2005-04-17 05:20:36 +07:00
|
|
|
/* 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@tungstengraphics.com>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#ifndef __RADEON_DRM_H__
|
|
|
|
#define __RADEON_DRM_H__
|
|
|
|
|
2015-07-22 15:29:00 +07:00
|
|
|
#include "drm.h"
|
2009-02-26 06:51:42 +07:00
|
|
|
|
2016-04-08 01:13:24 +07:00
|
|
|
#if defined(__cplusplus)
|
|
|
|
extern "C" {
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 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
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_UPLOAD_CLIPRECTS 0x00008000 /* handled client-side */
|
2005-04-17 05:20:36 +07:00
|
|
|
#define RADEON_REQUIRE_QUIESCENCE 0x00010000
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_UPLOAD_ZBIAS 0x00020000 /* version 1.2 and newer */
|
2005-04-17 05:20:36 +07:00
|
|
|
#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:
|
|
|
|
*/
|
2005-09-25 11:28:13 +07:00
|
|
|
#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 */
|
2005-04-17 05:20:36 +07:00
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_0 61
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_0 62
|
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_1 63
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_1 64
|
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_2 65
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_2 66
|
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_3 67
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_3 68
|
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_4 69
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_4 70
|
|
|
|
#define R200_EMIT_PP_CUBIC_FACES_5 71
|
|
|
|
#define R200_EMIT_PP_CUBIC_OFFSETS_5 72
|
|
|
|
#define RADEON_EMIT_PP_TEX_SIZE_0 73
|
|
|
|
#define RADEON_EMIT_PP_TEX_SIZE_1 74
|
|
|
|
#define RADEON_EMIT_PP_TEX_SIZE_2 75
|
|
|
|
#define R200_EMIT_RB3D_BLENDCOLOR 76
|
|
|
|
#define R200_EMIT_TCL_POINT_SPRITE_CNTL 77
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_FACES_0 78
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_OFFSETS_T0 79
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_FACES_1 80
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_OFFSETS_T1 81
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_FACES_2 82
|
|
|
|
#define RADEON_EMIT_PP_CUBIC_OFFSETS_T2 83
|
|
|
|
#define R200_EMIT_PP_TRI_PERF_CNTL 84
|
2005-09-11 16:55:53 +07:00
|
|
|
#define R200_EMIT_PP_AFS_0 85
|
|
|
|
#define R200_EMIT_PP_AFS_1 86
|
|
|
|
#define R200_EMIT_ATF_TFACTOR 87
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_0 88
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_1 89
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_2 90
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_3 91
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_4 92
|
|
|
|
#define R200_EMIT_PP_TXCTLALL_5 93
|
2006-06-24 14:04:07 +07:00
|
|
|
#define R200_EMIT_VAP_PVS_CNTL 94
|
|
|
|
#define RADEON_MAX_STATE_PACKETS 95
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Commands understood by cmd_buffer ioctl. More can be added but
|
|
|
|
* obviously these can't be removed or changed:
|
|
|
|
*/
|
2005-09-25 11:28:13 +07:00
|
|
|
#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 */
|
|
|
|
#define RADEON_CMD_WAIT 8 /* emit hw wait commands -- note:
|
|
|
|
* doesn't make the cpu wait, just
|
|
|
|
* the graphics hardware */
|
2006-06-24 14:04:07 +07:00
|
|
|
#define RADEON_CMD_VECLINEAR 9 /* another r200 stopgap */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
typedef union {
|
|
|
|
int i;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned char cmd_type, pad0, pad1, pad2;
|
|
|
|
} header;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned char cmd_type, packet_id, pad0, pad1;
|
|
|
|
} packet;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, offset, stride, count;
|
2005-04-17 05:20:36 +07:00
|
|
|
} scalars;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, offset, stride, count;
|
2005-04-17 05:20:36 +07:00
|
|
|
} vectors;
|
2006-06-24 14:04:07 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, addr_lo, addr_hi, count;
|
|
|
|
} veclinear;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, buf_idx, pad0, pad1;
|
2005-04-17 05:20:36 +07:00
|
|
|
} dma;
|
2005-09-25 11:28:13 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, flags, pad0, pad1;
|
2005-04-17 05:20:36 +07:00
|
|
|
} wait;
|
|
|
|
} drm_radeon_cmd_header_t;
|
|
|
|
|
|
|
|
#define RADEON_WAIT_2D 0x1
|
|
|
|
#define RADEON_WAIT_3D 0x2
|
|
|
|
|
2005-08-16 17:43:16 +07:00
|
|
|
/* Allowed parameters for R300_CMD_PACKET3
|
|
|
|
*/
|
|
|
|
#define R300_CMD_PACKET3_CLEAR 0
|
|
|
|
#define R300_CMD_PACKET3_RAW 1
|
|
|
|
|
|
|
|
/* Commands understood by cmd_buffer ioctl for R300.
|
|
|
|
* The interface has not been stabilized, so some of these may be removed
|
|
|
|
* and eventually reordered before stabilization.
|
|
|
|
*/
|
2005-09-25 11:28:13 +07:00
|
|
|
#define R300_CMD_PACKET0 1
|
|
|
|
#define R300_CMD_VPU 2 /* emit vertex program upload */
|
|
|
|
#define R300_CMD_PACKET3 3 /* emit a packet3 */
|
|
|
|
#define R300_CMD_END3D 4 /* emit sequence ending 3d rendering */
|
2005-08-16 17:43:16 +07:00
|
|
|
#define R300_CMD_CP_DELAY 5
|
|
|
|
#define R300_CMD_DMA_DISCARD 6
|
|
|
|
#define R300_CMD_WAIT 7
|
2007-11-05 09:50:58 +07:00
|
|
|
# define R300_WAIT_2D 0x1
|
|
|
|
# define R300_WAIT_3D 0x2
|
2008-03-30 04:51:49 +07:00
|
|
|
/* these two defines are DOING IT WRONG - however
|
|
|
|
* we have userspace which relies on using these.
|
|
|
|
* The wait interface is backwards compat new
|
|
|
|
* code should use the NEW_WAIT defines below
|
|
|
|
* THESE ARE NOT BIT FIELDS
|
|
|
|
*/
|
2007-11-05 09:50:58 +07:00
|
|
|
# define R300_WAIT_2D_CLEAN 0x3
|
|
|
|
# define R300_WAIT_3D_CLEAN 0x4
|
2008-03-30 04:51:49 +07:00
|
|
|
|
|
|
|
# define R300_NEW_WAIT_2D_3D 0x3
|
|
|
|
# define R300_NEW_WAIT_2D_2D_CLEAN 0x4
|
|
|
|
# define R300_NEW_WAIT_3D_3D_CLEAN 0x6
|
|
|
|
# define R300_NEW_WAIT_2D_2D_CLEAN_3D_3D_CLEAN 0x8
|
|
|
|
|
2006-03-19 15:45:26 +07:00
|
|
|
#define R300_CMD_SCRATCH 8
|
2008-05-28 10:52:28 +07:00
|
|
|
#define R300_CMD_R500FP 9
|
2005-08-16 17:43:16 +07:00
|
|
|
|
|
|
|
typedef union {
|
|
|
|
unsigned int u;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, pad0, pad1, pad2;
|
|
|
|
} header;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, count, reglo, reghi;
|
|
|
|
} packet0;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, count, adrlo, adrhi;
|
|
|
|
} vpu;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, packet, pad0, pad1;
|
|
|
|
} packet3;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, packet;
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned short count; /* amount of packet2 to emit */
|
2005-08-16 17:43:16 +07:00
|
|
|
} delay;
|
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, buf_idx, pad0, pad1;
|
|
|
|
} dma;
|
|
|
|
struct {
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned char cmd_type, flags, pad0, pad1;
|
2005-08-16 17:43:16 +07:00
|
|
|
} wait;
|
2006-03-19 15:45:26 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, reg, n_bufs, flags;
|
|
|
|
} scratch;
|
2008-05-28 10:52:28 +07:00
|
|
|
struct {
|
|
|
|
unsigned char cmd_type, count, adrlo, adrhi_flags;
|
|
|
|
} r500fp;
|
2005-08-16 17:43:16 +07:00
|
|
|
} drm_r300_cmd_header_t;
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#define RADEON_FRONT 0x1
|
|
|
|
#define RADEON_BACK 0x2
|
|
|
|
#define RADEON_DEPTH 0x4
|
|
|
|
#define RADEON_STENCIL 0x8
|
|
|
|
#define RADEON_CLEAR_FASTZ 0x80000000
|
|
|
|
#define RADEON_USE_HIERZ 0x40000000
|
|
|
|
#define RADEON_USE_COMP_ZBUF 0x20000000
|
|
|
|
|
2008-05-28 10:52:28 +07:00
|
|
|
#define R500FP_CONSTANT_TYPE (1 << 1)
|
|
|
|
#define R500FP_CONSTANT_CLAMP (1 << 2)
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 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
|
|
|
|
|
2009-02-25 02:02:13 +07:00
|
|
|
#define R600_SCRATCH_REG_OFFSET 256
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#define RADEON_NR_SAREA_CLIPRECTS 12
|
|
|
|
|
|
|
|
/* There are 2 heaps (local/GART). 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_GART_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
|
|
|
|
|
|
|
|
#define RADEON_MAX_SURFACES 8
|
|
|
|
|
|
|
|
/* Blits have strict offset rules. All blit offset must be aligned on
|
|
|
|
* a 1K-byte boundary.
|
|
|
|
*/
|
|
|
|
#define RADEON_OFFSET_SHIFT 10
|
|
|
|
#define RADEON_OFFSET_ALIGN (1 << RADEON_OFFSET_SHIFT)
|
|
|
|
#define RADEON_OFFSET_MASK (RADEON_OFFSET_ALIGN - 1)
|
|
|
|
|
2005-09-25 11:28:13 +07:00
|
|
|
#endif /* __RADEON_SAREA_DEFINES__ */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
unsigned int red;
|
|
|
|
unsigned int green;
|
|
|
|
unsigned int blue;
|
|
|
|
unsigned int alpha;
|
|
|
|
} radeon_color_regs_t;
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
/* Context state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int pp_misc; /* 0x1c14 */
|
2005-04-17 05:20:36 +07:00
|
|
|
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;
|
|
|
|
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int pp_cntl; /* 0x1c38 */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int rb3d_cntl;
|
|
|
|
unsigned int rb3d_coloroffset;
|
|
|
|
unsigned int re_width_height;
|
|
|
|
unsigned int rb3d_colorpitch;
|
|
|
|
unsigned int se_cntl;
|
|
|
|
|
|
|
|
/* Vertex format state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int se_coord_fmt; /* 0x1c50 */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Line state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int re_line_pattern; /* 0x1cd0 */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int re_line_state;
|
|
|
|
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int se_line_width; /* 0x1db8 */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Bumpmap state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int pp_lum_matrix; /* 0x1d00 */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int pp_rot_matrix_0; /* 0x1d58 */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int pp_rot_matrix_1;
|
|
|
|
|
|
|
|
/* Mask state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int rb3d_stencilrefmask; /* 0x1d7c */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int rb3d_ropcntl;
|
|
|
|
unsigned int rb3d_planemask;
|
|
|
|
|
|
|
|
/* Viewport state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int se_vport_xscale; /* 0x1d98 */
|
2005-04-17 05:20:36 +07:00
|
|
|
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 */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int se_cntl_status; /* 0x2140 */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Misc state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int re_top_left; /* 0x26c0 */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int re_misc;
|
|
|
|
} drm_radeon_context_regs_t;
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
/* Zbias state */
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int se_zbias_factor; /* 0x1dac */
|
2005-04-17 05:20:36 +07:00
|
|
|
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;
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int numverts:16; /* overloaded as offset/64 for elt prims */
|
|
|
|
unsigned int vc_format; /* vertex format */
|
2005-04-17 05:20:36 +07:00
|
|
|
} 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 {
|
|
|
|
/* 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.
|
|
|
|
*/
|
2007-07-11 12:27:12 +07:00
|
|
|
struct drm_clip_rect boxes[RADEON_NR_SAREA_CLIPRECTS];
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int nbox;
|
|
|
|
|
|
|
|
/* Counters for client-side throttling of rendering clients.
|
|
|
|
*/
|
|
|
|
unsigned int last_frame;
|
|
|
|
unsigned int last_dispatch;
|
|
|
|
unsigned int last_clear;
|
|
|
|
|
2007-07-11 12:27:12 +07:00
|
|
|
struct drm_tex_region tex_list[RADEON_NR_TEX_HEAPS][RADEON_NR_TEX_REGIONS +
|
2005-09-25 11:28:13 +07:00
|
|
|
1];
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int tex_age[RADEON_NR_TEX_HEAPS];
|
|
|
|
int ctx_owner;
|
2005-09-25 11:28:13 +07:00
|
|
|
int pfState; /* number of 3d windows (0,1,2ormore) */
|
|
|
|
int pfCurrentPage; /* which buffer is being displayed? */
|
|
|
|
int crtc2_base; /* CRTC2 frame offset */
|
2005-04-17 05:20:36 +07:00
|
|
|
int tiling_enabled; /* set by drm, read by 2d + 3d clients */
|
|
|
|
} 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.
|
|
|
|
*/
|
2005-09-25 11:28:13 +07:00
|
|
|
#define DRM_RADEON_CP_INIT 0x00
|
|
|
|
#define DRM_RADEON_CP_START 0x01
|
2005-04-17 05:20:36 +07:00
|
|
|
#define DRM_RADEON_CP_STOP 0x02
|
|
|
|
#define DRM_RADEON_CP_RESET 0x03
|
|
|
|
#define DRM_RADEON_CP_IDLE 0x04
|
2005-09-25 11:28:13 +07:00
|
|
|
#define DRM_RADEON_RESET 0x05
|
2005-04-17 05:20:36 +07:00
|
|
|
#define DRM_RADEON_FULLSCREEN 0x06
|
2005-09-25 11:28:13 +07:00
|
|
|
#define DRM_RADEON_SWAP 0x07
|
|
|
|
#define DRM_RADEON_CLEAR 0x08
|
2005-04-17 05:20:36 +07:00
|
|
|
#define DRM_RADEON_VERTEX 0x09
|
|
|
|
#define DRM_RADEON_INDICES 0x0A
|
|
|
|
#define DRM_RADEON_NOT_USED
|
|
|
|
#define DRM_RADEON_STIPPLE 0x0C
|
|
|
|
#define DRM_RADEON_INDIRECT 0x0D
|
|
|
|
#define DRM_RADEON_TEXTURE 0x0E
|
|
|
|
#define DRM_RADEON_VERTEX2 0x0F
|
|
|
|
#define DRM_RADEON_CMDBUF 0x10
|
|
|
|
#define DRM_RADEON_GETPARAM 0x11
|
|
|
|
#define DRM_RADEON_FLIP 0x12
|
|
|
|
#define DRM_RADEON_ALLOC 0x13
|
|
|
|
#define DRM_RADEON_FREE 0x14
|
|
|
|
#define DRM_RADEON_INIT_HEAP 0x15
|
|
|
|
#define DRM_RADEON_IRQ_EMIT 0x16
|
|
|
|
#define DRM_RADEON_IRQ_WAIT 0x17
|
|
|
|
#define DRM_RADEON_CP_RESUME 0x18
|
|
|
|
#define DRM_RADEON_SETPARAM 0x19
|
|
|
|
#define DRM_RADEON_SURF_ALLOC 0x1a
|
|
|
|
#define DRM_RADEON_SURF_FREE 0x1b
|
2009-06-05 19:42:42 +07:00
|
|
|
/* KMS ioctl */
|
|
|
|
#define DRM_RADEON_GEM_INFO 0x1c
|
|
|
|
#define DRM_RADEON_GEM_CREATE 0x1d
|
|
|
|
#define DRM_RADEON_GEM_MMAP 0x1e
|
|
|
|
#define DRM_RADEON_GEM_PREAD 0x21
|
|
|
|
#define DRM_RADEON_GEM_PWRITE 0x22
|
|
|
|
#define DRM_RADEON_GEM_SET_DOMAIN 0x23
|
|
|
|
#define DRM_RADEON_GEM_WAIT_IDLE 0x24
|
|
|
|
#define DRM_RADEON_CS 0x26
|
|
|
|
#define DRM_RADEON_INFO 0x27
|
2009-06-24 06:48:08 +07:00
|
|
|
#define DRM_RADEON_GEM_SET_TILING 0x28
|
|
|
|
#define DRM_RADEON_GEM_GET_TILING 0x29
|
2009-08-21 06:47:45 +07:00
|
|
|
#define DRM_RADEON_GEM_BUSY 0x2a
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
#define DRM_RADEON_GEM_VA 0x2b
|
2014-03-02 06:56:17 +07:00
|
|
|
#define DRM_RADEON_GEM_OP 0x2c
|
drm/radeon: add userptr support v8
This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.
It imposes several restrictions upon the memory being mapped:
1. It must be page aligned (both start/end addresses, i.e ptr and size).
2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).
3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.
4. The BO is only mapped readonly for now, so no write support.
5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.
Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.
v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2014-08-07 14:36:00 +07:00
|
|
|
#define DRM_RADEON_GEM_USERPTR 0x2d
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#define DRM_IOCTL_RADEON_CP_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_INIT, drm_radeon_init_t)
|
|
|
|
#define DRM_IOCTL_RADEON_CP_START DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_START)
|
|
|
|
#define DRM_IOCTL_RADEON_CP_STOP DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_STOP, drm_radeon_cp_stop_t)
|
|
|
|
#define DRM_IOCTL_RADEON_CP_RESET DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_RESET)
|
|
|
|
#define DRM_IOCTL_RADEON_CP_IDLE DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_IDLE)
|
|
|
|
#define DRM_IOCTL_RADEON_RESET DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_RESET)
|
|
|
|
#define DRM_IOCTL_RADEON_FULLSCREEN DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_FULLSCREEN, drm_radeon_fullscreen_t)
|
|
|
|
#define DRM_IOCTL_RADEON_SWAP DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_SWAP)
|
|
|
|
#define DRM_IOCTL_RADEON_CLEAR DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CLEAR, drm_radeon_clear_t)
|
|
|
|
#define DRM_IOCTL_RADEON_VERTEX DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_VERTEX, drm_radeon_vertex_t)
|
|
|
|
#define DRM_IOCTL_RADEON_INDICES DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_INDICES, drm_radeon_indices_t)
|
|
|
|
#define DRM_IOCTL_RADEON_STIPPLE DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_STIPPLE, drm_radeon_stipple_t)
|
|
|
|
#define DRM_IOCTL_RADEON_INDIRECT DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_INDIRECT, drm_radeon_indirect_t)
|
|
|
|
#define DRM_IOCTL_RADEON_TEXTURE DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_TEXTURE, drm_radeon_texture_t)
|
|
|
|
#define DRM_IOCTL_RADEON_VERTEX2 DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_VERTEX2, drm_radeon_vertex2_t)
|
|
|
|
#define DRM_IOCTL_RADEON_CMDBUF DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CMDBUF, drm_radeon_cmd_buffer_t)
|
|
|
|
#define DRM_IOCTL_RADEON_GETPARAM DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GETPARAM, drm_radeon_getparam_t)
|
|
|
|
#define DRM_IOCTL_RADEON_FLIP DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_FLIP)
|
|
|
|
#define DRM_IOCTL_RADEON_ALLOC DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_ALLOC, drm_radeon_mem_alloc_t)
|
|
|
|
#define DRM_IOCTL_RADEON_FREE DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_FREE, drm_radeon_mem_free_t)
|
|
|
|
#define DRM_IOCTL_RADEON_INIT_HEAP DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_INIT_HEAP, drm_radeon_mem_init_heap_t)
|
|
|
|
#define DRM_IOCTL_RADEON_IRQ_EMIT DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_IRQ_EMIT, drm_radeon_irq_emit_t)
|
|
|
|
#define DRM_IOCTL_RADEON_IRQ_WAIT DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_IRQ_WAIT, drm_radeon_irq_wait_t)
|
|
|
|
#define DRM_IOCTL_RADEON_CP_RESUME DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_RESUME)
|
|
|
|
#define DRM_IOCTL_RADEON_SETPARAM DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SETPARAM, drm_radeon_setparam_t)
|
|
|
|
#define DRM_IOCTL_RADEON_SURF_ALLOC DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SURF_ALLOC, drm_radeon_surface_alloc_t)
|
|
|
|
#define DRM_IOCTL_RADEON_SURF_FREE DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_SURF_FREE, drm_radeon_surface_free_t)
|
2009-06-05 19:42:42 +07:00
|
|
|
/* KMS */
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_INFO DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_INFO, struct drm_radeon_gem_info)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_CREATE DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_CREATE, struct drm_radeon_gem_create)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_MMAP DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_MMAP, struct drm_radeon_gem_mmap)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_PREAD DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_PREAD, struct drm_radeon_gem_pread)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_PWRITE DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_PWRITE, struct drm_radeon_gem_pwrite)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_SET_DOMAIN DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_SET_DOMAIN, struct drm_radeon_gem_set_domain)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_WAIT_IDLE DRM_IOW(DRM_COMMAND_BASE + DRM_RADEON_GEM_WAIT_IDLE, struct drm_radeon_gem_wait_idle)
|
|
|
|
#define DRM_IOCTL_RADEON_CS DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_CS, struct drm_radeon_cs)
|
|
|
|
#define DRM_IOCTL_RADEON_INFO DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_INFO, struct drm_radeon_info)
|
2010-08-14 17:20:34 +07:00
|
|
|
#define DRM_IOCTL_RADEON_GEM_SET_TILING DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_SET_TILING, struct drm_radeon_gem_set_tiling)
|
|
|
|
#define DRM_IOCTL_RADEON_GEM_GET_TILING DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_GET_TILING, struct drm_radeon_gem_get_tiling)
|
2009-08-21 06:47:45 +07:00
|
|
|
#define DRM_IOCTL_RADEON_GEM_BUSY DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_BUSY, struct drm_radeon_gem_busy)
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
#define DRM_IOCTL_RADEON_GEM_VA DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_VA, struct drm_radeon_gem_va)
|
2014-03-02 06:56:17 +07:00
|
|
|
#define DRM_IOCTL_RADEON_GEM_OP DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_OP, struct drm_radeon_gem_op)
|
drm/radeon: add userptr support v8
This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.
It imposes several restrictions upon the memory being mapped:
1. It must be page aligned (both start/end addresses, i.e ptr and size).
2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).
3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.
4. The BO is only mapped readonly for now, so no write support.
5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.
Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.
v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2014-08-07 14:36:00 +07:00
|
|
|
#define DRM_IOCTL_RADEON_GEM_USERPTR DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_USERPTR, struct drm_radeon_gem_userptr)
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
typedef struct drm_radeon_init {
|
|
|
|
enum {
|
2005-09-25 11:28:13 +07:00
|
|
|
RADEON_INIT_CP = 0x01,
|
2005-04-17 05:20:36 +07:00
|
|
|
RADEON_CLEANUP_CP = 0x02,
|
|
|
|
RADEON_INIT_R200_CP = 0x03,
|
2009-02-25 02:02:13 +07:00
|
|
|
RADEON_INIT_R300_CP = 0x04,
|
|
|
|
RADEON_INIT_R600_CP = 0x05
|
2005-04-17 05:20:36 +07:00
|
|
|
} func;
|
|
|
|
unsigned long sarea_priv_offset;
|
|
|
|
int is_pci;
|
|
|
|
int cp_mode;
|
|
|
|
int gart_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 gart_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 {
|
2005-09-25 11:28:13 +07:00
|
|
|
RADEON_INIT_FULLSCREEN = 0x01,
|
2005-04-17 05:20:36 +07:00
|
|
|
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;
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int depth_mask; /* misnamed field: should be stencil */
|
2005-04-17 05:20:36 +07:00
|
|
|
drm_radeon_clear_rect_t __user *depth_boxes;
|
|
|
|
} drm_radeon_clear_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_vertex {
|
|
|
|
int prim;
|
2005-09-25 11:28:13 +07:00
|
|
|
int idx; /* Index of vertex buffer */
|
|
|
|
int count; /* Number of vertices in buffer */
|
|
|
|
int discard; /* Client finished with buffer? */
|
2005-04-17 05:20:36 +07:00
|
|
|
} drm_radeon_vertex_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_indices {
|
|
|
|
int prim;
|
|
|
|
int idx;
|
|
|
|
int start;
|
|
|
|
int end;
|
2005-09-25 11:28:13 +07:00
|
|
|
int discard; /* Client finished with buffer? */
|
2005-04-17 05:20:36 +07:00
|
|
|
} 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 {
|
2005-09-25 11:28:13 +07:00
|
|
|
int idx; /* Index of vertex buffer */
|
|
|
|
int discard; /* Client finished with buffer? */
|
2005-04-17 05:20:36 +07:00
|
|
|
int nr_states;
|
|
|
|
drm_radeon_state_t __user *state;
|
|
|
|
int nr_prims;
|
|
|
|
drm_radeon_prim_t __user *prim;
|
|
|
|
} drm_radeon_vertex2_t;
|
|
|
|
|
|
|
|
/* v1.3 - obsoletes drm_radeon_vertex2
|
2011-03-31 08:57:33 +07:00
|
|
|
* - allows arbitrarily large cliprect list
|
2005-04-17 05:20:36 +07:00
|
|
|
* - allows updating of tcl packet, vector and scalar state
|
|
|
|
* - allows memory-efficient description of state updates
|
2005-09-25 11:28:13 +07:00
|
|
|
* - allows state to be emitted without a primitive
|
2005-04-17 05:20:36 +07:00
|
|
|
* (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 __user *buf;
|
|
|
|
int nbox;
|
2007-07-11 12:27:12 +07:00
|
|
|
struct drm_clip_rect __user *boxes;
|
2005-04-17 05:20:36 +07:00
|
|
|
} drm_radeon_cmd_buffer_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_tex_image {
|
2005-09-25 11:28:13 +07:00
|
|
|
unsigned int x, y; /* Blit coordinates */
|
2005-04-17 05:20:36 +07:00
|
|
|
unsigned int width, height;
|
|
|
|
const void __user *data;
|
|
|
|
} drm_radeon_tex_image_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_texture {
|
|
|
|
unsigned int offset;
|
|
|
|
int pitch;
|
|
|
|
int format;
|
2005-09-25 11:28:13 +07:00
|
|
|
int width; /* Texture image coordinates */
|
2005-04-17 05:20:36 +07:00
|
|
|
int height;
|
|
|
|
drm_radeon_tex_image_t __user *image;
|
|
|
|
} drm_radeon_texture_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_stipple {
|
|
|
|
unsigned int __user *mask;
|
|
|
|
} drm_radeon_stipple_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_indirect {
|
|
|
|
int idx;
|
|
|
|
int start;
|
|
|
|
int end;
|
|
|
|
int discard;
|
|
|
|
} drm_radeon_indirect_t;
|
|
|
|
|
2006-01-02 17:32:48 +07:00
|
|
|
/* enum for card type parameters */
|
|
|
|
#define RADEON_CARD_PCI 0
|
|
|
|
#define RADEON_CARD_AGP 1
|
|
|
|
#define RADEON_CARD_PCIE 2
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 1.3: An ioctl to get parameters that aren't available to the 3d
|
2005-09-25 11:28:13 +07:00
|
|
|
* client any other way.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_PARAM_GART_BUFFER_OFFSET 1 /* card offset of 1st GART buffer */
|
2005-04-17 05:20:36 +07:00
|
|
|
#define RADEON_PARAM_LAST_FRAME 2
|
|
|
|
#define RADEON_PARAM_LAST_DISPATCH 3
|
|
|
|
#define RADEON_PARAM_LAST_CLEAR 4
|
|
|
|
/* Added with DRM version 1.6. */
|
|
|
|
#define RADEON_PARAM_IRQ_NR 5
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_PARAM_GART_BASE 6 /* card offset of GART base */
|
2005-04-17 05:20:36 +07:00
|
|
|
/* Added with DRM version 1.8. */
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_PARAM_REGISTER_HANDLE 7 /* for drmMap() */
|
2005-04-17 05:20:36 +07:00
|
|
|
#define RADEON_PARAM_STATUS_HANDLE 8
|
|
|
|
#define RADEON_PARAM_SAREA_HANDLE 9
|
|
|
|
#define RADEON_PARAM_GART_TEX_HANDLE 10
|
|
|
|
#define RADEON_PARAM_SCRATCH_OFFSET 11
|
2006-01-02 17:32:48 +07:00
|
|
|
#define RADEON_PARAM_CARD_TYPE 12
|
2007-07-11 09:16:01 +07:00
|
|
|
#define RADEON_PARAM_VBLANK_CRTC 13 /* VBLANK CRTC */
|
2008-02-07 12:01:05 +07:00
|
|
|
#define RADEON_PARAM_FB_LOCATION 14 /* FB location */
|
2008-05-28 08:57:40 +07:00
|
|
|
#define RADEON_PARAM_NUM_GB_PIPES 15 /* num GB pipes */
|
2009-06-05 19:42:42 +07:00
|
|
|
#define RADEON_PARAM_DEVICE_ID 16
|
2009-08-20 06:11:39 +07:00
|
|
|
#define RADEON_PARAM_NUM_Z_PIPES 17 /* num Z pipes */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
typedef struct drm_radeon_getparam {
|
|
|
|
int param;
|
|
|
|
void __user *value;
|
|
|
|
} drm_radeon_getparam_t;
|
|
|
|
|
|
|
|
/* 1.6: Set up a memory manager for regions of shared memory:
|
|
|
|
*/
|
|
|
|
#define RADEON_MEM_REGION_GART 1
|
|
|
|
#define RADEON_MEM_REGION_FB 2
|
|
|
|
|
|
|
|
typedef struct drm_radeon_mem_alloc {
|
|
|
|
int region;
|
|
|
|
int alignment;
|
|
|
|
int size;
|
|
|
|
int __user *region_offset; /* offset from start of fb or GART */
|
|
|
|
} drm_radeon_mem_alloc_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_mem_free {
|
|
|
|
int region;
|
|
|
|
int region_offset;
|
|
|
|
} drm_radeon_mem_free_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_mem_init_heap {
|
|
|
|
int region;
|
|
|
|
int size;
|
2005-09-25 11:28:13 +07:00
|
|
|
int start;
|
2005-04-17 05:20:36 +07:00
|
|
|
} drm_radeon_mem_init_heap_t;
|
|
|
|
|
|
|
|
/* 1.6: Userspace can request & wait on irq's:
|
|
|
|
*/
|
|
|
|
typedef struct drm_radeon_irq_emit {
|
|
|
|
int __user *irq_seq;
|
|
|
|
} drm_radeon_irq_emit_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_irq_wait {
|
|
|
|
int irq_seq;
|
|
|
|
} drm_radeon_irq_wait_t;
|
|
|
|
|
|
|
|
/* 1.10: Clients tell the DRM where they think the framebuffer is located in
|
|
|
|
* the card's address space, via a new generic ioctl to set parameters
|
|
|
|
*/
|
|
|
|
|
|
|
|
typedef struct drm_radeon_setparam {
|
|
|
|
unsigned int param;
|
2009-02-26 06:51:42 +07:00
|
|
|
__s64 value;
|
2005-04-17 05:20:36 +07:00
|
|
|
} drm_radeon_setparam_t;
|
|
|
|
|
|
|
|
#define RADEON_SETPARAM_FB_LOCATION 1 /* determined framebuffer location */
|
|
|
|
#define RADEON_SETPARAM_SWITCH_TILING 2 /* enable/disable color tiling */
|
2005-09-25 11:28:13 +07:00
|
|
|
#define RADEON_SETPARAM_PCIGART_LOCATION 3 /* PCI Gart Location */
|
2006-03-19 15:37:55 +07:00
|
|
|
#define RADEON_SETPARAM_NEW_MEMMAP 4 /* Use new memory map */
|
2007-05-08 12:19:23 +07:00
|
|
|
#define RADEON_SETPARAM_PCIGART_TABLE_SIZE 5 /* PCI GART Table Size */
|
2007-07-11 09:16:01 +07:00
|
|
|
#define RADEON_SETPARAM_VBLANK_CRTC 6 /* VBLANK CRTC */
|
2005-04-17 05:20:36 +07:00
|
|
|
/* 1.14: Clients can allocate/free a surface
|
|
|
|
*/
|
|
|
|
typedef struct drm_radeon_surface_alloc {
|
|
|
|
unsigned int address;
|
|
|
|
unsigned int size;
|
|
|
|
unsigned int flags;
|
|
|
|
} drm_radeon_surface_alloc_t;
|
|
|
|
|
|
|
|
typedef struct drm_radeon_surface_free {
|
|
|
|
unsigned int address;
|
|
|
|
} drm_radeon_surface_free_t;
|
|
|
|
|
2007-11-05 09:50:58 +07:00
|
|
|
#define DRM_RADEON_VBLANK_CRTC1 1
|
|
|
|
#define DRM_RADEON_VBLANK_CRTC2 2
|
2007-07-11 09:16:01 +07:00
|
|
|
|
2009-06-05 19:42:42 +07:00
|
|
|
/*
|
|
|
|
* Kernel modesetting world below.
|
|
|
|
*/
|
|
|
|
#define RADEON_GEM_DOMAIN_CPU 0x1
|
|
|
|
#define RADEON_GEM_DOMAIN_GTT 0x2
|
|
|
|
#define RADEON_GEM_DOMAIN_VRAM 0x4
|
|
|
|
|
|
|
|
struct drm_radeon_gem_info {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 gart_size;
|
|
|
|
__u64 vram_size;
|
|
|
|
__u64 vram_visible;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
2014-07-17 17:01:07 +07:00
|
|
|
#define RADEON_GEM_NO_BACKING_STORE (1 << 0)
|
|
|
|
#define RADEON_GEM_GTT_UC (1 << 1)
|
|
|
|
#define RADEON_GEM_GTT_WC (1 << 2)
|
2014-08-28 13:56:00 +07:00
|
|
|
/* BO is expected to be accessed by the CPU */
|
|
|
|
#define RADEON_GEM_CPU_ACCESS (1 << 3)
|
2014-08-28 21:59:05 +07:00
|
|
|
/* CPU access is not expected to work for this BO */
|
|
|
|
#define RADEON_GEM_NO_CPU_ACCESS (1 << 4)
|
2009-06-05 19:42:42 +07:00
|
|
|
|
|
|
|
struct drm_radeon_gem_create {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 size;
|
|
|
|
__u64 alignment;
|
|
|
|
__u32 handle;
|
|
|
|
__u32 initial_domain;
|
|
|
|
__u32 flags;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
drm/radeon: add userptr support v8
This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.
It imposes several restrictions upon the memory being mapped:
1. It must be page aligned (both start/end addresses, i.e ptr and size).
2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).
3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.
4. The BO is only mapped readonly for now, so no write support.
5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.
Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.
v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2014-08-07 14:36:00 +07:00
|
|
|
/*
|
|
|
|
* This is not a reliable API and you should expect it to fail for any
|
|
|
|
* number of reasons and have fallback path that do not use userptr to
|
|
|
|
* perform any operation.
|
|
|
|
*/
|
|
|
|
#define RADEON_GEM_USERPTR_READONLY (1 << 0)
|
2014-08-07 14:36:01 +07:00
|
|
|
#define RADEON_GEM_USERPTR_ANONONLY (1 << 1)
|
2014-08-07 14:36:02 +07:00
|
|
|
#define RADEON_GEM_USERPTR_VALIDATE (1 << 2)
|
2014-08-07 14:36:03 +07:00
|
|
|
#define RADEON_GEM_USERPTR_REGISTER (1 << 3)
|
drm/radeon: add userptr support v8
This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.
It imposes several restrictions upon the memory being mapped:
1. It must be page aligned (both start/end addresses, i.e ptr and size).
2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).
3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.
4. The BO is only mapped readonly for now, so no write support.
5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.
Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.
v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2014-08-07 14:36:00 +07:00
|
|
|
|
|
|
|
struct drm_radeon_gem_userptr {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 addr;
|
|
|
|
__u64 size;
|
|
|
|
__u32 flags;
|
|
|
|
__u32 handle;
|
drm/radeon: add userptr support v8
This patch adds an IOCTL for turning a pointer supplied by
userspace into a buffer object.
It imposes several restrictions upon the memory being mapped:
1. It must be page aligned (both start/end addresses, i.e ptr and size).
2. It must be normal system memory, not a pointer into another map of IO
space (e.g. it must not be a GTT mmapping of another object).
3. The BO is mapped into GTT, so the maximum amount of memory mapped at
all times is still the GTT limit.
4. The BO is only mapped readonly for now, so no write support.
5. List of backing pages is only acquired once, so they represent a
snapshot of the first use.
Exporting and sharing as well as mapping of buffer objects created by
this function is forbidden and results in an -EPERM.
v2: squash all previous changes into first public version
v3: fix tabs, map readonly, don't use MM callback any more
v4: set TTM_PAGE_FLAG_SG so that TTM never messes with the pages,
pin/unpin pages on bind/unbind instead of populate/unpopulate
v5: rebased on 3.17-wip, IOCTL renamed to userptr, reject any unknown
flags, better handle READONLY flag, improve permission check
v6: fix ptr cast warning, use set_page_dirty/mark_page_accessed on unpin
v7: add warning about it's availability in the API definition
v8: drop access_ok check, fix VM mapping bits
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v4)
Reviewed-by: Jérôme Glisse <jglisse@redhat.com> (v4)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2014-08-07 14:36:00 +07:00
|
|
|
};
|
|
|
|
|
2011-12-17 05:03:42 +07:00
|
|
|
#define RADEON_TILING_MACRO 0x1
|
|
|
|
#define RADEON_TILING_MICRO 0x2
|
|
|
|
#define RADEON_TILING_SWAP_16BIT 0x4
|
|
|
|
#define RADEON_TILING_SWAP_32BIT 0x8
|
|
|
|
/* this object requires a surface when mapped - i.e. front buffer */
|
|
|
|
#define RADEON_TILING_SURFACE 0x10
|
|
|
|
#define RADEON_TILING_MICRO_SQUARE 0x20
|
|
|
|
#define RADEON_TILING_EG_BANKW_SHIFT 8
|
|
|
|
#define RADEON_TILING_EG_BANKW_MASK 0xf
|
|
|
|
#define RADEON_TILING_EG_BANKH_SHIFT 12
|
|
|
|
#define RADEON_TILING_EG_BANKH_MASK 0xf
|
|
|
|
#define RADEON_TILING_EG_MACRO_TILE_ASPECT_SHIFT 16
|
|
|
|
#define RADEON_TILING_EG_MACRO_TILE_ASPECT_MASK 0xf
|
|
|
|
#define RADEON_TILING_EG_TILE_SPLIT_SHIFT 24
|
|
|
|
#define RADEON_TILING_EG_TILE_SPLIT_MASK 0xf
|
|
|
|
#define RADEON_TILING_EG_STENCIL_TILE_SPLIT_SHIFT 28
|
|
|
|
#define RADEON_TILING_EG_STENCIL_TILE_SPLIT_MASK 0xf
|
2009-06-24 06:48:08 +07:00
|
|
|
|
|
|
|
struct drm_radeon_gem_set_tiling {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 tiling_flags;
|
|
|
|
__u32 pitch;
|
2009-06-24 06:48:08 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_get_tiling {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 tiling_flags;
|
|
|
|
__u32 pitch;
|
2009-06-24 06:48:08 +07:00
|
|
|
};
|
|
|
|
|
2009-06-05 19:42:42 +07:00
|
|
|
struct drm_radeon_gem_mmap {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 pad;
|
|
|
|
__u64 offset;
|
|
|
|
__u64 size;
|
|
|
|
__u64 addr_ptr;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_set_domain {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 read_domains;
|
|
|
|
__u32 write_domain;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_wait_idle {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 pad;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_busy {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 domain;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_pread {
|
|
|
|
/** Handle for the object being read. */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 pad;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Offset into the object to read from */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 offset;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Length of data to read */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 size;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Pointer to write the data into. */
|
|
|
|
/* void *, but pointers are not 32/64 compatible */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 data_ptr;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_gem_pwrite {
|
|
|
|
/** Handle for the object being written to. */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 pad;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Offset into the object to write to */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 offset;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Length of data to write */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 size;
|
2009-06-05 19:42:42 +07:00
|
|
|
/** Pointer to read the data from. */
|
|
|
|
/* void *, but pointers are not 32/64 compatible */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 data_ptr;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
2014-03-02 06:56:17 +07:00
|
|
|
/* Sets or returns a value associated with a buffer. */
|
|
|
|
struct drm_radeon_gem_op {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle; /* buffer */
|
|
|
|
__u32 op; /* RADEON_GEM_OP_* */
|
|
|
|
__u64 value; /* input or return value */
|
2014-03-02 06:56:17 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
#define RADEON_GEM_OP_GET_INITIAL_DOMAIN 0
|
|
|
|
#define RADEON_GEM_OP_SET_INITIAL_DOMAIN 1
|
|
|
|
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
#define RADEON_VA_MAP 1
|
|
|
|
#define RADEON_VA_UNMAP 2
|
|
|
|
|
|
|
|
#define RADEON_VA_RESULT_OK 0
|
|
|
|
#define RADEON_VA_RESULT_ERROR 1
|
|
|
|
#define RADEON_VA_RESULT_VA_EXIST 2
|
|
|
|
|
|
|
|
#define RADEON_VM_PAGE_VALID (1 << 0)
|
|
|
|
#define RADEON_VM_PAGE_READABLE (1 << 1)
|
|
|
|
#define RADEON_VM_PAGE_WRITEABLE (1 << 2)
|
|
|
|
#define RADEON_VM_PAGE_SYSTEM (1 << 3)
|
|
|
|
#define RADEON_VM_PAGE_SNOOPED (1 << 4)
|
|
|
|
|
|
|
|
struct drm_radeon_gem_va {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 operation;
|
|
|
|
__u32 vm_id;
|
|
|
|
__u32 flags;
|
|
|
|
__u64 offset;
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
};
|
|
|
|
|
2009-06-05 19:42:42 +07:00
|
|
|
#define RADEON_CHUNK_ID_RELOCS 0x01
|
|
|
|
#define RADEON_CHUNK_ID_IB 0x02
|
2011-10-25 06:38:45 +07:00
|
|
|
#define RADEON_CHUNK_ID_FLAGS 0x03
|
2012-03-21 04:18:14 +07:00
|
|
|
#define RADEON_CHUNK_ID_CONST_IB 0x04
|
2011-10-25 06:38:45 +07:00
|
|
|
|
|
|
|
/* The first dword of RADEON_CHUNK_ID_FLAGS is a uint32 of these flags: */
|
|
|
|
#define RADEON_CS_KEEP_TILING_FLAGS 0x01
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
#define RADEON_CS_USE_VM 0x02
|
2012-12-03 03:03:33 +07:00
|
|
|
#define RADEON_CS_END_OF_FRAME 0x04 /* a hint from userspace which CS is the last one */
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
/* The second dword of RADEON_CHUNK_ID_FLAGS is a uint32 that sets the ring type */
|
|
|
|
#define RADEON_CS_RING_GFX 0
|
|
|
|
#define RADEON_CS_RING_COMPUTE 1
|
2012-12-14 00:27:28 +07:00
|
|
|
#define RADEON_CS_RING_DMA 2
|
2013-04-08 17:41:29 +07:00
|
|
|
#define RADEON_CS_RING_UVD 3
|
2013-05-23 17:10:04 +07:00
|
|
|
#define RADEON_CS_RING_VCE 4
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
/* The third dword of RADEON_CHUNK_ID_FLAGS is a sint32 that sets the priority */
|
|
|
|
/* 0 = normal, + = higher priority, - = lower priority */
|
2009-06-05 19:42:42 +07:00
|
|
|
|
|
|
|
struct drm_radeon_cs_chunk {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 chunk_id;
|
|
|
|
__u32 length_dw;
|
|
|
|
__u64 chunk_data;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
2012-01-06 10:11:06 +07:00
|
|
|
/* drm_radeon_cs_reloc.flags */
|
2014-08-15 16:52:53 +07:00
|
|
|
#define RADEON_RELOC_PRIO_MASK (0xf << 0)
|
2012-01-06 10:11:06 +07:00
|
|
|
|
2009-06-05 19:42:42 +07:00
|
|
|
struct drm_radeon_cs_reloc {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 handle;
|
|
|
|
__u32 read_domains;
|
|
|
|
__u32 write_domain;
|
|
|
|
__u32 flags;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
struct drm_radeon_cs {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 num_chunks;
|
|
|
|
__u32 cs_id;
|
|
|
|
/* this points to __u64 * which point to cs chunks */
|
|
|
|
__u64 chunks;
|
2009-06-05 19:42:42 +07:00
|
|
|
/* updates to the limits after this CS ioctl */
|
2014-08-31 18:11:54 +07:00
|
|
|
__u64 gart_limit;
|
|
|
|
__u64 vram_limit;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
#define RADEON_INFO_DEVICE_ID 0x00
|
|
|
|
#define RADEON_INFO_NUM_GB_PIPES 0x01
|
2009-08-20 06:11:39 +07:00
|
|
|
#define RADEON_INFO_NUM_Z_PIPES 0x02
|
2009-09-16 20:24:21 +07:00
|
|
|
#define RADEON_INFO_ACCEL_WORKING 0x03
|
2010-05-12 23:01:13 +07:00
|
|
|
#define RADEON_INFO_CRTC_FROM_ID 0x04
|
2010-06-04 06:00:03 +07:00
|
|
|
#define RADEON_INFO_ACCEL_WORKING2 0x05
|
2010-06-05 00:10:12 +07:00
|
|
|
#define RADEON_INFO_TILING_CONFIG 0x06
|
2010-07-13 08:11:11 +07:00
|
|
|
#define RADEON_INFO_WANT_HYPERZ 0x07
|
2011-01-05 11:46:48 +07:00
|
|
|
#define RADEON_INFO_WANT_CMASK 0x08 /* get access to CMASK on r300 */
|
2011-01-25 05:14:26 +07:00
|
|
|
#define RADEON_INFO_CLOCK_CRYSTAL_FREQ 0x09 /* clock crystal frequency */
|
2011-03-01 11:32:27 +07:00
|
|
|
#define RADEON_INFO_NUM_BACKENDS 0x0a /* DB/backends for r600+ - need for OQ */
|
2011-04-27 00:27:43 +07:00
|
|
|
#define RADEON_INFO_NUM_TILE_PIPES 0x0b /* tile pipes for r600+ */
|
2011-05-04 06:28:02 +07:00
|
|
|
#define RADEON_INFO_FUSION_GART_WORKING 0x0c /* fusion writes to GTT were broken before this */
|
2011-07-16 02:53:52 +07:00
|
|
|
#define RADEON_INFO_BACKEND_MAP 0x0d /* pipe to backend map, needed by mesa */
|
drm/radeon: GPU virtual memory support v22
Virtual address space are per drm client (opener of /dev/drm).
Client are in charge of virtual address space, they need to
map bo into it by calling DRM_RADEON_GEM_VA ioctl.
First 16M of virtual address space is reserved by the kernel.
Once using 2 level page table we should be able to have a small
vram memory footprint for each pt (there would be one pt for all
gart, one for all vram and then one first level for each virtual
address space).
Plan include using the sub allocator for a common vm page table
area and using memcpy to copy vm page table in & out. Or use
a gart object and copy things in & out using dma.
v2: agd5f fixes:
- Add vram base offset for vram pages. The GPU physical address of a
vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete
cards and the physical bus address of the stolen memory on
integrated chips.
- VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1
v3: agd5f:
- integrate with the semaphore/multi-ring stuff
v4:
- rebase on top ttm dma & multi-ring stuff
- userspace is now in charge of the address space
- no more specific cs vm ioctl, instead cs ioctl has a new
chunk
v5:
- properly handle mem == NULL case from move_notify callback
- fix the vm cleanup path
v6:
- fix update of page table to only happen on valid mem placement
v7:
- add tlb flush for each vm context
- add flags to define mapping property (readable, writeable, snooped)
- make ring id implicit from ib->fence->ring, up to each asic callback
to then do ring specific scheduling if vm ib scheduling function
v8:
- add query for ib limit and kernel reserved virtual space
- rename vm->size to max_pfn (maximum number of page)
- update gem_va ioctl to also allow unmap operation
- bump kernel version to allow userspace to query for vm support
v9:
- rebuild page table only when bind and incrementaly depending
on bo referenced by cs and that have been moved
- allow virtual address space to grow
- use sa allocator for vram page table
- return invalid when querying vm limit on non cayman GPU
- dump vm fault register on lockup
v10: agd5f:
- Move the vm schedule_ib callback to a standalone function, remove
the callback and use the existing ib_execute callback for VM IBs.
v11:
- rebase on top of lastest Linus
v12: agd5f:
- remove spurious backslash
- set IB vm_id to 0 in radeon_ib_get()
v13: agd5f:
- fix handling of RADEON_CHUNK_ID_FLAGS
v14:
- fix va destruction
- fix suspend resume
- forbid bo to have several different va in same vm
v15:
- rebase
v16:
- cleanup left over of vm init/fini
v17: agd5f:
- cs checker
v18: agd5f:
- reworks the CS ioctl to better support multiple rings and
VM. Rather than adding a new chunk id for VM, just re-use the
IB chunk id and add a new flags for VM mode. Also define additional
dwords for the flags chunk id to define the what ring we want to use
(gfx, compute, uvd, etc.) and the priority.
v19:
- fix cs fini in weird case of no ib
- semi working flush fix for ni
- rebase on top of sa allocator changes
v20: agd5f:
- further CS ioctl cleanups from Christian's comments
v21: agd5f:
- integrate CS checker improvements
v22: agd5f:
- final cleanups for release, only allow VM CS on cayman
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 10:11:05 +07:00
|
|
|
/* virtual address start, va < start are reserved by the kernel */
|
|
|
|
#define RADEON_INFO_VA_START 0x0e
|
|
|
|
/* maximum size of ib using the virtual memory cs */
|
|
|
|
#define RADEON_INFO_IB_VM_MAX_SIZE 0x0f
|
2012-03-21 04:17:55 +07:00
|
|
|
/* max pipes - needed for compute shaders */
|
|
|
|
#define RADEON_INFO_MAX_PIPES 0x10
|
2012-08-09 21:34:17 +07:00
|
|
|
/* timestamp for GL_ARB_timer_query (OpenGL), returns the current GPU clock */
|
|
|
|
#define RADEON_INFO_TIMESTAMP 0x11
|
2012-12-05 00:55:37 +07:00
|
|
|
/* max shader engines (SE) - needed for geometry shaders, etc. */
|
|
|
|
#define RADEON_INFO_MAX_SE 0x12
|
|
|
|
/* max SH per SE */
|
|
|
|
#define RADEON_INFO_MAX_SH_PER_SE 0x13
|
2013-04-09 04:25:47 +07:00
|
|
|
/* fast fb access is enabled */
|
|
|
|
#define RADEON_INFO_FASTFB_WORKING 0x14
|
2013-04-09 21:35:42 +07:00
|
|
|
/* query if a RADEON_CS_RING_* submission is supported */
|
|
|
|
#define RADEON_INFO_RING_WORKING 0x15
|
2013-04-09 22:17:08 +07:00
|
|
|
/* SI tile mode array */
|
|
|
|
#define RADEON_INFO_SI_TILE_MODE_ARRAY 0x16
|
2013-08-17 04:47:39 +07:00
|
|
|
/* query if CP DMA is supported on the compute ring */
|
|
|
|
#define RADEON_INFO_SI_CP_DMA_COMPUTE 0x17
|
2013-11-18 16:26:00 +07:00
|
|
|
/* CIK macrotile mode array */
|
|
|
|
#define RADEON_INFO_CIK_MACROTILE_MODE_ARRAY 0x18
|
2013-12-22 08:18:01 +07:00
|
|
|
/* query the number of render backends */
|
|
|
|
#define RADEON_INFO_SI_BACKEND_ENABLED_MASK 0x19
|
2014-01-21 06:20:29 +07:00
|
|
|
/* max engine clock - needed for OpenCL */
|
|
|
|
#define RADEON_INFO_MAX_SCLK 0x1a
|
2014-01-23 23:50:49 +07:00
|
|
|
/* version of VCE firmware */
|
|
|
|
#define RADEON_INFO_VCE_FW_VERSION 0x1b
|
|
|
|
/* version of VCE feedback */
|
|
|
|
#define RADEON_INFO_VCE_FB_VERSION 0x1c
|
2014-03-02 06:56:18 +07:00
|
|
|
#define RADEON_INFO_NUM_BYTES_MOVED 0x1d
|
|
|
|
#define RADEON_INFO_VRAM_USAGE 0x1e
|
|
|
|
#define RADEON_INFO_GTT_USAGE 0x1f
|
2014-06-03 03:13:21 +07:00
|
|
|
#define RADEON_INFO_ACTIVE_CU_COUNT 0x20
|
2014-09-30 21:04:40 +07:00
|
|
|
#define RADEON_INFO_CURRENT_GPU_TEMP 0x21
|
2014-09-30 22:33:30 +07:00
|
|
|
#define RADEON_INFO_CURRENT_GPU_SCLK 0x22
|
|
|
|
#define RADEON_INFO_CURRENT_GPU_MCLK 0x23
|
2014-10-01 22:26:50 +07:00
|
|
|
#define RADEON_INFO_READ_REG 0x24
|
2015-06-16 15:28:16 +07:00
|
|
|
#define RADEON_INFO_VA_UNMAP_WORKING 0x25
|
Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux
Pull drm updates from Dave Airlie:
"This is the main drm pull request for v4.2.
I've one other new driver from freescale on my radar, it's been posted
and reviewed, I'd just like to get someone to give it a last look, so
maybe I'll send it or maybe I'll leave it.
There is no major nouveau changes in here, Ben was working on
something big, and we agreed it was a bit late, there wasn't anything
else he considered urgent to merge.
There might be another msm pull for some bits that are waiting on
arm-soc, I'll see how we time it.
This touches some "of" stuff, acks are in place except for the fixes
to the build in various configs,t hat I just applied.
Summary:
New drivers:
- virtio-gpu:
KMS only pieces of driver for virtio-gpu in qemu.
This is just the first part of this driver, enough to run
unaccelerated userspace on. As qemu merges more we'll start
adding the 3D features for the virgl 3d work.
- amdgpu:
a new driver from AMD to driver their newer GPUs. (VI+)
It contains a new cleaner userspace API, and is a clean
break from radeon moving forward, that AMD are going to
concentrate on. It also contains a set of register headers
auto generated from AMD internal database.
core:
- atomic modesetting API completed, enabled by default now.
- Add support for mode_id blob to atomic ioctl to complete interface.
- bunch of Displayport MST fixes
- lots of misc fixes.
panel:
- new simple panels
- fix some long-standing build issues with bridge drivers
radeon:
- VCE1 support
- add a GPU reset counter for userspace
- lots of fixes.
amdkfd:
- H/W debugger support module
- static user-mode queues
- support killing all the waves when a process terminates
- use standard DECLARE_BITMAP
i915:
- Add Broxton support
- S3, rotation support for Skylake
- RPS booting tuning
- CPT modeset sequence fixes
- ns2501 dither support
- enable cmd parser on haswell
- cdclk handling fixes
- gen8 dynamic pte allocation
- lots of atomic conversion work
exynos:
- Add atomic modesetting support
- Add iommu support
- Consolidate drm driver initialization
- and MIC, DECON and MIPI-DSI support for exynos5433
omapdrm:
- atomic modesetting support (fixes lots of things in rewrite)
tegra:
- DP aux transaction fixes
- iommu support fix
msm:
- adreno a306 support
- various dsi bits
- various 64-bit fixes
- NV12MT support
rcar-du:
- atomic and misc fixes
sti:
- fix HDMI timing complaince
tilcdc:
- use drm component API to access tda998x driver
- fix module unloading
qxl:
- stability fixes"
* 'drm-next' of git://people.freedesktop.org/~airlied/linux: (872 commits)
drm/nouveau: Pause between setting gpu to D3hot and cutting the power
drm/dp/mst: close deadlock in connector destruction.
drm: Always enable atomic API
drm/vgem: Set unique to "vgem"
of: fix a build error to of_graph_get_endpoint_by_regs function
drm/dp/mst: take lock around looking up the branch device on hpd irq
drm/dp/mst: make sure mst_primary mstb is valid in work function
of: add EXPORT_SYMBOL for of_graph_get_endpoint_by_regs
ARM: dts: rename the clock of MIPI DSI 'pll_clk' to 'sclk_mipi'
drm/atomic: Don't set crtc_state->enable manually
drm/exynos: dsi: do not set TE GPIO direction by input
drm/exynos: dsi: add support for MIC driver as a bridge
drm/exynos: dsi: add support for Exynos5433
drm/exynos: dsi: make use of array for clock access
drm/exynos: dsi: make use of driver data for static values
drm/exynos: dsi: add macros for register access
drm/exynos: dsi: rename pll_clk to sclk_clk
drm/exynos: mic: add MIC driver
of: add helper for getting endpoint node of specific identifiers
drm/exynos: add Exynos5433 decon driver
...
2015-06-27 03:18:51 +07:00
|
|
|
#define RADEON_INFO_GPU_RESET_COUNTER 0x26
|
2009-06-05 19:42:42 +07:00
|
|
|
|
|
|
|
struct drm_radeon_info {
|
2014-08-31 18:11:54 +07:00
|
|
|
__u32 request;
|
|
|
|
__u32 pad;
|
|
|
|
__u64 value;
|
2009-06-05 19:42:42 +07:00
|
|
|
};
|
|
|
|
|
2013-04-09 22:17:08 +07:00
|
|
|
/* Those correspond to the tile index to use, this is to explicitly state
|
|
|
|
* the API that is implicitly defined by the tile mode array.
|
|
|
|
*/
|
|
|
|
#define SI_TILE_MODE_COLOR_LINEAR_ALIGNED 8
|
|
|
|
#define SI_TILE_MODE_COLOR_1D 13
|
|
|
|
#define SI_TILE_MODE_COLOR_1D_SCANOUT 9
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_8BPP 14
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_16BPP 15
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_32BPP 16
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_64BPP 17
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_SCANOUT_16BPP 11
|
|
|
|
#define SI_TILE_MODE_COLOR_2D_SCANOUT_32BPP 12
|
|
|
|
#define SI_TILE_MODE_DEPTH_STENCIL_1D 4
|
|
|
|
#define SI_TILE_MODE_DEPTH_STENCIL_2D 0
|
|
|
|
#define SI_TILE_MODE_DEPTH_STENCIL_2D_2AA 3
|
|
|
|
#define SI_TILE_MODE_DEPTH_STENCIL_2D_4AA 3
|
|
|
|
#define SI_TILE_MODE_DEPTH_STENCIL_2D_8AA 2
|
|
|
|
|
2013-09-18 23:23:51 +07:00
|
|
|
#define CIK_TILE_MODE_DEPTH_STENCIL_1D 5
|
|
|
|
|
2016-04-08 01:13:24 +07:00
|
|
|
#if defined(__cplusplus)
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif
|