slipping through + * and screwing with the clear operation. + */ + dev_priv->depth_clear.rb3d_cntl = (RADEON_PLANE_MASK_ENABLE | + (dev_priv->color_fmt << 10) | + (1<<15)); + + dev_priv->depth_clear.rb3d_zstencilcntl = + (dev_priv->depth_fmt | + RADEON_Z_TEST_ALWAYS | + RADEON_STENCIL_TEST_ALWAYS | + RADEON_STENCIL_S_FAIL_REPLACE | + RADEON_STENCIL_ZPASS_REPLACE | + RADEON_STENCIL_ZFAIL_REPLACE | + RADEON_Z_WRITE_ENABLE); + + dev_priv->depth_clear.se_cntl = (RADEON_FFACE_CULL_CW | + RADEON_BFACE_SOLID | + RADEON_FFACE_SOLID | + RADEON_FLAT_SHADE_VTX_LAST | + RADEON_DIFFUSE_SHADE_FLAT | + RADEON_ALPHA_SHADE_FLAT | + RADEON_SPECULAR_SHADE_FLAT | + RADEON_FOG_SHADE_FLAT | + RADEON_VTX_PIX_CENTER_OGL | + RADEON_ROUND_MODE_TRUNC | + RADEON_ROUND_PREC_8TH_PIX); + + DRM_GETSAREA(); + + if(!dev_priv->sarea) { + DRM_ERROR("could not find sarea!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + + DRM_FIND_MAP( dev_priv->fb, init->fb_offset ); + if(!dev_priv->fb) { + DRM_ERROR("could not find framebuffer!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + DRM_FIND_MAP( dev_priv->mmio, init->mmio_offset ); + if(!dev_priv->mmio) { + DRM_ERROR("could not find mmio region!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + DRM_FIND_MAP( dev_priv->cp_ring, init->ring_offset ); + if(!dev_priv->cp_ring) { + DRM_ERROR("could not find cp ring region!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + DRM_FIND_MAP( dev_priv->ring_rptr, init->ring_rptr_offset ); + if(!dev_priv->ring_rptr) { + DRM_ERROR("could not find ring read pointer!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + DRM_FIND_MAP( dev_priv->buffers, init->buffers_offset ); + if(!dev_priv->buffers) { + DRM_ERROR("could not find dma buffer region!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + + if ( !dev_priv->is_pci ) { + DRM_FIND_MAP( dev_priv->agp_textures, + init->agp_textures_offset ); + if(!dev_priv->agp_textures) { + DRM_ERROR("could not find agp texture region!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + } + + dev_priv->sarea_priv = + (drm_radeon_sarea_t *)((u8 *)dev_priv->sarea->handle + + init->sarea_priv_offset); + + if ( !dev_priv->is_pci ) { + DRM_IOREMAP( dev_priv->cp_ring ); + DRM_IOREMAP( dev_priv->ring_rptr ); + DRM_IOREMAP( dev_priv->buffers ); + if(!dev_priv->cp_ring->handle || + !dev_priv->ring_rptr->handle || + !dev_priv->buffers->handle) { + DRM_ERROR("could not find ioremap agp regions!\n"); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(EINVAL); + } + } else { + dev_priv->cp_ring->handle = + (void *)dev_priv->cp_ring->offset; + dev_priv->ring_rptr->handle = + (void *)dev_priv->ring_rptr->offset; + dev_priv->buffers->handle = (void *)dev_priv->buffers->offset; + + DRM_DEBUG( "dev_priv->cp_ring->handle %p\n", + dev_priv->cp_ring->handle ); + DRM_DEBUG( "dev_priv->ring_rptr->handle %p\n", + dev_priv->ring_rptr->handle ); + DRM_DEBUG( "dev_priv->buffers->handle %p\n", + dev_priv->buffers->handle ); + } + + + dev_priv->agp_size = init->agp_size; + dev_priv->agp_vm_start = RADEON_READ( RADEON_CONFIG_APER_SIZE ); +#if __REALLY_HAVE_AGP + if ( !dev_priv->is_pci ) + dev_priv->agp_buffers_offset = (dev_priv->buffers->offset + - dev->agp->base + + dev_priv->agp_vm_start); + else +#endif + dev_priv->agp_buffers_offset = (dev_priv->buffers->offset + - dev->sg->handle + + dev_priv->agp_vm_start); + + DRM_DEBUG( "dev_priv->agp_size %d\n", + dev_priv->agp_size ); + DRM_DEBUG( "dev_priv->agp_vm_start 0x%x\n", + dev_priv->agp_vm_start ); + DRM_DEBUG( "dev_priv->agp_buffers_offset 0x%lx\n", + dev_priv->agp_buffers_offset ); + + dev_priv->ring.head = ((__volatile__ u32 *) + dev_priv->ring_rptr->handle); + + dev_priv->ring.start = (u32 *)dev_priv->cp_ring->handle; + dev_priv->ring.end = ((u32 *)dev_priv->cp_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->ring.high_mark = RADEON_RING_HIGH_MARK; + +#if __REALLY_HAVE_SG + if ( dev_priv->is_pci ) { + if (!DRM(ati_pcigart_init)( dev, &dev_priv->phys_pci_gart, + &dev_priv->bus_pci_gart)) { + DRM_ERROR( "failed to init PCI GART!\n" ); + dev->dev_private = (void *)dev_priv; + radeon_do_cleanup_cp(dev); + return DRM_ERR(ENOMEM); + } + /* Turn on PCI GART + */ + tmp = RADEON_READ( RADEON_AIC_CNTL ) + | RADEON_PCIGART_TRANSLATE_EN; + RADEON_WRITE( RADEON_AIC_CNTL, tmp ); + + /* set PCI GART page-table base address + */ + RADEON_WRITE( RADEON_AIC_PT_BASE, dev_priv->bus_pci_gart ); + + /* set address range for PCI address translate + */ + RADEON_WRITE( RADEON_AIC_LO_ADDR, dev_priv->agp_vm_start ); + RADEON_WRITE( RADEON_AIC_HI_ADDR, dev_priv->agp_vm_start + + dev_priv->agp_size - 1); + + /* Turn off AGP aperture -- is this required for PCIGART? + */ + RADEON_WRITE( RADEON_MC_AGP_LOCATION, 0xffffffc0 ); /* ?? */ + RADEON_WRITE( RADEON_AGP_COMMAND, 0 ); /* clear AGP_COMMAND */ + } else { +#endif /* __REALLY_HAVE_SG */ + /* Turn off PCI GART + */ + tmp = RADEON_READ( RADEON_AIC_CNTL ) + & ~RADEON_PCIGART_TRANSLATE_EN; + RADEON_WRITE( RADEON_AIC_CNTL, tmp ); +#if __REALLY_HAVE_SG + } +#endif /* __REALLY_HAVE_SG */ + + radeon_cp_load_microcode( dev_priv ); + radeon_cp_init_ring_buffer( dev, dev_priv ); + + dev_priv->last_buf = 0; + + dev->dev_private = (void *)dev_priv; + + radeon_do_engine_reset( dev ); + + return 0; +} + +int radeon_do_cleanup_cp( drm_device_t *dev ) +{ + DRM_DEBUG( "\n" ); + + if ( dev->dev_private ) { + drm_radeon_private_t *dev_priv = dev->dev_private; + + if ( !dev_priv->is_pci ) { + DRM_IOREMAPFREE( dev_priv->cp_ring ); + DRM_IOREMAPFREE( dev_priv->ring_rptr ); + DRM_IOREMAPFREE( dev_priv->buffers ); + } else { +#if __REALLY_HAVE_SG + if (!DRM(ati_pcigart_cleanup)( dev, + dev_priv->phys_pci_gart, + dev_priv->bus_pci_gart )) + DRM_ERROR( "failed to cleanup PCI GART!\n" ); +#endif /* __REALLY_HAVE_SG */ + } + + DRM(free)( dev->dev_private, sizeof(drm_radeon_private_t), + DRM_MEM_DRIVER ); + dev->dev_private = NULL; + } + + return 0; +} + +int radeon_cp_init( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_init_t init; + + DRM_COPY_FROM_USER_IOCTL( init, (drm_radeon_init_t *)data, sizeof(init) ); + + switch ( init.func ) { + case RADEON_INIT_CP: + case RADEON_INIT_R200_CP: + return radeon_do_init_cp( dev, &init ); + case RADEON_CLEANUP_CP: + return radeon_do_cleanup_cp( dev ); + } + + return DRM_ERR(EINVAL); +} + +int radeon_cp_start( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + if ( dev_priv->cp_running ) { + DRM_DEBUG( "%s while CP running\n", __func__ ); + return 0; + } + if ( dev_priv->cp_mode == RADEON_CSQ_PRIDIS_INDDIS ) { + DRM_DEBUG( "%s called with bogus CP mode (%d)\n", + __func__, dev_priv->cp_mode ); + return 0; + } + + radeon_do_cp_start( dev_priv ); + + return 0; +} + +/* Stop the CP. The engine must have been idled before calling this + * routine. + */ +int radeon_cp_stop( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_cp_stop_t stop; + int ret; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + DRM_COPY_FROM_USER_IOCTL( stop, (drm_radeon_cp_stop_t *)data, sizeof(stop) ); + + /* Flush any pending CP commands. This ensures any outstanding + * commands are exectuted by the engine before we turn it off. + */ + if ( stop.flush ) { + radeon_do_cp_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 = radeon_do_cp_idle( dev_priv ); + if ( ret ) return ret; + } + + /* Finally, we can turn off the CP. If the engine isn't idle, + * we will get some dropped triangles as they won't be fully + * rendered before the CP is shut down. + */ + radeon_do_cp_stop( dev_priv ); + + /* Reset the engine */ + radeon_do_engine_reset( dev ); + + return 0; +} + +/* Just reset the CP ring. Called as part of an X Server engine reset. + */ +int radeon_cp_reset( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_DEBUG( "%s called before init done\n", __func__ ); + return DRM_ERR(EINVAL); + } + + radeon_do_cp_reset( dev_priv ); + + /* The CP is no longer running after an engine reset */ + dev_priv->cp_running = 0; + + return 0; +} + +int radeon_cp_idle( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + return radeon_do_cp_idle( dev_priv ); +} + +int radeon_engine_reset( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + return radeon_do_engine_reset( dev ); +} + + +/* ================================================================ + * Fullscreen mode + */ + +/* KW: Deprecated to say the least: + */ +int radeon_fullscreen( DRM_IOCTL_ARGS ) +{ + return 0; +} + + +/* ================================================================ + * Freelist management + */ + +/* Original comment: FIXME: ROTATE_BUFS is a hack to cycle through + * bufs until freelist code is used. Note this hides a problem with + * the scratch register * (used to keep track of last buffer + * completed) being written to before * the last buffer has actually + * completed rendering. + * + * KW: It's also a good way to find free buffers quickly. + */ + +drm_buf_t *radeon_freelist_get( drm_device_t *dev ) +{ + drm_device_dma_t *dma = dev->dma; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_buf_priv_t *buf_priv; + drm_buf_t *buf; + int i, t; + int start; + + if ( ++dev_priv->last_buf >= dma->buf_count ) + dev_priv->last_buf = 0; + + start = dev_priv->last_buf; + + for ( t = 0 ; t < dev_priv->usec_timeout ; t++ ) { + u32 done_age = DRM_READ32(&dev_priv->scratch[1]); + DRM_DEBUG("done_age = %d\n",done_age); + for ( i = start ; i < dma->buf_count ; i++ ) { + buf = dma->buflist[i]; + buf_priv = buf->dev_private; + if ( buf->pid == 0 || (buf->pending && + buf_priv->age <= done_age) ) { + buf->pending = 0; + return buf; + } + start = 0; + } + DRM_UDELAY( 1 ); + } + + DRM_ERROR( "returning NULL!\n" ); + return NULL; +} + +void radeon_freelist_reset( drm_device_t *dev ) +{ + drm_device_dma_t *dma = dev->dma; + drm_radeon_private_t *dev_priv = dev->dev_private; + int i; + + dev_priv->last_buf = 0; + for ( i = 0 ; i < dma->buf_count ; i++ ) { + drm_buf_t *buf = dma->buflist[i]; + drm_radeon_buf_priv_t *buf_priv = buf->dev_private; + buf_priv->age = 0; + } +} + + +/* ================================================================ + * CP command submission + */ + +int radeon_wait_ring( drm_radeon_private_t *dev_priv, int n ) +{ + drm_radeon_ring_buffer_t *ring = &dev_priv->ring; + int i; + + for ( i = 0 ; i < dev_priv->usec_timeout ; i++ ) { + radeon_update_ring_snapshot( ring ); + if ( ring->space > n ) + return 0; + DRM_UDELAY( 1 ); + } + + /* FIXME: This return value is ignored in the BEGIN_RING macro! */ +#if RADEON_FIFO_DEBUG + radeon_status( dev_priv ); + DRM_ERROR( "failed!\n" ); +#endif + return DRM_ERR(EBUSY); +} + +static int radeon_cp_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 = radeon_freelist_get( dev ); + if ( !buf ) return DRM_ERR(EAGAIN); + + buf->pid = DRM_CURRENTPID; + + if ( DRM_COPY_TO_USER( &d->request_indices[i], &buf->idx, + sizeof(buf->idx) ) ) + return DRM_ERR(EFAULT); + if ( DRM_COPY_TO_USER( &d->request_sizes[i], &buf->total, + sizeof(buf->total) ) ) + return DRM_ERR(EFAULT); + + d->granted_count++; + } + return 0; +} + +int radeon_cp_buffers( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_device_dma_t *dma = dev->dma; + int ret = 0; + drm_dma_t d; + + LOCK_TEST_WITH_RETURN( dev ); + + DRM_COPY_FROM_USER_IOCTL( d, (drm_dma_t *)data, sizeof(d) ); + + /* Please don't send us buffers. + */ + if ( d.send_count != 0 ) { + DRM_ERROR( "Process %d trying to send %d buffers via drmDMA\n", + DRM_CURRENTPID, d.send_count ); + return DRM_ERR(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", + DRM_CURRENTPID, d.request_count, dma->buf_count ); + return DRM_ERR(EINVAL); + } + + d.granted_count = 0; + + if ( d.request_count ) { + ret = radeon_cp_get_buffers( dev, &d ); + } + + DRM_COPY_TO_USER_IOCTL( (drm_dma_t *)data, d, sizeof(d) ); + + return ret; +} diff --git a/shared-core/radeon_drm.h b/shared-core/radeon_drm.h new file mode 100644 index 00000000..463a1fa1 --- /dev/null +++ b/shared-core/radeon_drm.h @@ -0,0 +1,512 @@ +/* radeon_drm.h -- Public header for the radeon driver -*- linux-c -*- + * + * Copyright 2000 Precision Insight, Inc., Cedar Park, Texas. + * Copyright 2000 VA Linux Systems, Inc., Fremont, California. + * Copyright 2002 Tungsten Graphics, Inc., Cedar Park, Texas. + * 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 <martin@valinux.com> + * Gareth Hughes <gareth@valinux.com> + * Keith Whitwell <keith_whitwell@yahoo.com> + */ + +#ifndef __RADEON_DRM_H__ +#define __RADEON_DRM_H__ + +/* WARNING: If you change any of these defines, make sure to change the + * defines in the X server file (radeon_sarea.h) + */ +#ifndef __RADEON_SAREA_DEFINES__ +#define __RADEON_SAREA_DEFINES__ + +/* Old style state flags, required for sarea interface (1.1 and 1.2 + * clears) and 1.2 drm_vertex2 ioctl. + */ +#define RADEON_UPLOAD_CONTEXT 0x00000001 +#define RADEON_UPLOAD_VERTFMT 0x00000002 +#define RADEON_UPLOAD_LINE 0x00000004 +#define RADEON_UPLOAD_BUMPMAP 0x00000008 +#define RADEON_UPLOAD_MASKS 0x00000010 +#define RADEON_UPLOAD_VIEWPORT 0x00000020 +#define RADEON_UPLOAD_SETUP 0x00000040 +#define RADEON_UPLOAD_TCL 0x00000080 +#define RADEON_UPLOAD_MISC 0x00000100 +#define RADEON_UPLOAD_TEX0 0x00000200 +#define RADEON_UPLOAD_TEX1 0x00000400 +#define RADEON_UPLOAD_TEX2 0x00000800 +#define RADEON_UPLOAD_TEX0IMAGES 0x00001000 +#define RADEON_UPLOAD_TEX1IMAGES 0x00002000 +#define RADEON_UPLOAD_TEX2IMAGES 0x00004000 +#define RADEON_UPLOAD_CLIPRECTS 0x00008000 /* handled client-side */ +#define RADEON_REQUIRE_QUIESCENCE 0x00010000 +#define RADEON_UPLOAD_ZBIAS 0x00020000 /* version 1.2 and newer */ +#define RADEON_UPLOAD_ALL 0x003effff +#define RADEON_UPLOAD_CONTEXT_ALL 0x003e01ff + + +/* New style per-packet identifiers for use in cmd_buffer ioctl with + * the RADEON_EMIT_PACKET command. Comments relate new packets to old + * state bits and the packet size: + */ +#define RADEON_EMIT_PP_MISC 0 /* context/7 */ +#define RADEON_EMIT_PP_CNTL 1 /* context/3 */ +#define RADEON_EMIT_RB3D_COLORPITCH 2 /* context/1 */ +#define RADEON_EMIT_RE_LINE_PATTERN 3 /* line/2 */ +#define RADEON_EMIT_SE_LINE_WIDTH 4 /* line/1 */ +#define RADEON_EMIT_PP_LUM_MATRIX 5 /* bumpmap/1 */ +#define RADEON_EMIT_PP_ROT_MATRIX_0 6 /* bumpmap/2 */ +#define RADEON_EMIT_RB3D_STENCILREFMASK 7 /* masks/3 */ +#define RADEON_EMIT_SE_VPORT_XSCALE 8 /* viewport/6 */ +#define RADEON_EMIT_SE_CNTL 9 /* setup/2 */ +#define RADEON_EMIT_SE_CNTL_STATUS 10 /* setup/1 */ +#define RADEON_EMIT_RE_MISC 11 /* misc/1 */ +#define RADEON_EMIT_PP_TXFILTER_0 12 /* tex0/6 */ +#define RADEON_EMIT_PP_BORDER_COLOR_0 13 /* tex0/1 */ +#define RADEON_EMIT_PP_TXFILTER_1 14 /* tex1/6 */ +#define RADEON_EMIT_PP_BORDER_COLOR_1 15 /* tex1/1 */ +#define RADEON_EMIT_PP_TXFILTER_2 16 /* tex2/6 */ +#define RADEON_EMIT_PP_BORDER_COLOR_2 17 /* tex2/1 */ +#define RADEON_EMIT_SE_ZBIAS_FACTOR 18 /* zbias/2 */ +#define RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT 19 /* tcl/11 */ +#define RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED 20 /* material/17 */ +#define R200_EMIT_PP_TXCBLEND_0 21 /* tex0/4 */ +#define R200_EMIT_PP_TXCBLEND_1 22 /* tex1/4 */ +#define R200_EMIT_PP_TXCBLEND_2 23 /* tex2/4 */ +#define R200_EMIT_PP_TXCBLEND_3 24 /* tex3/4 */ +#define R200_EMIT_PP_TXCBLEND_4 25 /* tex4/4 */ +#define R200_EMIT_PP_TXCBLEND_5 26 /* tex5/4 */ +#define R200_EMIT_PP_TXCBLEND_6 27 /* ???/4 */ +#define R200_EMIT_PP_TXCBLEND_7 28 /* ???/4 */ +#define R200_EMIT_TCL_LIGHT_MODEL_CTL_0 29 /* tcl/7 */ +#define R200_EMIT_TFACTOR_0 30 /* tf/7 */ +#define R200_EMIT_VTX_FMT_0 31 /* vtx/5 */ +#define R200_EMIT_VAP_CTL 32 /* vap/1 */ +#define R200_EMIT_MATRIX_SELECT_0 33 /* msl/5 */ +#define R200_EMIT_TEX_PROC_CTL_2 34 /* tcg/5 */ +#define R200_EMIT_TCL_UCP_VERT_BLEND_CTL 35 /* tcl/1 */ +#define R200_EMIT_PP_TXFILTER_0 36 /* tex0/6 */ +#define R200_EMIT_PP_TXFILTER_1 37 /* tex1/6 */ +#define R200_EMIT_PP_TXFILTER_2 38 /* tex2/6 */ +#define R200_EMIT_PP_TXFILTER_3 39 /* tex3/6 */ +#define R200_EMIT_PP_TXFILTER_4 40 /* tex4/6 */ +#define R200_EMIT_PP_TXFILTER_5 41 /* tex5/6 */ +#define R200_EMIT_PP_TXOFFSET_0 42 /* tex0/1 */ +#define R200_EMIT_PP_TXOFFSET_1 43 /* tex1/1 */ +#define R200_EMIT_PP_TXOFFSET_2 44 /* tex2/1 */ +#define R200_EMIT_PP_TXOFFSET_3 45 /* tex3/1 */ +#define R200_EMIT_PP_TXOFFSET_4 46 /* tex4/1 */ +#define R200_EMIT_PP_TXOFFSET_5 47 /* tex5/1 */ +#define R200_EMIT_VTE_CNTL 48 /* vte/1 */ +#define R200_EMIT_OUTPUT_VTX_COMP_SEL 49 /* vtx/1 */ +#define R200_EMIT_PP_TAM_DEBUG3 50 /* tam/1 */ +#define R200_EMIT_PP_CNTL_X 51 /* cst/1 */ +#define R200_EMIT_RB3D_DEPTHXY_OFFSET 52 /* cst/1 */ +#define R200_EMIT_RE_AUX_SCISSOR_CNTL 53 /* cst/1 */ +#define R200_EMIT_RE_SCISSOR_TL_0 54 /* cst/2 */ +#define R200_EMIT_RE_SCISSOR_TL_1 55 /* cst/2 */ +#define R200_EMIT_RE_SCISSOR_TL_2 56 /* cst/2 */ +#define R200_EMIT_SE_VAP_CNTL_STATUS 57 /* cst/1 */ +#define R200_EMIT_SE_VTX_STATE_CNTL 58 /* cst/1 */ +#define R200_EMIT_RE_POINTSIZE 59 /* cst/1 */ +#define R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0 60 /* cst/4 */ +#define RADEON_MAX_STATE_PACKETS 61 + + +/* Commands understood by cmd_buffer ioctl. More can be added but + * obviously these can't be removed or changed: + */ +#define RADEON_CMD_PACKET 1 /* emit one of the register packets above */ +#define RADEON_CMD_SCALARS 2 /* emit scalar data */ +#define RADEON_CMD_VECTORS 3 /* emit vector data */ +#define RADEON_CMD_DMA_DISCARD 4 /* discard current dma buf */ +#define RADEON_CMD_PACKET3 5 /* emit hw packet */ +#define RADEON_CMD_PACKET3_CLIP 6 /* emit hw packet wrapped in cliprects */ +#define RADEON_CMD_SCALARS2 7 /* r200 stopgap */ + + +typedef union { + int i; + struct { + unsigned char cmd_type, pad0, pad1, pad2; + } header; + struct { + unsigned char cmd_type, packet_id, pad0, pad1; + } packet; + struct { + unsigned char cmd_type, offset, stride, count; + } scalars; + struct { + unsigned char cmd_type, offset, stride, count; + } vectors; + struct { + unsigned char cmd_type, buf_idx, pad0, pad1; + } dma; +} drm_radeon_cmd_header_t; + + +#define RADEON_FRONT 0x1 +#define RADEON_BACK 0x2 +#define RADEON_DEPTH 0x4 +#define RADEON_STENCIL 0x8 + +/* Primitive types + */ +#define RADEON_POINTS 0x1 +#define RADEON_LINES 0x2 +#define RADEON_LINE_STRIP 0x3 +#define RADEON_TRIANGLES 0x4 +#define RADEON_TRIANGLE_FAN 0x5 +#define RADEON_TRIANGLE_STRIP 0x6 + +/* Vertex/indirect buffer size + */ +#define RADEON_BUFFER_SIZE 65536 + +/* Byte offsets for indirect buffer data + */ +#define RADEON_INDEX_PRIM_OFFSET 20 + +#define RADEON_SCRATCH_REG_OFFSET 32 + +#define RADEON_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 RADEON_LOCAL_TEX_HEAP 0 +#define RADEON_AGP_TEX_HEAP 1 +#define RADEON_NR_TEX_HEAPS 2 +#define RADEON_NR_TEX_REGIONS 64 +#define RADEON_LOG_TEX_GRANULARITY 16 + +#define RADEON_MAX_TEXTURE_LEVELS 12 +#define RADEON_MAX_TEXTURE_UNITS 3 + +#endif /* __RADEON_SAREA_DEFINES__ */ + +typedef struct { + unsigned int red; + unsigned int green; + unsigned int blue; + unsigned int alpha; +} radeon_color_regs_t; + +typedef struct { + /* Context state */ + unsigned int pp_misc; /* 0x1c14 */ + unsigned int pp_fog_color; + unsigned int re_solid_color; + unsigned int rb3d_blendcntl; + unsigned int rb3d_depthoffset; + unsigned int rb3d_depthpitch; + unsigned int rb3d_zstencilcntl; + + unsigned int pp_cntl; /* 0x1c38 */ + unsigned int rb3d_cntl; + unsigned int rb3d_coloroffset; + unsigned int re_width_height; + unsigned int rb3d_colorpitch; + unsigned int se_cntl; + + /* Vertex format state */ + unsigned int se_coord_fmt; /* 0x1c50 */ + + /* Line state */ + unsigned int re_line_pattern; /* 0x1cd0 */ + unsigned int re_line_state; + + unsigned int se_line_width; /* 0x1db8 */ + + /* Bumpmap state */ + unsigned int pp_lum_matrix; /* 0x1d00 */ + + unsigned int pp_rot_matrix_0; /* 0x1d58 */ + unsigned int pp_rot_matrix_1; + + /* Mask state */ + unsigned int rb3d_stencilrefmask; /* 0x1d7c */ + unsigned int rb3d_ropcntl; + unsigned int rb3d_planemask; + + /* Viewport state */ + unsigned int se_vport_xscale; /* 0x1d98 */ + unsigned int se_vport_xoffset; + unsigned int se_vport_yscale; + unsigned int se_vport_yoffset; + unsigned int se_vport_zscale; + unsigned int se_vport_zoffset; + + /* Setup state */ + unsigned int se_cntl_status; /* 0x2140 */ + + /* Misc state */ + unsigned int re_top_left; /* 0x26c0 */ + unsigned int re_misc; +} drm_radeon_context_regs_t; + +typedef struct { + /* Zbias state */ + unsigned int se_zbias_factor; /* 0x1dac */ + unsigned int se_zbias_constant; +} drm_radeon_context2_regs_t; + + +/* Setup registers for each texture unit + */ +typedef struct { + unsigned int pp_txfilter; + unsigned int pp_txformat; + unsigned int pp_txoffset; + unsigned int pp_txcblend; + unsigned int pp_txablend; + unsigned int pp_tfactor; + unsigned int pp_border_color; +} drm_radeon_texture_regs_t; + +typedef struct { + unsigned int start; + unsigned int finish; + unsigned int prim:8; + unsigned int stateidx:8; + unsigned int numverts:16; /* overloaded as offset/64 for elt prims */ + unsigned int vc_format; /* vertex format */ +} drm_radeon_prim_t; + + +typedef struct { + drm_radeon_context_regs_t context; + drm_radeon_texture_regs_t tex[RADEON_MAX_TEXTURE_UNITS]; + drm_radeon_context2_regs_t context2; + unsigned int dirty; +} drm_radeon_state_t; + + +typedef struct { + unsigned char next, prev; + unsigned char in_use; + int age; +} drm_radeon_tex_region_t; + +typedef struct { + /* The channel for communication of state information to the + * kernel on firing a vertex buffer with either of the + * obsoleted vertex/index ioctls. + */ + drm_radeon_context_regs_t context_state; + drm_radeon_texture_regs_t tex_state[RADEON_MAX_TEXTURE_UNITS]; + unsigned int dirty; + unsigned int vertsize; + unsigned int vc_format; + + /* The current cliprects, or a subset thereof. + */ + drm_clip_rect_t boxes[RADEON_NR_SAREA_CLIPRECTS]; + unsigned int nbox; + + /* Counters for client-side throttling of rendering clients. + */ + unsigned int last_frame; + unsigned int last_dispatch; + unsigned int last_clear; + + drm_radeon_tex_region_t tex_list[RADEON_NR_TEX_HEAPS][RADEON_NR_TEX_REGIONS+1]; + int tex_age[RADEON_NR_TEX_HEAPS]; + int ctx_owner; + int pfState; /* number of 3d windows (0,1,2ormore) */ + int pfCurrentPage; /* which buffer is being displayed? */ +} drm_radeon_sarea_t; + + +/* WARNING: If you change any of these defines, make sure to change the + * defines in the Xserver file (xf86drmRadeon.h) + * + * KW: actually it's illegal to change any of this (backwards compatibility). + */ + +/* Radeon specific ioctls + * The device specific ioctl range is 0x40 to 0x79. + */ +#define DRM_IOCTL_RADEON_CP_INIT DRM_IOW( 0x40, drm_radeon_init_t) +#define DRM_IOCTL_RADEON_CP_START DRM_IO( 0x41) +#define DRM_IOCTL_RADEON_CP_STOP DRM_IOW( 0x42, drm_radeon_cp_stop_t) +#define DRM_IOCTL_RADEON_CP_RESET DRM_IO( 0x43) +#define DRM_IOCTL_RADEON_CP_IDLE DRM_IO( 0x44) +#define DRM_IOCTL_RADEON_RESET DRM_IO( 0x45) +#define DRM_IOCTL_RADEON_FULLSCREEN DRM_IOW( 0x46, drm_radeon_fullscreen_t) +#define DRM_IOCTL_RADEON_SWAP DRM_IO( 0x47) +#define DRM_IOCTL_RADEON_CLEAR DRM_IOW( 0x48, drm_radeon_clear_t) +#define DRM_IOCTL_RADEON_VERTEX DRM_IOW( 0x49, drm_radeon_vertex_t) +#define DRM_IOCTL_RADEON_INDICES DRM_IOW( 0x4a, drm_radeon_indices_t) +#define DRM_IOCTL_RADEON_STIPPLE DRM_IOW( 0x4c, drm_radeon_stipple_t) +#define DRM_IOCTL_RADEON_INDIRECT DRM_IOWR(0x4d, drm_radeon_indirect_t) +#define DRM_IOCTL_RADEON_TEXTURE DRM_IOWR(0x4e, drm_radeon_texture_t) +#define DRM_IOCTL_RADEON_VERTEX2 DRM_IOW( 0x4f, drm_radeon_vertex_t) +#define DRM_IOCTL_RADEON_CMDBUF DRM_IOW( 0x50, drm_radeon_cmd_buffer_t) +#define DRM_IOCTL_RADEON_GETPARAM DRM_IOWR(0x51, drm_radeon_getparam_t) +#define DRM_IOCTL_RADEON_FLIP DRM_IO( 0x52) + +typedef struct drm_radeon_init { + enum { + RADEON_INIT_CP = 0x01, + RADEON_CLEANUP_CP = 0x02, + RADEON_INIT_R200_CP = 0x03, + } func; + unsigned long sarea_priv_offset; + int is_pci; + int cp_mode; + int agp_size; + int ring_size; + int usec_timeout; + + 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 long fb_offset; + unsigned long mmio_offset; + unsigned long ring_offset; + unsigned long ring_rptr_offset; + unsigned long buffers_offset; + unsigned long agp_textures_offset; +} drm_radeon_init_t; + +typedef struct drm_radeon_cp_stop { + int flush; + int idle; +} drm_radeon_cp_stop_t; + +typedef struct drm_radeon_fullscreen { + enum { + RADEON_INIT_FULLSCREEN = 0x01, + RADEON_CLEANUP_FULLSCREEN = 0x02 + } func; +} drm_radeon_fullscreen_t; + +#define CLEAR_X1 0 +#define CLEAR_Y1 1 +#define CLEAR_X2 2 +#define CLEAR_Y2 3 +#define CLEAR_DEPTH 4 + +typedef union drm_radeon_clear_rect { + float f[5]; + unsigned int ui[5]; +} drm_radeon_clear_rect_t; + +typedef struct drm_radeon_clear { + unsigned int flags; + unsigned int clear_color; + unsigned int clear_depth; + unsigned int color_mask; + unsigned int depth_mask; /* misnamed field: should be stencil */ + drm_radeon_clear_rect_t *depth_boxes; +} drm_radeon_clear_t; + +typedef struct drm_radeon_vertex { + int prim; + int idx; /* Index of vertex buffer */ + int count; /* Number of vertices in buffer */ + int discard; /* Client finished with buffer? */ +} drm_radeon_vertex_t; + +typedef struct drm_radeon_indices { + int prim; + int idx; + int start; + int end; + int discard; /* Client finished with buffer? */ +} drm_radeon_indices_t; + +/* v1.2 - obsoletes drm_radeon_vertex and drm_radeon_indices + * - allows multiple primitives and state changes in a single ioctl + * - supports driver change to emit native primitives + */ +typedef struct drm_radeon_vertex2 { + int idx; /* Index of vertex buffer */ + int discard; /* Client finished with buffer? */ + int nr_states; + drm_radeon_state_t *state; + int nr_prims; + drm_radeon_prim_t *prim; +} drm_radeon_vertex2_t; + +/* v1.3 - obsoletes drm_radeon_vertex2 + * - allows arbitarily large cliprect list + * - allows updating of tcl packet, vector and scalar state + * - allows memory-efficient description of state updates + * - allows state to be emitted without a primitive + * (for clears, ctx switches) + * - allows more than one dma buffer to be referenced per ioctl + * - supports tcl driver + * - may be extended in future versions with new cmd types, packets + */ +typedef struct drm_radeon_cmd_buffer { + int bufsz; + char *buf; + int nbox; + drm_clip_rect_t *boxes; +} drm_radeon_cmd_buffer_t; + +typedef struct drm_radeon_tex_image { + unsigned int x, y; /* Blit coordinates */ + unsigned int width, height; + const void *data; +} drm_radeon_tex_image_t; + +typedef struct drm_radeon_texture { + int offset; + int pitch; + int format; + int width; /* Texture image coordinates */ + int height; + drm_radeon_tex_image_t *image; +} drm_radeon_texture_t; + +typedef struct drm_radeon_stipple { + unsigned int *mask; +} drm_radeon_stipple_t; + +typedef struct drm_radeon_indirect { + int idx; + int start; + int end; + int discard; +} drm_radeon_indirect_t; + + +/* 1.3: An ioctl to get parameters that aren't available to the 3d + * client any other way. + */ +#define RADEON_PARAM_AGP_BUFFER_OFFSET 0x1 +#define RADEON_PARAM_LAST_FRAME 0x2 +#define RADEON_PARAM_LAST_DISPATCH 0x3 +#define RADEON_PARAM_LAST_CLEAR 0x4 + +typedef struct drm_radeon_getparam { + int param; + int *value; +} drm_radeon_getparam_t; + +#endif diff --git a/shared-core/radeon_drv.h b/shared-core/radeon_drv.h index ae66f12e..aa92ff2c 100644 --- a/shared-core/radeon_drv.h +++ b/shared-core/radeon_drv.h @@ -78,6 +78,9 @@ typedef struct drm_radeon_private { volatile u32 *scratch; int usec_timeout; + + int is_r200; + int is_pci; unsigned long phys_pci_gart; dma_addr_t bus_pci_gart; @@ -277,6 +280,7 @@ extern int radeon_cp_flip( DRM_IOCTL_ARGS ); # define RADEON_STENCIL_ENABLE (1 << 7) # define RADEON_Z_ENABLE (1 << 8) #define RADEON_RB3D_DEPTHOFFSET 0x1c24 +#define RADEON_RB3D_DEPTHPITCH 0x1c28 #define RADEON_RB3D_PLANEMASK 0x1d84 #define RADEON_RB3D_STENCILREFMASK 0x1d7c #define RADEON_RB3D_ZCACHE_MODE 0x3250 @@ -508,6 +512,62 @@ extern int radeon_cp_flip( DRM_IOCTL_ARGS ); #define RADEON_TXFORMAT_ARGB8888 6 #define RADEON_TXFORMAT_RGBA8888 7 +#define R200_PP_TXCBLEND_0 0x2f00 +#define R200_PP_TXCBLEND_1 0x2f10 +#define R200_PP_TXCBLEND_2 0x2f20 +#define R200_PP_TXCBLEND_3 0x2f30 +#define R200_PP_TXCBLEND_4 0x2f40 +#define R200_PP_TXCBLEND_5 0x2f50 +#define R200_PP_TXCBLEND_6 0x2f60 +#define R200_PP_TXCBLEND_7 0x2f70 +#define R200_SE_TCL_LIGHT_MODEL_CTL_0 0x2268 +#define R200_PP_TFACTOR_0 0x2ee0 +#define R200_SE_VTX_FMT_0 0x2088 +#define R200_SE_VAP_CNTL 0x2080 +#define R200_SE_TCL_MATRIX_SEL_0 0x2230 +#define R200_SE_TCL_TEX_PROC_CTL_2 0x22a8 +#define R200_SE_TCL_UCP_VERT_BLEND_CTL 0x22c0 +#define R200_PP_TXFILTER_5 0x2ca0 +#define R200_PP_TXFILTER_4 0x2c80 +#define R200_PP_TXFILTER_3 0x2c60 +#define R200_PP_TXFILTER_2 0x2c40 +#define R200_PP_TXFILTER_1 0x2c20 +#define R200_PP_TXFILTER_0 0x2c00 +#define R200_PP_TXOFFSET_5 0x2d78 +#define R200_PP_TXOFFSET_4 0x2d60 +#define R200_PP_TXOFFSET_3 0x2d48 +#define R200_PP_TXOFFSET_2 0x2d30 +#define R200_PP_TXOFFSET_1 0x2d18 +#define R200_PP_TXOFFSET_0 0x2d00 +#define R200_RE_AUX_SCISSOR_CNTL 0x26f0 +#define R200_SE_VTE_CNTL 0x20b0 +#define R200_SE_TCL_OUTPUT_VTX_COMP_SEL 0x2250 +#define R200_PP_TAM_DEBUG3 0x2d9c +#define R200_PP_CNTL_X 0x2cc4 +#define R200_SE_VAP_CNTL_STATUS 0x2140 +#define R200_RE_SCISSOR_TL_0 0x1cd8 +#define R200_RE_SCISSOR_TL_1 0x1ce0 +#define R200_RE_SCISSOR_TL_2 0x1ce8 +#define R200_RB3D_DEPTHXY_OFFSET 0x1d60 +#define R200_RE_AUX_SCISSOR_CNTL 0x26f0 +#define R200_SE_VTX_STATE_CNTL 0x2180 +#define R200_RE_POINTSIZE 0x2648 +#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0 0x2554 + + +#define SE_VAP_CNTL__TCL_ENA_MASK 0x00000001 +#define SE_VAP_CNTL__FORCE_W_TO_ONE_MASK 0x00010000 +#define SE_VAP_CNTL__VF_MAX_VTX_NUM__SHIFT 0x00000012 +#define SE_VTE_CNTL__VTX_XY_FMT_MASK 0x00000100 +#define SE_VTE_CNTL__VTX_Z_FMT_MASK 0x00000200 +#define SE_VTX_FMT_0__VTX_Z0_PRESENT_MASK 0x00000001 +#define SE_VTX_FMT_0__VTX_W0_PRESENT_MASK 0x00000002 +#define SE_VTX_FMT_0__VTX_COLOR_0_FMT__SHIFT 0x0000000b +#define R200_3D_DRAW_IMMD_2 0xC0003500 +#define R200_SE_VTX_FMT_1 0x208c +#define R200_RE_CNTL 0x1c50 + + /* Constants */ #define RADEON_MAX_USEC_TIMEOUT 100000 /* 100 ms */ @@ -705,16 +765,15 @@ do { \ } \ if (((dev_priv->ring.tail + _nr) & mask) != write) { \ DRM_ERROR( \ - "ADVANCE_RING(): mismatch: nr: %x write: %x\n", \ + "ADVANCE_RING(): mismatch: nr: %x write: %x line: %d\n", \ ((dev_priv->ring.tail + _nr) & mask), \ - write); \ + write, __LINE__); \ } else \ dev_priv->ring.tail = write; \ } while (0) #define COMMIT_RING() do { \ - radeon_flush_write_combine(); \ - RADEON_WRITE( RADEON_CP_RB_WPTR, dev_priv->ring.tail ); \ + RADEON_WRITE( RADEON_CP_RB_WPTR, dev_priv->ring.tail ); \ } while (0) #define OUT_RING( x ) do { \ diff --git a/shared-core/radeon_state.c b/shared-core/radeon_state.c new file mode 100644 index 00000000..b89a35eb --- /dev/null +++ b/shared-core/radeon_state.c @@ -0,0 +1,2074 @@ +/* radeon_state.c -- State support for Radeon -*- linux-c -*- + * + * Copyright 2000 VA Linux Systems, Inc., Fremont, 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 <gareth@valinux.com> + * Kevin E. Martin <martin@valinux.com> + */ + +#include "radeon.h" +#include "drmP.h" +#include "drm.h" +#include "radeon_drm.h" +#include "radeon_drv.h" + + +/* ================================================================ + * CP hardware state programming functions + */ + +static __inline__ void radeon_emit_clip_rect( drm_radeon_private_t *dev_priv, + drm_clip_rect_t *box ) +{ + RING_LOCALS; + + DRM_DEBUG( " box: x1=%d y1=%d x2=%d y2=%d\n", + box->x1, box->y1, box->x2, box->y2 ); + + BEGIN_RING( 4 ); + OUT_RING( CP_PACKET0( RADEON_RE_TOP_LEFT, 0 ) ); + OUT_RING( (box->y1 << 16) | box->x1 ); + OUT_RING( CP_PACKET0( RADEON_RE_WIDTH_HEIGHT, 0 ) ); +/* OUT_RING( ((box->y2 - 1) << 16) | (box->x2 - 1) );*/ + OUT_RING( (box->y2 << 16) | box->x2 ); + ADVANCE_RING(); +} + +/* Emit 1.1 state + */ +static void radeon_emit_state( drm_radeon_private_t *dev_priv, + drm_radeon_context_regs_t *ctx, + drm_radeon_texture_regs_t *tex, + unsigned int dirty ) +{ + RING_LOCALS; + DRM_DEBUG( "dirty=0x%08x\n", dirty ); + + if ( dirty & RADEON_UPLOAD_CONTEXT ) { + BEGIN_RING( 14 ); + OUT_RING( CP_PACKET0( RADEON_PP_MISC, 6 ) ); + OUT_RING( ctx->pp_misc ); + OUT_RING( ctx->pp_fog_color ); + OUT_RING( ctx->re_solid_color ); + OUT_RING( ctx->rb3d_blendcntl ); + OUT_RING( ctx->rb3d_depthoffset ); + OUT_RING( ctx->rb3d_depthpitch ); + OUT_RING( ctx->rb3d_zstencilcntl ); + OUT_RING( CP_PACKET0( RADEON_PP_CNTL, 2 ) ); + OUT_RING( ctx->pp_cntl ); + OUT_RING( ctx->rb3d_cntl ); + OUT_RING( ctx->rb3d_coloroffset ); + OUT_RING( CP_PACKET0( RADEON_RB3D_COLORPITCH, 0 ) ); + OUT_RING( ctx->rb3d_colorpitch ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_VERTFMT ) { + BEGIN_RING( 2 ); + OUT_RING( CP_PACKET0( RADEON_SE_COORD_FMT, 0 ) ); + OUT_RING( ctx->se_coord_fmt ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_LINE ) { + BEGIN_RING( 5 ); + OUT_RING( CP_PACKET0( RADEON_RE_LINE_PATTERN, 1 ) ); + OUT_RING( ctx->re_line_pattern ); + OUT_RING( ctx->re_line_state ); + OUT_RING( CP_PACKET0( RADEON_SE_LINE_WIDTH, 0 ) ); + OUT_RING( ctx->se_line_width ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_BUMPMAP ) { + BEGIN_RING( 5 ); + OUT_RING( CP_PACKET0( RADEON_PP_LUM_MATRIX, 0 ) ); + OUT_RING( ctx->pp_lum_matrix ); + OUT_RING( CP_PACKET0( RADEON_PP_ROT_MATRIX_0, 1 ) ); + OUT_RING( ctx->pp_rot_matrix_0 ); + OUT_RING( ctx->pp_rot_matrix_1 ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_MASKS ) { + BEGIN_RING( 4 ); + OUT_RING( CP_PACKET0( RADEON_RB3D_STENCILREFMASK, 2 ) ); + OUT_RING( ctx->rb3d_stencilrefmask ); + OUT_RING( ctx->rb3d_ropcntl ); + OUT_RING( ctx->rb3d_planemask ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_VIEWPORT ) { + BEGIN_RING( 7 ); + OUT_RING( CP_PACKET0( RADEON_SE_VPORT_XSCALE, 5 ) ); + OUT_RING( ctx->se_vport_xscale ); + OUT_RING( ctx->se_vport_xoffset ); + OUT_RING( ctx->se_vport_yscale ); + OUT_RING( ctx->se_vport_yoffset ); + OUT_RING( ctx->se_vport_zscale ); + OUT_RING( ctx->se_vport_zoffset ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_SETUP ) { + BEGIN_RING( 4 ); + OUT_RING( CP_PACKET0( RADEON_SE_CNTL, 0 ) ); + OUT_RING( ctx->se_cntl ); + OUT_RING( CP_PACKET0( RADEON_SE_CNTL_STATUS, 0 ) ); + OUT_RING( ctx->se_cntl_status ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_MISC ) { + BEGIN_RING( 2 ); + OUT_RING( CP_PACKET0( RADEON_RE_MISC, 0 ) ); + OUT_RING( ctx->re_misc ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_TEX0 ) { + BEGIN_RING( 9 ); + OUT_RING( CP_PACKET0( RADEON_PP_TXFILTER_0, 5 ) ); + OUT_RING( tex[0].pp_txfilter ); + OUT_RING( tex[0].pp_txformat ); + OUT_RING( tex[0].pp_txoffset ); + OUT_RING( tex[0].pp_txcblend ); + OUT_RING( tex[0].pp_txablend ); + OUT_RING( tex[0].pp_tfactor ); + OUT_RING( CP_PACKET0( RADEON_PP_BORDER_COLOR_0, 0 ) ); + OUT_RING( tex[0].pp_border_color ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_TEX1 ) { + BEGIN_RING( 9 ); + OUT_RING( CP_PACKET0( RADEON_PP_TXFILTER_1, 5 ) ); + OUT_RING( tex[1].pp_txfilter ); + OUT_RING( tex[1].pp_txformat ); + OUT_RING( tex[1].pp_txoffset ); + OUT_RING( tex[1].pp_txcblend ); + OUT_RING( tex[1].pp_txablend ); + OUT_RING( tex[1].pp_tfactor ); + OUT_RING( CP_PACKET0( RADEON_PP_BORDER_COLOR_1, 0 ) ); + OUT_RING( tex[1].pp_border_color ); + ADVANCE_RING(); + } + + if ( dirty & RADEON_UPLOAD_TEX2 ) { + BEGIN_RING( 9 ); + OUT_RING( CP_PACKET0( RADEON_PP_TXFILTER_2, 5 ) ); + OUT_RING( tex[2].pp_txfilter ); + OUT_RING( tex[2].pp_txformat ); + OUT_RING( tex[2].pp_txoffset ); + OUT_RING( tex[2].pp_txcblend ); + OUT_RING( tex[2].pp_txablend ); + OUT_RING( tex[2].pp_tfactor ); + OUT_RING( CP_PACKET0( RADEON_PP_BORDER_COLOR_2, 0 ) ); + OUT_RING( tex[2].pp_border_color ); + ADVANCE_RING(); + } +} + +/* Emit 1.2 state + */ +static void radeon_emit_state2( drm_radeon_private_t *dev_priv, + drm_radeon_state_t *state ) +{ + RING_LOCALS; + + if (state->dirty & RADEON_UPLOAD_ZBIAS) { + BEGIN_RING( 3 ); + OUT_RING( CP_PACKET0( RADEON_SE_ZBIAS_FACTOR, 1 ) ); + OUT_RING( state->context2.se_zbias_factor ); + OUT_RING( state->context2.se_zbias_constant ); + ADVANCE_RING(); + } + + radeon_emit_state( dev_priv, &state->context, + state->tex, state->dirty ); +} + +/* New (1.3) state mechanism. 3 commands (packet, scalar, vector) in + * 1.3 cmdbuffers allow all previous state to be updated as well as + * the tcl scalar and vector areas. + */ +static struct { + int start; + int len; + const char *name; +} packet[RADEON_MAX_STATE_PACKETS] = { + { RADEON_PP_MISC,7,"RADEON_PP_MISC" }, + { RADEON_PP_CNTL,3,"RADEON_PP_CNTL" }, + { RADEON_RB3D_COLORPITCH,1,"RADEON_RB3D_COLORPITCH" }, + { RADEON_RE_LINE_PATTERN,2,"RADEON_RE_LINE_PATTERN" }, + { RADEON_SE_LINE_WIDTH,1,"RADEON_SE_LINE_WIDTH" }, + { RADEON_PP_LUM_MATRIX,1,"RADEON_PP_LUM_MATRIX" }, + { RADEON_PP_ROT_MATRIX_0,2,"RADEON_PP_ROT_MATRIX_0" }, + { RADEON_RB3D_STENCILREFMASK,3,"RADEON_RB3D_STENCILREFMASK" }, + { RADEON_SE_VPORT_XSCALE,6,"RADEON_SE_VPORT_XSCALE" }, + { RADEON_SE_CNTL,2,"RADEON_SE_CNTL" }, + { RADEON_SE_CNTL_STATUS,1,"RADEON_SE_CNTL_STATUS" }, + { RADEON_RE_MISC,1,"RADEON_RE_MISC" }, + { RADEON_PP_TXFILTER_0,6,"RADEON_PP_TXFILTER_0" }, + { RADEON_PP_BORDER_COLOR_0,1,"RADEON_PP_BORDER_COLOR_0" }, + { RADEON_PP_TXFILTER_1,6,"RADEON_PP_TXFILTER_1" }, + { RADEON_PP_BORDER_COLOR_1,1,"RADEON_PP_BORDER_COLOR_1" }, + { RADEON_PP_TXFILTER_2,6,"RADEON_PP_TXFILTER_2" }, + { RADEON_PP_BORDER_COLOR_2,1,"RADEON_PP_BORDER_COLOR_2" }, + { RADEON_SE_ZBIAS_FACTOR,2,"RADEON_SE_ZBIAS_FACTOR" }, + { RADEON_SE_TCL_OUTPUT_VTX_FMT,11,"RADEON_SE_TCL_OUTPUT_VTX_FMT" }, + { RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED,17,"RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED" }, + { R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0" }, + { R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1" }, + { R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2" }, + { R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3" }, + { R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4" }, + { R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5" }, + { R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6" }, + { R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7" }, + { R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0" }, + { R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0" }, + { R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0" }, + { R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL" }, + { R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0" }, + { R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2" }, + { R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL" }, + { R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0" }, + { R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1" }, + { R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2" }, + { R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3" }, + { R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4" }, + { R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5" }, + { R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0" }, + { R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1" }, + { R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2" }, + { R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3" }, + { R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4" }, + { R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5" }, + { R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL" }, + { R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1, "R200_SE_TCL_OUTPUT_VTX_COMP_SEL" }, + { R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3" }, + { R200_PP_CNTL_X, 1, "R200_PP_CNTL_X" }, + { R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET" }, + { R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL" }, + { R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0" }, + { R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1" }, + { R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2" }, + { R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS" }, + { R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL" }, + { R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE" }, + { R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4, "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0" }, +}; + + + + + + + + + + +#if RADEON_PERFORMANCE_BOXES +/* ================================================================ + * Performance monitoring functions + */ + +static void radeon_clear_box( drm_radeon_private_t *dev_priv, + int x, int y, int w, int h, + int r, int g, int b ) +{ + u32 pitch, offset; + u32 color; + RING_LOCALS; + + switch ( dev_priv->color_fmt ) { + case RADEON_COLOR_FORMAT_RGB565: + color = (((r & 0xf8) << 8) | + ((g & 0xfc) << 3) | + ((b & 0xf8) >> 3)); + break; + case RADEON_COLOR_FORMAT_ARGB8888: + default: + 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( CP_PACKET3( RADEON_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_SOLID_COLOR | + (dev_priv->color_fmt << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP3_P | + RADEON_GMC_CLR_CMP_CNTL_DIS ); + + OUT_RING( (pitch << 22) | (offset >> 5) ); + OUT_RING( color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); +} + +static void radeon_cp_performance_boxes( drm_radeon_private_t *dev_priv ) +{ + if ( atomic_read( &dev_priv->idle_count ) == 0 ) { + radeon_clear_box( dev_priv, 64, 4, 8, 8, 0, 255, 0 ); + } else { + atomic_set( &dev_priv->idle_count, 0 ); + } +} + +#endif + + +/* ================================================================ + * CP command dispatch functions + */ + +static void radeon_cp_dispatch_clear( drm_device_t *dev, + drm_radeon_clear_t *clear, + drm_radeon_clear_rect_t *depth_boxes ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_radeon_depth_clear_t *depth_clear = &dev_priv->depth_clear; + int nbox = sarea_priv->nbox; + drm_clip_rect_t *pbox = sarea_priv->boxes; + unsigned int flags = clear->flags; + u32 rb3d_cntl = 0, rb3d_stencilrefmask= 0; + int i; + RING_LOCALS; + DRM_DEBUG( "flags = 0x%x\n", flags ); + + if ( dev_priv->page_flipping && dev_priv->current_page == 1 ) { + unsigned int tmp = flags; + + flags &= ~(RADEON_FRONT | RADEON_BACK); + if ( tmp & RADEON_FRONT ) flags |= RADEON_BACK; + if ( tmp & RADEON_BACK ) flags |= RADEON_FRONT; + } + + if ( flags & (RADEON_FRONT | RADEON_BACK) ) { + + BEGIN_RING( 4 ); + + /* Ensure the 3D stream is idle before doing a + * 2D fill to clear the front or back buffer. + */ + RADEON_WAIT_UNTIL_3D_IDLE(); + + OUT_RING( CP_PACKET0( RADEON_DP_WRITE_MASK, 0 ) ); + OUT_RING( clear->color_mask ); + + ADVANCE_RING(); + + /* Make sure we restore the 3D state next time. + */ + dev_priv->sarea_priv->ctx_owner = 0; + + 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", + x, y, w, h, flags ); + + if ( flags & RADEON_FRONT ) { + BEGIN_RING( 6 ); + + OUT_RING( CP_PACKET3( RADEON_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_SOLID_COLOR | + (dev_priv->color_fmt << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP3_P | + RADEON_GMC_CLR_CMP_CNTL_DIS ); + + OUT_RING( dev_priv->front_pitch_offset ); + OUT_RING( clear->clear_color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); + } + + if ( flags & RADEON_BACK ) { + BEGIN_RING( 6 ); + + OUT_RING( CP_PACKET3( RADEON_CNTL_PAINT_MULTI, 4 ) ); + OUT_RING( RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_SOLID_COLOR | + (dev_priv->color_fmt << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP3_P | + RADEON_GMC_CLR_CMP_CNTL_DIS ); + + OUT_RING( dev_priv->back_pitch_offset ); + OUT_RING( clear->clear_color ); + + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + ADVANCE_RING(); + } + } + } + + /* We have to clear the depth and/or stencil buffers by + * rendering a quad into just those buffers. Thus, we have to + * make sure the 3D engine is configured correctly. + */ + if ( dev_priv->is_r200 && + (flags & (RADEON_DEPTH | RADEON_STENCIL)) ) { + + int tempPP_CNTL; + int tempRE_CNTL; + int tempRB3D_CNTL; + int tempRB3D_ZSTENCILCNTL; + int tempRB3D_STENCILREFMASK; + int tempRB3D_PLANEMASK; + int tempSE_CNTL; + int tempSE_VTE_CNTL; + int tempSE_VTX_FMT_0; + int tempSE_VTX_FMT_1; + int tempSE_VAP_CNTL; + int tempRE_AUX_SCISSOR_CNTL; + + tempPP_CNTL = 0; + tempRE_CNTL = 0; + + tempRB3D_CNTL = depth_clear->rb3d_cntl; + tempRB3D_CNTL &= ~(1<<15); /* unset radeon magic flag */ + + tempRB3D_ZSTENCILCNTL = depth_clear->rb3d_zstencilcntl; + tempRB3D_STENCILREFMASK = 0x0; + + tempSE_CNTL = depth_clear->se_cntl; + + + + /* Disable TCL */ + + tempSE_VAP_CNTL = (/* SE_VAP_CNTL__FORCE_W_TO_ONE_MASK | */ + (0x9 << SE_VAP_CNTL__VF_MAX_VTX_NUM__SHIFT)); + + tempRB3D_PLANEMASK = 0x0; + + tempRE_AUX_SCISSOR_CNTL = 0x0; + + tempSE_VTE_CNTL = + SE_VTE_CNTL__VTX_XY_FMT_MASK | + SE_VTE_CNTL__VTX_Z_FMT_MASK; + + /* Vertex format (X, Y, Z, W)*/ + tempSE_VTX_FMT_0 = + SE_VTX_FMT_0__VTX_Z0_PRESENT_MASK | + SE_VTX_FMT_0__VTX_W0_PRESENT_MASK; + tempSE_VTX_FMT_1 = 0x0; + + + /* + * Depth buffer specific enables + */ + if (flags & RADEON_DEPTH) { + /* Enable depth buffer */ + tempRB3D_CNTL |= RADEON_Z_ENABLE; + } else { + /* Disable depth buffer */ + tempRB3D_CNTL &= ~RADEON_Z_ENABLE; + } + + /* + * Stencil buffer specific enables + */ + if ( flags & RADEON_STENCIL ) { + tempRB3D_CNTL |= RADEON_STENCIL_ENABLE; + tempRB3D_STENCILREFMASK = clear->depth_mask; + } else { + tempRB3D_CNTL &= ~RADEON_STENCIL_ENABLE; + tempRB3D_STENCILREFMASK = 0x00000000; + } + + BEGIN_RING( 26 ); + RADEON_WAIT_UNTIL_2D_IDLE(); + + OUT_RING_REG( RADEON_PP_CNTL, tempPP_CNTL ); + OUT_RING_REG( R200_RE_CNTL, tempRE_CNTL ); + OUT_RING_REG( RADEON_RB3D_CNTL, tempRB3D_CNTL ); + OUT_RING_REG( RADEON_RB3D_ZSTENCILCNTL, + tempRB3D_ZSTENCILCNTL ); + OUT_RING_REG( RADEON_RB3D_STENCILREFMASK, + tempRB3D_STENCILREFMASK ); + OUT_RING_REG( RADEON_RB3D_PLANEMASK, tempRB3D_PLANEMASK ); + OUT_RING_REG( RADEON_SE_CNTL, tempSE_CNTL ); + OUT_RING_REG( R200_SE_VTE_CNTL, tempSE_VTE_CNTL ); + OUT_RING_REG( R200_SE_VTX_FMT_0, tempSE_VTX_FMT_0 ); + OUT_RING_REG( R200_SE_VTX_FMT_1, tempSE_VTX_FMT_1 ); + OUT_RING_REG( R200_SE_VAP_CNTL, tempSE_VAP_CNTL ); + OUT_RING_REG( R200_RE_AUX_SCISSOR_CNTL, + tempRE_AUX_SCISSOR_CNTL ); + ADVANCE_RING(); + + /* Make sure we restore the 3D state next time. + */ + dev_priv->sarea_priv->ctx_owner = 0; + + for ( i = 0 ; i < nbox ; i++ ) { + + /* Funny that this should be required -- + * sets top-left? + */ + radeon_emit_clip_rect( dev_priv, + &sarea_priv->boxes[i] ); + + BEGIN_RING( 14 ); + OUT_RING( CP_PACKET3( R200_3D_DRAW_IMMD_2, 12 ) ); + OUT_RING( (RADEON_PRIM_TYPE_RECT_LIST | + RADEON_PRIM_WALK_RING | + (3 << RADEON_NUM_VERTICES_SHIFT)) ); + OUT_RING( depth_boxes[i].ui[CLEAR_X1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x3f800000 ); + OUT_RING( depth_boxes[i].ui[CLEAR_X1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x3f800000 ); + OUT_RING( depth_boxes[i].ui[CLEAR_X2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x3f800000 ); + ADVANCE_RING(); + } + } + else if ( (flags & (RADEON_DEPTH | RADEON_STENCIL)) ) { + + rb3d_cntl = depth_clear->rb3d_cntl; + + if ( flags & RADEON_DEPTH ) { + rb3d_cntl |= RADEON_Z_ENABLE; + } else { + rb3d_cntl &= ~RADEON_Z_ENABLE; + } + + if ( flags & RADEON_STENCIL ) { + rb3d_cntl |= RADEON_STENCIL_ENABLE; + rb3d_stencilrefmask = clear->depth_mask; /* misnamed field */ + } else { + rb3d_cntl &= ~RADEON_STENCIL_ENABLE; + rb3d_stencilrefmask = 0x00000000; + } + + BEGIN_RING( 13 ); + RADEON_WAIT_UNTIL_2D_IDLE(); + + OUT_RING( CP_PACKET0( RADEON_PP_CNTL, 1 ) ); + OUT_RING( 0x00000000 ); + OUT_RING( rb3d_cntl ); + + OUT_RING_REG( RADEON_RB3D_ZSTENCILCNTL, + depth_clear->rb3d_zstencilcntl ); + OUT_RING_REG( RADEON_RB3D_STENCILREFMASK, + rb3d_stencilrefmask ); + OUT_RING_REG( RADEON_RB3D_PLANEMASK, + 0x00000000 ); + OUT_RING_REG( RADEON_SE_CNTL, + depth_clear->se_cntl ); + ADVANCE_RING(); + + /* Make sure we restore the 3D state next time. + */ + dev_priv->sarea_priv->ctx_owner = 0; + + for ( i = 0 ; i < nbox ; i++ ) { + + /* Funny that this should be required -- + * sets top-left? + */ + radeon_emit_clip_rect( dev_priv, + &sarea_priv->boxes[i] ); + + BEGIN_RING( 15 ); + + OUT_RING( CP_PACKET3( RADEON_3D_DRAW_IMMD, 13 ) ); + OUT_RING( RADEON_VTX_Z_PRESENT | + RADEON_VTX_PKCOLOR_PRESENT); + OUT_RING( (RADEON_PRIM_TYPE_RECT_LIST | + RADEON_PRIM_WALK_RING | + RADEON_MAOS_ENABLE | + RADEON_VTX_FMT_RADEON_MODE | + (3 << RADEON_NUM_VERTICES_SHIFT)) ); + + + OUT_RING( depth_boxes[i].ui[CLEAR_X1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x0 ); + + OUT_RING( depth_boxes[i].ui[CLEAR_X1] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x0 ); + + OUT_RING( depth_boxes[i].ui[CLEAR_X2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_Y2] ); + OUT_RING( depth_boxes[i].ui[CLEAR_DEPTH] ); + OUT_RING( 0x0 ); + + ADVANCE_RING(); + } + } + + /* Increment the clear counter. The client-side 3D driver must + * wait on this value before performing the clear ioctl. We + * need this because the card's so damned fast... + */ + dev_priv->sarea_priv->last_clear++; + + BEGIN_RING( 4 ); + + RADEON_CLEAR_AGE( dev_priv->sarea_priv->last_clear ); + RADEON_WAIT_UNTIL_IDLE(); + + ADVANCE_RING(); +} + +static void radeon_cp_dispatch_swap( drm_device_t *dev ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + int nbox = sarea_priv->nbox; + drm_clip_rect_t *pbox = sarea_priv->boxes; + int i; + RING_LOCALS; + DRM_DEBUG( "\n" ); + +#if RADEON_PERFORMANCE_BOXES + /* Do some trivial performance monitoring... + */ + radeon_cp_performance_boxes( dev_priv ); +#endif + + /* Wait for the 3D stream to idle before dispatching the bitblt. + * This will prevent data corruption between the two streams. + */ + BEGIN_RING( 2 ); + + RADEON_WAIT_UNTIL_3D_IDLE(); + + ADVANCE_RING(); + + 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 swap %d,%d-%d,%d\n", + x, y, w, h ); + + BEGIN_RING( 7 ); + + OUT_RING( CP_PACKET3( RADEON_CNTL_BITBLT_MULTI, 5 ) ); + OUT_RING( RADEON_GMC_SRC_PITCH_OFFSET_CNTL | + RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_NONE | + (dev_priv->color_fmt << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP3_S | + RADEON_DP_SRC_SOURCE_MEMORY | + RADEON_GMC_CLR_CMP_CNTL_DIS | + RADEON_GMC_WR_MSK_DIS ); + + /* Make this work even if front & back are flipped: + */ + if (dev_priv->current_page == 0) { + OUT_RING( dev_priv->back_pitch_offset ); + OUT_RING( dev_priv->front_pitch_offset ); + } + else { + OUT_RING( dev_priv->front_pitch_offset ); + OUT_RING( dev_priv->back_pitch_offset ); + } + + OUT_RING( (x << 16) | y ); + OUT_RING( (x << 16) | y ); + OUT_RING( (w << 16) | h ); + + 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( 4 ); + + RADEON_FRAME_AGE( dev_priv->sarea_priv->last_frame ); + RADEON_WAIT_UNTIL_2D_IDLE(); + + ADVANCE_RING(); +} + +static void radeon_cp_dispatch_flip( drm_device_t *dev ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + RING_LOCALS; + DRM_DEBUG( "%s: page=%d pfCurrentPage=%d\n", + __FUNCTION__, + dev_priv->current_page, + dev_priv->sarea_priv->pfCurrentPage); + +#if RADEON_PERFORMANCE_BOXES + /* Do some trivial performance monitoring... + */ + radeon_cp_performance_boxes( dev_priv ); +#endif + + BEGIN_RING( 4 ); + + RADEON_WAIT_UNTIL_3D_IDLE(); +/* + RADEON_WAIT_UNTIL_PAGE_FLIPPED(); +*/ + OUT_RING( CP_PACKET0( RADEON_CRTC_OFFSET, 0 ) ); + + if ( dev_priv->current_page == 0 ) { + OUT_RING( dev_priv->back_offset ); + dev_priv->current_page = 1; + } else { + OUT_RING( dev_priv->front_offset ); + dev_priv->current_page = 0; + } + + 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++; + dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page; + + BEGIN_RING( 2 ); + + RADEON_FRAME_AGE( dev_priv->sarea_priv->last_frame ); + + ADVANCE_RING(); +} + +static int bad_prim_vertex_nr( int primitive, int nr ) +{ + switch (primitive & RADEON_PRIM_TYPE_MASK) { + case RADEON_PRIM_TYPE_NONE: + case RADEON_PRIM_TYPE_POINT: + return nr < 1; + case RADEON_PRIM_TYPE_LINE: + return (nr & 1) || nr == 0; + case RADEON_PRIM_TYPE_LINE_STRIP: + return nr < 2; + case RADEON_PRIM_TYPE_TRI_LIST: + case RADEON_PRIM_TYPE_3VRT_POINT_LIST: + case RADEON_PRIM_TYPE_3VRT_LINE_LIST: + case RADEON_PRIM_TYPE_RECT_LIST: + return nr % 3 || nr == 0; + case RADEON_PRIM_TYPE_TRI_FAN: + case RADEON_PRIM_TYPE_TRI_STRIP: + return nr < 3; + default: + return 1; + } +} + + + +typedef struct { + unsigned int start; + unsigned int finish; + unsigned int prim; + unsigned int numverts; + unsigned int offset; + unsigned int vc_format; +} drm_radeon_tcl_prim_t; + +static void radeon_cp_dispatch_vertex( drm_device_t *dev, + drm_buf_t *buf, + drm_radeon_tcl_prim_t *prim, + drm_clip_rect_t *boxes, + int nbox ) + +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_clip_rect_t box; + int offset = dev_priv->agp_buffers_offset + buf->offset + prim->start; + int numverts = (int)prim->numverts; + int i = 0; + RING_LOCALS; + + DRM_DEBUG("hwprim 0x%x vfmt 0x%x %d..%d %d verts\n", + prim->prim, + prim->vc_format, + prim->start, + prim->finish, + prim->numverts); + + if (bad_prim_vertex_nr( prim->prim, prim->numverts )) { + DRM_ERROR( "bad prim %x numverts %d\n", + prim->prim, prim->numverts ); + return; + } + + do { + /* Emit the next cliprect */ + if ( i < nbox ) { + if (DRM_COPY_FROM_USER_UNCHECKED( &box, &boxes[i], sizeof(box) )) + return; + + radeon_emit_clip_rect( dev_priv, &box ); + } + + /* Emit the vertex buffer rendering commands */ + BEGIN_RING( 5 ); + + OUT_RING( CP_PACKET3( RADEON_3D_RNDR_GEN_INDX_PRIM, 3 ) ); + OUT_RING( offset ); + OUT_RING( numverts ); + OUT_RING( prim->vc_format ); + OUT_RING( prim->prim | RADEON_PRIM_WALK_LIST | + RADEON_COLOR_ORDER_RGBA | + RADEON_VTX_FMT_RADEON_MODE | + (numverts << RADEON_NUM_VERTICES_SHIFT) ); + + ADVANCE_RING(); + + i++; + } while ( i < nbox ); +} + + + +static void radeon_cp_discard_buffer( drm_device_t *dev, drm_buf_t *buf ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_buf_priv_t *buf_priv = buf->dev_private; + RING_LOCALS; + + buf_priv->age = ++dev_priv->sarea_priv->last_dispatch; + + /* Emit the vertex buffer age */ + BEGIN_RING( 2 ); + RADEON_DISPATCH_AGE( buf_priv->age ); + ADVANCE_RING(); + + buf->pending = 1; + buf->used = 0; +} + +static void radeon_cp_dispatch_indirect( drm_device_t *dev, + drm_buf_t *buf, + int start, int end ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + RING_LOCALS; + DRM_DEBUG( "indirect: buf=%d s=0x%x e=0x%x\n", + buf->idx, start, end ); + + if ( start != end ) { + int offset = (dev_priv->agp_buffers_offset + + 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 CP packet. + */ + if ( dwords & 1 ) { + u32 *data = (u32 *) + ((char *)dev_priv->buffers->handle + + buf->offset + start); + data[dwords++] = RADEON_CP_PACKET2; + } + + /* Fire off the indirect buffer */ + BEGIN_RING( 3 ); + + OUT_RING( CP_PACKET0( RADEON_CP_IB_BASE, 1 ) ); + OUT_RING( offset ); + OUT_RING( dwords ); + + ADVANCE_RING(); + } +} + + +static void radeon_cp_dispatch_indices( drm_device_t *dev, + drm_buf_t *elt_buf, + drm_radeon_tcl_prim_t *prim, + drm_clip_rect_t *boxes, + int nbox ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_clip_rect_t box; + int offset = dev_priv->agp_buffers_offset + prim->offset; + u32 *data; + int dwords; + int i = 0; + int start = prim->start + RADEON_INDEX_PRIM_OFFSET; + int count = (prim->finish - start) / sizeof(u16); + + DRM_DEBUG("hwprim 0x%x vfmt 0x%x %d..%d offset: %x nr %d\n", + prim->prim, + prim->vc_format, + prim->start, + prim->finish, + prim->offset, + prim->numverts); + + if (bad_prim_vertex_nr( prim->prim, count )) { + DRM_ERROR( "bad prim %x count %d\n", + prim->prim, count ); + return; + } + + + if ( start >= prim->finish || + (prim->start & 0x7) ) { + DRM_ERROR( "buffer prim %d\n", prim->prim ); + return; + } + + dwords = (prim->finish - prim->start + 3) / sizeof(u32); + + data = (u32 *)((char *)dev_priv->buffers->handle + + elt_buf->offset + prim->start); + + data[0] = CP_PACKET3( RADEON_3D_RNDR_GEN_INDX_PRIM, dwords-2 ); + data[1] = offset; + data[2] = prim->numverts; + data[3] = prim->vc_format; + data[4] = (prim->prim | + RADEON_PRIM_WALK_IND | + RADEON_COLOR_ORDER_RGBA | + RADEON_VTX_FMT_RADEON_MODE | + (count << RADEON_NUM_VERTICES_SHIFT) ); + + do { + if ( i < nbox ) { + if (DRM_COPY_FROM_USER_UNCHECKED( &box, &boxes[i], sizeof(box) )) + return; + + radeon_emit_clip_rect( dev_priv, &box ); + } + + radeon_cp_dispatch_indirect( dev, elt_buf, + prim->start, + prim->finish ); + + i++; + } while ( i < nbox ); + +} + +#define RADEON_MAX_TEXTURE_SIZE (RADEON_BUFFER_SIZE - 8 * sizeof(u32)) + +static int radeon_cp_dispatch_texture( drm_device_t *dev, + drm_radeon_texture_t *tex, + drm_radeon_tex_image_t *image ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_buf_t *buf; + u32 format; + u32 *buffer; + const u8 *data; + int size, dwords, tex_width, blit_width; + u32 y, height; + int ret = 0, i; + RING_LOCALS; + + /* FIXME: Be smarter about this... + */ + buf = radeon_freelist_get( dev ); + if ( !buf ) return DRM_ERR(EAGAIN); + + DRM_DEBUG( "tex: ofs=0x%x p=%d f=%d x=%hd y=%hd w=%hd h=%hd\n", + tex->offset >> 10, tex->pitch, tex->format, + image->x, image->y, image->width, image->height ); + + /* 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 ( tex->format ) { + case RADEON_TXFORMAT_ARGB8888: + case RADEON_TXFORMAT_RGBA8888: + format = RADEON_COLOR_FORMAT_ARGB8888; + tex_width = tex->width * 4; + blit_width = image->width * 4; + break; + case RADEON_TXFORMAT_AI88: + case RADEON_TXFORMAT_ARGB1555: + case RADEON_TXFORMAT_RGB565: + case RADEON_TXFORMAT_ARGB4444: + format = RADEON_COLOR_FORMAT_RGB565; + tex_width = tex->width * 2; + blit_width = image->width * 2; + break; + case RADEON_TXFORMAT_I8: + case RADEON_TXFORMAT_RGB332: + format = RADEON_COLOR_FORMAT_CI8; + tex_width = tex->width * 1; + blit_width = image->width * 1; + break; + default: + DRM_ERROR( "invalid texture format %d\n", tex->format ); + return DRM_ERR(EINVAL); + } + + DRM_DEBUG( " tex=%dx%d blit=%d\n", + tex_width, tex->height, blit_width ); + + /* Flush the pixel cache. 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( 4 ); + + RADEON_FLUSH_CACHE(); + RADEON_WAIT_UNTIL_IDLE(); + + ADVANCE_RING(); + +#ifdef __BIG_ENDIAN + /* The Mesa texture functions provide the data in little endian as the + * chip wants it, but we need to compensate for the fact that the CP + * ring gets byte-swapped + */ + BEGIN_RING( 2 ); + OUT_RING_REG( RADEON_RBBM_GUICNTL, RADEON_HOST_DATA_SWAP_32BIT ); + ADVANCE_RING(); +#endif + + /* Make a copy of the parameters in case we have to update them + * for a multi-pass texture blit. + */ + y = image->y; + height = image->height; + data = (const u8 *)image->data; + + size = height * blit_width; + + if ( size > RADEON_MAX_TEXTURE_SIZE ) { + /* Texture image is too large, do a multipass upload */ + ret = DRM_ERR(EAGAIN); + + /* Adjust the blit size to fit the indirect buffer */ + height = RADEON_MAX_TEXTURE_SIZE / blit_width; + size = height * blit_width; + + /* Update the input parameters for next time */ + image->y += height; + image->height -= height; + image->data = (const char *)image->data + size; + + if ( DRM_COPY_TO_USER( tex->image, image, sizeof(*image) ) ) { + DRM_ERROR( "EFAULT on tex->image\n" ); + return DRM_ERR(EFAULT); + } + } else if ( size < 4 && size > 0 ) { + size = 4; + } + + dwords = size / 4; + + /* Dispatch the indirect buffer. + */ + buffer = (u32 *)((char *)dev_priv->buffers->handle + buf->offset); + + buffer[0] = CP_PACKET3( RADEON_CNTL_HOSTDATA_BLT, dwords + 6 ); + buffer[1] = (RADEON_GMC_DST_PITCH_OFFSET_CNTL | + RADEON_GMC_BRUSH_NONE | + (format << 8) | + RADEON_GMC_SRC_DATATYPE_COLOR | + RADEON_ROP3_S | + RADEON_DP_SRC_SOURCE_HOST_DATA | + RADEON_GMC_CLR_CMP_CNTL_DIS | + RADEON_GMC_WR_MSK_DIS); + + buffer[2] = (tex->pitch << 22) | (tex->offset >> 10); + buffer[3] = 0xffffffff; + buffer[4] = 0xffffffff; + buffer[5] = (y << 16) | image->x; + buffer[6] = (height << 16) | image->width; + buffer[7] = dwords; + + buffer += 8; + + if ( tex_width >= 32 ) { + /* Texture image width is larger than the minimum, so we + * can upload it directly. + */ + if ( DRM_COPY_FROM_USER( buffer, data, dwords * sizeof(u32) ) ) { + DRM_ERROR( "EFAULT on data, %d dwords\n", dwords ); + return DRM_ERR(EFAULT); + } + } else { + /* Texture image width is less than the minimum, so we + * need to pad out each image scanline to the minimum + * width. + */ + for ( i = 0 ; i < tex->height ; i++ ) { + if ( DRM_COPY_FROM_USER( buffer, data, tex_width ) ) { + DRM_ERROR( "EFAULT on pad, %d bytes\n", + tex_width ); + return DRM_ERR(EFAULT); + } + buffer += 8; + data += tex_width; + } + } + + buf->pid = DRM_CURRENTPID; + buf->used = (dwords + 8) * sizeof(u32); + + radeon_cp_dispatch_indirect( dev, buf, 0, buf->used ); + radeon_cp_discard_buffer( dev, buf ); + + /* Flush the pixel cache after the blit completes. This ensures + * the texture data is written out to memory before rendering + * continues. + */ + BEGIN_RING( 4 ); + + RADEON_FLUSH_CACHE(); + RADEON_WAIT_UNTIL_2D_IDLE(); + + ADVANCE_RING(); + + return ret; +} + +static void radeon_cp_dispatch_stipple( drm_device_t *dev, u32 *stipple ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + int i; + RING_LOCALS; + DRM_DEBUG( "\n" ); + + BEGIN_RING( 35 ); + + OUT_RING( CP_PACKET0( RADEON_RE_STIPPLE_ADDR, 0 ) ); + OUT_RING( 0x00000000 ); + + OUT_RING( CP_PACKET0_TABLE( RADEON_RE_STIPPLE_DATA, 31 ) ); + for ( i = 0 ; i < 32 ; i++ ) { + OUT_RING( stipple[i] ); + } + + ADVANCE_RING(); +} + + +/* ================================================================ + * IOCTL functions + */ + +int radeon_cp_clear( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_radeon_clear_t clear; + drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS]; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + DRM_COPY_FROM_USER_IOCTL( clear, (drm_radeon_clear_t *)data, + sizeof(clear) ); + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + + if ( sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS ) + sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS; + + if ( DRM_COPY_FROM_USER( &depth_boxes, clear.depth_boxes, + sarea_priv->nbox * sizeof(depth_boxes[0]) ) ) + return DRM_ERR(EFAULT); + + radeon_cp_dispatch_clear( dev, &clear, depth_boxes ); + + COMMIT_RING(); + return 0; +} + + + +/* Not sure why this isn't set all the time: + */ +static int radeon_do_init_pageflip( drm_device_t *dev ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + RING_LOCALS; + + DRM_DEBUG( "\n" ); + + dev_priv->crtc_offset_cntl = RADEON_READ( RADEON_CRTC_OFFSET_CNTL ); + + BEGIN_RING( 4 ); + RADEON_WAIT_UNTIL_3D_IDLE(); + OUT_RING( CP_PACKET0( RADEON_CRTC_OFFSET_CNTL, 0 ) ); + OUT_RING( dev_priv->crtc_offset_cntl | RADEON_CRTC_OFFSET_FLIP_CNTL ); + ADVANCE_RING(); + + dev_priv->page_flipping = 1; + dev_priv->current_page = 0; + dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page; + + return 0; +} + +int radeon_do_cleanup_pageflip( drm_device_t *dev ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "\n" ); + + if (dev_priv->current_page != 0) + radeon_cp_dispatch_flip( dev ); + + /* FIXME: If the X server changes screen resolution, it + * clobbers the value of RADEON_CRTC_OFFSET_CNTL, above, + * leading to a flashing efect. + */ + dev_priv->page_flipping = 0; + return 0; +} + +/* Swapping and flipping are different operations, need different ioctls. + * They can & should be intermixed to support multiple 3d windows. + */ +int radeon_cp_flip( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + + if (!dev_priv->page_flipping) + radeon_do_init_pageflip( dev ); + + radeon_cp_dispatch_flip( dev ); + + COMMIT_RING(); + return 0; +} + +int radeon_cp_swap( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + DRM_DEBUG( "\n" ); + + LOCK_TEST_WITH_RETURN( dev ); + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + + if ( sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS ) + sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS; + + radeon_cp_dispatch_swap( dev ); + dev_priv->sarea_priv->ctx_owner = 0; + + COMMIT_RING(); + return 0; +} + +int radeon_cp_vertex( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_radeon_vertex_t vertex; + drm_radeon_tcl_prim_t prim; + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( vertex, (drm_radeon_vertex_t *)data, + sizeof(vertex) ); + + DRM_DEBUG( "pid=%d index=%d count=%d discard=%d\n", + DRM_CURRENTPID, + 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 DRM_ERR(EINVAL); + } + if ( vertex.prim < 0 || + vertex.prim > RADEON_PRIM_TYPE_3VRT_LINE_LIST ) { + DRM_ERROR( "buffer prim %d\n", vertex.prim ); + return DRM_ERR(EINVAL); + } + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + buf = dma->buflist[vertex.idx]; + + if ( buf->pid != DRM_CURRENTPID ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + DRM_CURRENTPID, buf->pid ); + return DRM_ERR(EINVAL); + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", vertex.idx ); + return DRM_ERR(EINVAL); + } + + /* Build up a prim_t record: + */ + if (vertex.count) { + buf->used = vertex.count; /* not used? */ + + if ( sarea_priv->dirty & ~RADEON_UPLOAD_CLIPRECTS ) { + radeon_emit_state( dev_priv, + &sarea_priv->context_state, + sarea_priv->tex_state, + sarea_priv->dirty ); + + sarea_priv->dirty &= ~(RADEON_UPLOAD_TEX0IMAGES | + RADEON_UPLOAD_TEX1IMAGES | + RADEON_UPLOAD_TEX2IMAGES | + RADEON_REQUIRE_QUIESCENCE); + } + + prim.start = 0; + prim.finish = vertex.count; /* unused */ + prim.prim = vertex.prim; + prim.numverts = vertex.count; + prim.vc_format = dev_priv->sarea_priv->vc_format; + + radeon_cp_dispatch_vertex( dev, buf, &prim, + dev_priv->sarea_priv->boxes, + dev_priv->sarea_priv->nbox ); + } + + if (vertex.discard) { + radeon_cp_discard_buffer( dev, buf ); + } + + COMMIT_RING(); + return 0; +} + +int radeon_cp_indices( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_radeon_indices_t elts; + drm_radeon_tcl_prim_t prim; + int count; + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( elts, (drm_radeon_indices_t *)data, + sizeof(elts) ); + + DRM_DEBUG( "pid=%d index=%d start=%d end=%d discard=%d\n", + DRM_CURRENTPID, + 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 DRM_ERR(EINVAL); + } + if ( elts.prim < 0 || + elts.prim > RADEON_PRIM_TYPE_3VRT_LINE_LIST ) { + DRM_ERROR( "buffer prim %d\n", elts.prim ); + return DRM_ERR(EINVAL); + } + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + buf = dma->buflist[elts.idx]; + + if ( buf->pid != DRM_CURRENTPID ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + DRM_CURRENTPID, buf->pid ); + return DRM_ERR(EINVAL); + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", elts.idx ); + return DRM_ERR(EINVAL); + } + + count = (elts.end - elts.start) / sizeof(u16); + elts.start -= RADEON_INDEX_PRIM_OFFSET; + + if ( elts.start & 0x7 ) { + DRM_ERROR( "misaligned buffer 0x%x\n", elts.start ); + return DRM_ERR(EINVAL); + } + if ( elts.start < buf->used ) { + DRM_ERROR( "no header 0x%x - 0x%x\n", elts.start, buf->used ); + return DRM_ERR(EINVAL); + } + + buf->used = elts.end; + + if ( sarea_priv->dirty & ~RADEON_UPLOAD_CLIPRECTS ) { + radeon_emit_state( dev_priv, + &sarea_priv->context_state, + sarea_priv->tex_state, + sarea_priv->dirty ); + + sarea_priv->dirty &= ~(RADEON_UPLOAD_TEX0IMAGES | + RADEON_UPLOAD_TEX1IMAGES | + RADEON_UPLOAD_TEX2IMAGES | + RADEON_REQUIRE_QUIESCENCE); + } + + + /* Build up a prim_t record: + */ + prim.start = elts.start; + prim.finish = elts.end; + prim.prim = elts.prim; + prim.offset = 0; /* offset from start of dma buffers */ + prim.numverts = RADEON_MAX_VB_VERTS; /* duh */ + prim.vc_format = dev_priv->sarea_priv->vc_format; + + radeon_cp_dispatch_indices( dev, buf, &prim, + dev_priv->sarea_priv->boxes, + dev_priv->sarea_priv->nbox ); + if (elts.discard) { + radeon_cp_discard_buffer( dev, buf ); + } + + COMMIT_RING(); + return 0; +} + +int radeon_cp_texture( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_texture_t tex; + drm_radeon_tex_image_t image; + int ret; + + LOCK_TEST_WITH_RETURN( dev ); + + DRM_COPY_FROM_USER_IOCTL( tex, (drm_radeon_texture_t *)data, sizeof(tex) ); + + if ( tex.image == NULL ) { + DRM_ERROR( "null texture image!\n" ); + return DRM_ERR(EINVAL); + } + + if ( DRM_COPY_FROM_USER( &image, + (drm_radeon_tex_image_t *)tex.image, + sizeof(image) ) ) + return DRM_ERR(EFAULT); + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + ret = radeon_cp_dispatch_texture( dev, &tex, &image ); + + COMMIT_RING(); + return ret; +} + +int radeon_cp_stipple( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_stipple_t stipple; + u32 mask[32]; + + LOCK_TEST_WITH_RETURN( dev ); + + DRM_COPY_FROM_USER_IOCTL( stipple, (drm_radeon_stipple_t *)data, + sizeof(stipple) ); + + if ( DRM_COPY_FROM_USER( &mask, stipple.mask, 32 * sizeof(u32) ) ) + return DRM_ERR(EFAULT); + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + + radeon_cp_dispatch_stipple( dev, mask ); + + COMMIT_RING(); + return 0; +} + +int radeon_cp_indirect( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_radeon_indirect_t indirect; + RING_LOCALS; + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( indirect, (drm_radeon_indirect_t *)data, + sizeof(indirect) ); + + DRM_DEBUG( "indirect: idx=%d s=%d e=%d d=%d\n", + indirect.idx, indirect.start, + indirect.end, indirect.discard ); + + if ( indirect.idx < 0 || indirect.idx >= dma->buf_count ) { + DRM_ERROR( "buffer index %d (of %d max)\n", + indirect.idx, dma->buf_count - 1 ); + return DRM_ERR(EINVAL); + } + + buf = dma->buflist[indirect.idx]; + + if ( buf->pid != DRM_CURRENTPID ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + DRM_CURRENTPID, buf->pid ); + return DRM_ERR(EINVAL); + } + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", indirect.idx ); + return DRM_ERR(EINVAL); + } + + if ( indirect.start < buf->used ) { + DRM_ERROR( "reusing indirect: start=0x%x actual=0x%x\n", + indirect.start, buf->used ); + return DRM_ERR(EINVAL); + } + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + buf->used = indirect.end; + + /* Wait for the 3D stream to idle before the indirect buffer + * containing 2D acceleration commands is processed. + */ + BEGIN_RING( 2 ); + + RADEON_WAIT_UNTIL_3D_IDLE(); + + ADVANCE_RING(); + + /* Dispatch the indirect buffer full of commands from the + * X server. This is insecure and is thus only available to + * privileged clients. + */ + radeon_cp_dispatch_indirect( dev, buf, indirect.start, indirect.end ); + if (indirect.discard) { + radeon_cp_discard_buffer( dev, buf ); + } + + + COMMIT_RING(); + return 0; +} + +int radeon_cp_vertex2( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf; + drm_radeon_vertex2_t vertex; + int i; + unsigned char laststate; + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( vertex, (drm_radeon_vertex2_t *)data, + sizeof(vertex) ); + + DRM_DEBUG( "pid=%d index=%d discard=%d\n", + DRM_CURRENTPID, + vertex.idx, 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 DRM_ERR(EINVAL); + } + + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + buf = dma->buflist[vertex.idx]; + + if ( buf->pid != DRM_CURRENTPID ) { + DRM_ERROR( "process %d using buffer owned by %d\n", + DRM_CURRENTPID, buf->pid ); + return DRM_ERR(EINVAL); + } + + if ( buf->pending ) { + DRM_ERROR( "sending pending buffer %d\n", vertex.idx ); + return DRM_ERR(EINVAL); + } + + if (sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS) + return DRM_ERR(EINVAL); + + for (laststate = 0xff, i = 0 ; i < vertex.nr_prims ; i++) { + drm_radeon_prim_t prim; + drm_radeon_tcl_prim_t tclprim; + + if ( DRM_COPY_FROM_USER( &prim, &vertex.prim[i], sizeof(prim) ) ) + return DRM_ERR(EFAULT); + + if ( prim.stateidx != laststate ) { + drm_radeon_state_t state; + + if ( DRM_COPY_FROM_USER( &state, + &vertex.state[prim.stateidx], + sizeof(state) ) ) + return DRM_ERR(EFAULT); + + radeon_emit_state2( dev_priv, &state ); + + laststate = prim.stateidx; + } + + tclprim.start = prim.start; + tclprim.finish = prim.finish; + tclprim.prim = prim.prim; + tclprim.vc_format = prim.vc_format; + + if ( prim.prim & RADEON_PRIM_WALK_IND ) { + tclprim.offset = prim.numverts * 64; + tclprim.numverts = RADEON_MAX_VB_VERTS; /* duh */ + + radeon_cp_dispatch_indices( dev, buf, &tclprim, + sarea_priv->boxes, + sarea_priv->nbox); + } else { + tclprim.numverts = prim.numverts; + tclprim.offset = 0; /* not used */ + + radeon_cp_dispatch_vertex( dev, buf, &tclprim, + sarea_priv->boxes, + sarea_priv->nbox); + } + + if (sarea_priv->nbox == 1) + sarea_priv->nbox = 0; + } + + if ( vertex.discard ) { + radeon_cp_discard_buffer( dev, buf ); + } + + COMMIT_RING(); + return 0; +} + + +static int radeon_emit_packets( + drm_radeon_private_t *dev_priv, + drm_radeon_cmd_header_t header, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + int id = (int)header.packet.packet_id; + int sz = packet[id].len; + int reg = packet[id].start; + int *data = (int *)cmdbuf->buf; + RING_LOCALS; + + if (sz * sizeof(int) > cmdbuf->bufsz) + return DRM_ERR(EINVAL); + + BEGIN_RING(sz+1); + OUT_RING( CP_PACKET0( reg, (sz-1) ) ); + OUT_RING_USER_TABLE( data, sz ); + ADVANCE_RING(); + + cmdbuf->buf += sz * sizeof(int); + cmdbuf->bufsz -= sz * sizeof(int); + return 0; +} + +static __inline__ int radeon_emit_scalars( + drm_radeon_private_t *dev_priv, + drm_radeon_cmd_header_t header, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + int sz = header.scalars.count; + int *data = (int *)cmdbuf->buf; + int start = header.scalars.offset; + int stride = header.scalars.stride; + RING_LOCALS; + + BEGIN_RING( 3+sz ); + OUT_RING( CP_PACKET0( RADEON_SE_TCL_SCALAR_INDX_REG, 0 ) ); + OUT_RING( start | (stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); + OUT_RING( CP_PACKET0_TABLE( RADEON_SE_TCL_SCALAR_DATA_REG, sz-1 ) ); + OUT_RING_USER_TABLE( data, sz ); + ADVANCE_RING(); + cmdbuf->buf += sz * sizeof(int); + cmdbuf->bufsz -= sz * sizeof(int); + return 0; +} + +/* God this is ugly + */ +static inline int radeon_emit_scalars2( + drm_radeon_private_t *dev_priv, + drm_radeon_cmd_header_t header, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + int sz = header.scalars.count; + int *data = (int *)cmdbuf->buf; + int start = header.scalars.offset + 0x100; + int stride = header.scalars.stride; + RING_LOCALS; + + BEGIN_RING( 3+sz ); + OUT_RING( CP_PACKET0( RADEON_SE_TCL_SCALAR_INDX_REG, 0 ) ); + OUT_RING( start | (stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); + OUT_RING( CP_PACKET0_TABLE( RADEON_SE_TCL_SCALAR_DATA_REG, sz-1 ) ); + OUT_RING_USER_TABLE( data, sz ); + ADVANCE_RING(); + cmdbuf->buf += sz * sizeof(int); + cmdbuf->bufsz -= sz * sizeof(int); + return 0; +} + +static inline int radeon_emit_vectors( + drm_radeon_private_t *dev_priv, + drm_radeon_cmd_header_t header, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + int sz = header.vectors.count; + int *data = (int *)cmdbuf->buf; + int start = header.vectors.offset; + int stride = header.vectors.stride; + RING_LOCALS; + + BEGIN_RING( 3+sz ); + OUT_RING( CP_PACKET0( RADEON_SE_TCL_VECTOR_INDX_REG, 0 ) ); + OUT_RING( start | (stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); + OUT_RING( CP_PACKET0_TABLE( RADEON_SE_TCL_VECTOR_DATA_REG, (sz-1) ) ); + OUT_RING_USER_TABLE( data, sz ); + ADVANCE_RING(); + + cmdbuf->buf += sz * sizeof(int); + cmdbuf->bufsz -= sz * sizeof(int); + return 0; +} + + +static int radeon_emit_packet3( drm_device_t *dev, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + int cmdsz, tmp; + int *cmd = (int *)cmdbuf->buf; + RING_LOCALS; + + + DRM_DEBUG("\n"); + + if (DRM_GET_USER_UNCHECKED( tmp, &cmd[0])) + return DRM_ERR(EFAULT); + + cmdsz = 2 + ((tmp & RADEON_CP_PACKET_COUNT_MASK) >> 16); + + if ((tmp & 0xc0000000) != RADEON_CP_PACKET3 || + cmdsz * 4 > cmdbuf->bufsz) + return DRM_ERR(EINVAL); + + BEGIN_RING( cmdsz ); + OUT_RING_USER_TABLE( cmd, cmdsz ); + ADVANCE_RING(); + + cmdbuf->buf += cmdsz * 4; + cmdbuf->bufsz -= cmdsz * 4; + return 0; +} + + +static int radeon_emit_packet3_cliprect( drm_device_t *dev, + drm_radeon_cmd_buffer_t *cmdbuf ) +{ + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_clip_rect_t box; + int cmdsz, tmp; + int *cmd = (int *)cmdbuf->buf; + drm_clip_rect_t *boxes = cmdbuf->boxes; + int i = 0; + RING_LOCALS; + + DRM_DEBUG("\n"); + + if (DRM_GET_USER_UNCHECKED( tmp, &cmd[0])) + return DRM_ERR(EFAULT); + + cmdsz = 2 + ((tmp & RADEON_CP_PACKET_COUNT_MASK) >> 16); + + if ((tmp & 0xc0000000) != RADEON_CP_PACKET3 || + cmdsz * 4 > cmdbuf->bufsz) + return DRM_ERR(EINVAL); + + do { + if ( i < cmdbuf->nbox ) { + if (DRM_COPY_FROM_USER_UNCHECKED( &box, &boxes[i], sizeof(box) )) + return DRM_ERR(EFAULT); + radeon_emit_clip_rect( dev_priv, &box ); + } + + BEGIN_RING( cmdsz ); + OUT_RING_USER_TABLE( cmd, cmdsz ); + ADVANCE_RING(); + + } while ( ++i < cmdbuf->nbox ); + + if (cmdbuf->nbox == 1) + cmdbuf->nbox = 0; + + cmdbuf->buf += cmdsz * 4; + cmdbuf->bufsz -= cmdsz * 4; + return 0; +} + + + +int radeon_cp_cmdbuf( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_device_dma_t *dma = dev->dma; + drm_buf_t *buf = 0; + int idx; + drm_radeon_cmd_buffer_t cmdbuf; + drm_radeon_cmd_header_t header; + + LOCK_TEST_WITH_RETURN( dev ); + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( cmdbuf, (drm_radeon_cmd_buffer_t *)data, + sizeof(cmdbuf) ); + + DRM_DEBUG( "pid=%d\n", DRM_CURRENTPID ); + RING_SPACE_TEST_WITH_RETURN( dev_priv ); + VB_AGE_TEST_WITH_RETURN( dev_priv ); + + + if (DRM_VERIFYAREA_READ( cmdbuf.buf, cmdbuf.bufsz )) + return DRM_ERR(EFAULT); + + if (cmdbuf.nbox && + DRM_VERIFYAREA_READ(cmdbuf.boxes, + cmdbuf.nbox * sizeof(drm_clip_rect_t))) + return DRM_ERR(EFAULT); + + while ( cmdbuf.bufsz >= sizeof(header) ) { + + if (DRM_GET_USER_UNCHECKED( header.i, (int *)cmdbuf.buf )) { + DRM_ERROR("__get_user %p\n", cmdbuf.buf); + return DRM_ERR(EFAULT); + } + + cmdbuf.buf += sizeof(header); + cmdbuf.bufsz -= sizeof(header); + + switch (header.header.cmd_type) { + case RADEON_CMD_PACKET: + DRM_DEBUG("RADEON_CMD_PACKET\n"); + if (radeon_emit_packets( dev_priv, header, &cmdbuf )) { + DRM_ERROR("radeon_emit_packets failed\n"); + return DRM_ERR(EINVAL); + } + break; + + case RADEON_CMD_SCALARS: + DRM_DEBUG("RADEON_CMD_SCALARS\n"); + if (radeon_emit_scalars( dev_priv, header, &cmdbuf )) { + DRM_ERROR("radeon_emit_scalars failed\n"); + return DRM_ERR(EINVAL); + } + break; + + case RADEON_CMD_VECTORS: + DRM_DEBUG("RADEON_CMD_VECTORS\n"); + if (radeon_emit_vectors( dev_priv, header, &cmdbuf )) { + DRM_ERROR("radeon_emit_vectors failed\n"); + return DRM_ERR(EINVAL); + } + break; + + case RADEON_CMD_DMA_DISCARD: + DRM_DEBUG("RADEON_CMD_DMA_DISCARD\n"); + idx = header.dma.buf_idx; + if ( idx < 0 || idx >= dma->buf_count ) { + DRM_ERROR( "buffer index %d (of %d max)\n", + idx, dma->buf_count - 1 ); + return DRM_ERR(EINVAL); + } + + buf = dma->buflist[idx]; + if ( buf->pid != DRM_CURRENTPID || buf->pending ) { + DRM_ERROR( "bad buffer\n" ); + return DRM_ERR(EINVAL); + } + + radeon_cp_discard_buffer( dev, buf ); + break; + + case RADEON_CMD_PACKET3: + DRM_DEBUG("RADEON_CMD_PACKET3\n"); + if (radeon_emit_packet3( dev, &cmdbuf )) { + DRM_ERROR("radeon_emit_packet3 failed\n"); + return DRM_ERR(EINVAL); + } + break; + + case RADEON_CMD_PACKET3_CLIP: + DRM_DEBUG("RADEON_CMD_PACKET3_CLIP\n"); + if (radeon_emit_packet3_cliprect( dev, &cmdbuf )) { + DRM_ERROR("radeon_emit_packet3_clip failed\n"); + return DRM_ERR(EINVAL); + } + break; + + case RADEON_CMD_SCALARS2: + DRM_DEBUG("RADEON_CMD_SCALARS2\n"); + if (radeon_emit_scalars2( dev_priv, header, &cmdbuf )) { + DRM_ERROR("radeon_emit_scalars2 failed\n"); + return DRM_ERR(EINVAL); + } + break; + default: + DRM_ERROR("bad cmd_type %d at %p\n", + header.header.cmd_type, + cmdbuf.buf - sizeof(header)); + return DRM_ERR(EINVAL); + } + } + + + DRM_DEBUG("DONE\n"); + COMMIT_RING(); + return 0; +} + + + +int radeon_cp_getparam( DRM_IOCTL_ARGS ) +{ + DRM_DEVICE; + drm_radeon_private_t *dev_priv = dev->dev_private; + drm_radeon_getparam_t param; + int value; + + if ( !dev_priv ) { + DRM_ERROR( "%s called with no initialization\n", __func__ ); + return DRM_ERR(EINVAL); + } + + DRM_COPY_FROM_USER_IOCTL( param, (drm_radeon_getparam_t *)data, + sizeof(param) ); + + DRM_DEBUG( "pid=%d\n", DRM_CURRENTPID ); + + switch( param.param ) { + case RADEON_PARAM_AGP_BUFFER_OFFSET: + value = dev_priv->agp_buffers_offset; + break; + case RADEON_PARAM_LAST_FRAME: + value = DRM_READ32(&dev_priv->scratch[0]); + break; + case RADEON_PARAM_LAST_DISPATCH: + value = DRM_READ32(&dev_priv->scratch[1]); + break; + case RADEON_PARAM_LAST_CLEAR: + value = DRM_READ32(&dev_priv->scratch[2]); + break; + default: + return DRM_ERR(EINVAL); + } + + if ( DRM_COPY_TO_USER( param.value, &value, sizeof(int) ) ) { + DRM_ERROR( "copy_to_user\n" ); + return DRM_ERR(EFAULT); + } + + return 0; +} |