linux/drivers/gpu/drm/radeon/evergreen_cs.c

1575 lines
43 KiB
C
Raw Normal View History

/*
* Copyright 2010 Advanced Micro Devices, Inc.
* Copyright 2008 Red Hat Inc.
* Copyright 2009 Jerome Glisse.
*
* 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 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
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) 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: Dave Airlie
* Alex Deucher
* Jerome Glisse
*/
#include "drmP.h"
#include "radeon.h"
#include "evergreend.h"
#include "evergreen_reg_safe.h"
#include "cayman_reg_safe.h"
static int evergreen_cs_packet_next_reloc(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc);
struct evergreen_cs_track {
u32 group_size;
u32 nbanks;
u32 npipes;
u32 row_size;
/* value we track */
u32 nsamples;
u32 cb_color_base_last[12];
struct radeon_bo *cb_color_bo[12];
u32 cb_color_bo_offset[12];
struct radeon_bo *cb_color_fmask_bo[8];
struct radeon_bo *cb_color_cmask_bo[8];
u32 cb_color_info[12];
u32 cb_color_view[12];
u32 cb_color_pitch_idx[12];
u32 cb_color_slice_idx[12];
u32 cb_color_dim_idx[12];
u32 cb_color_dim[12];
u32 cb_color_pitch[12];
u32 cb_color_slice[12];
u32 cb_color_cmask_slice[8];
u32 cb_color_fmask_slice[8];
u32 cb_target_mask;
u32 cb_shader_mask;
u32 vgt_strmout_config;
u32 vgt_strmout_buffer_config;
u32 db_depth_control;
u32 db_depth_view;
u32 db_depth_size;
u32 db_depth_size_idx;
u32 db_z_info;
u32 db_z_idx;
u32 db_z_read_offset;
u32 db_z_write_offset;
struct radeon_bo *db_z_read_bo;
struct radeon_bo *db_z_write_bo;
u32 db_s_info;
u32 db_s_idx;
u32 db_s_read_offset;
u32 db_s_write_offset;
struct radeon_bo *db_s_read_bo;
struct radeon_bo *db_s_write_bo;
};
static u32 evergreen_cs_get_aray_mode(u32 tiling_flags)
{
if (tiling_flags & RADEON_TILING_MACRO)
return ARRAY_2D_TILED_THIN1;
else if (tiling_flags & RADEON_TILING_MICRO)
return ARRAY_1D_TILED_THIN1;
else
return ARRAY_LINEAR_GENERAL;
}
static u32 evergreen_cs_get_num_banks(u32 nbanks)
{
switch (nbanks) {
case 2:
return ADDR_SURF_2_BANK;
case 4:
return ADDR_SURF_4_BANK;
case 8:
default:
return ADDR_SURF_8_BANK;
case 16:
return ADDR_SURF_16_BANK;
}
}
static u32 evergreen_cs_get_tile_split(u32 row_size)
{
switch (row_size) {
case 1:
default:
return ADDR_SURF_TILE_SPLIT_1KB;
case 2:
return ADDR_SURF_TILE_SPLIT_2KB;
case 4:
return ADDR_SURF_TILE_SPLIT_4KB;
}
}
static void evergreen_cs_track_init(struct evergreen_cs_track *track)
{
int i;
for (i = 0; i < 8; i++) {
track->cb_color_fmask_bo[i] = NULL;
track->cb_color_cmask_bo[i] = NULL;
track->cb_color_cmask_slice[i] = 0;
track->cb_color_fmask_slice[i] = 0;
}
for (i = 0; i < 12; i++) {
track->cb_color_base_last[i] = 0;
track->cb_color_bo[i] = NULL;
track->cb_color_bo_offset[i] = 0xFFFFFFFF;
track->cb_color_info[i] = 0;
track->cb_color_view[i] = 0;
track->cb_color_pitch_idx[i] = 0;
track->cb_color_slice_idx[i] = 0;
track->cb_color_dim[i] = 0;
track->cb_color_pitch[i] = 0;
track->cb_color_slice[i] = 0;
track->cb_color_dim[i] = 0;
}
track->cb_target_mask = 0xFFFFFFFF;
track->cb_shader_mask = 0xFFFFFFFF;
track->db_depth_view = 0xFFFFC000;
track->db_depth_size = 0xFFFFFFFF;
track->db_depth_size_idx = 0;
track->db_depth_control = 0xFFFFFFFF;
track->db_z_info = 0xFFFFFFFF;
track->db_z_idx = 0xFFFFFFFF;
track->db_z_read_offset = 0xFFFFFFFF;
track->db_z_write_offset = 0xFFFFFFFF;
track->db_z_read_bo = NULL;
track->db_z_write_bo = NULL;
track->db_s_info = 0xFFFFFFFF;
track->db_s_idx = 0xFFFFFFFF;
track->db_s_read_offset = 0xFFFFFFFF;
track->db_s_write_offset = 0xFFFFFFFF;
track->db_s_read_bo = NULL;
track->db_s_write_bo = NULL;
}
static int evergreen_cs_track_check(struct radeon_cs_parser *p)
{
struct evergreen_cs_track *track = p->track;
/* we don't support stream out buffer yet */
if (track->vgt_strmout_config || track->vgt_strmout_buffer_config) {
dev_warn(p->dev, "this kernel doesn't support SMX output buffer\n");
return -EINVAL;
}
/* XXX fill in */
return 0;
}
/**
* evergreen_cs_packet_parse() - parse cp packet and point ib index to next packet
* @parser: parser structure holding parsing context.
* @pkt: where to store packet informations
*
* Assume that chunk_ib_index is properly set. Will return -EINVAL
* if packet is bigger than remaining ib size. or if packets is unknown.
**/
int evergreen_cs_packet_parse(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt,
unsigned idx)
{
struct radeon_cs_chunk *ib_chunk = &p->chunks[p->chunk_ib_idx];
uint32_t header;
if (idx >= ib_chunk->length_dw) {
DRM_ERROR("Can not parse packet at %d after CS end %d !\n",
idx, ib_chunk->length_dw);
return -EINVAL;
}
header = radeon_get_ib_value(p, idx);
pkt->idx = idx;
pkt->type = CP_PACKET_GET_TYPE(header);
pkt->count = CP_PACKET_GET_COUNT(header);
pkt->one_reg_wr = 0;
switch (pkt->type) {
case PACKET_TYPE0:
pkt->reg = CP_PACKET0_GET_REG(header);
break;
case PACKET_TYPE3:
pkt->opcode = CP_PACKET3_GET_OPCODE(header);
break;
case PACKET_TYPE2:
pkt->count = -1;
break;
default:
DRM_ERROR("Unknown packet type %d at %d !\n", pkt->type, idx);
return -EINVAL;
}
if ((pkt->count + 1 + pkt->idx) >= ib_chunk->length_dw) {
DRM_ERROR("Packet (%d:%d:%d) end after CS buffer (%d) !\n",
pkt->idx, pkt->type, pkt->count, ib_chunk->length_dw);
return -EINVAL;
}
return 0;
}
/**
* evergreen_cs_packet_next_reloc() - parse next packet which should be reloc packet3
* @parser: parser structure holding parsing context.
* @data: pointer to relocation data
* @offset_start: starting offset
* @offset_mask: offset mask (to align start offset on)
* @reloc: reloc informations
*
* Check next packet is relocation packet3, do bo validation and compute
* GPU offset using the provided start.
**/
static int evergreen_cs_packet_next_reloc(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc)
{
struct radeon_cs_chunk *relocs_chunk;
struct radeon_cs_packet p3reloc;
unsigned idx;
int r;
if (p->chunk_relocs_idx == -1) {
DRM_ERROR("No relocation chunk !\n");
return -EINVAL;
}
*cs_reloc = NULL;
relocs_chunk = &p->chunks[p->chunk_relocs_idx];
r = evergreen_cs_packet_parse(p, &p3reloc, p->idx);
if (r) {
return r;
}
p->idx += p3reloc.count + 2;
if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
DRM_ERROR("No packet3 for relocation for packet at %d.\n",
p3reloc.idx);
return -EINVAL;
}
idx = radeon_get_ib_value(p, p3reloc.idx + 1);
if (idx >= relocs_chunk->length_dw) {
DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
idx, relocs_chunk->length_dw);
return -EINVAL;
}
/* FIXME: we assume reloc size is 4 dwords */
*cs_reloc = p->relocs_ptr[(idx / 4)];
return 0;
}
/**
* evergreen_cs_packet_next_vline() - parse userspace VLINE packet
* @parser: parser structure holding parsing context.
*
* Userspace sends a special sequence for VLINE waits.
* PACKET0 - VLINE_START_END + value
* PACKET3 - WAIT_REG_MEM poll vline status reg
* RELOC (P3) - crtc_id in reloc.
*
* This function parses this and relocates the VLINE START END
* and WAIT_REG_MEM packets to the correct crtc.
* It also detects a switched off crtc and nulls out the
* wait in that case.
*/
static int evergreen_cs_packet_parse_vline(struct radeon_cs_parser *p)
{
struct drm_mode_object *obj;
struct drm_crtc *crtc;
struct radeon_crtc *radeon_crtc;
struct radeon_cs_packet p3reloc, wait_reg_mem;
int crtc_id;
int r;
uint32_t header, h_idx, reg, wait_reg_mem_info;
volatile uint32_t *ib;
ib = p->ib->ptr;
/* parse the WAIT_REG_MEM */
r = evergreen_cs_packet_parse(p, &wait_reg_mem, p->idx);
if (r)
return r;
/* check its a WAIT_REG_MEM */
if (wait_reg_mem.type != PACKET_TYPE3 ||
wait_reg_mem.opcode != PACKET3_WAIT_REG_MEM) {
DRM_ERROR("vline wait missing WAIT_REG_MEM segment\n");
return -EINVAL;
}
wait_reg_mem_info = radeon_get_ib_value(p, wait_reg_mem.idx + 1);
/* bit 4 is reg (0) or mem (1) */
if (wait_reg_mem_info & 0x10) {
DRM_ERROR("vline WAIT_REG_MEM waiting on MEM rather than REG\n");
return -EINVAL;
}
/* waiting for value to be equal */
if ((wait_reg_mem_info & 0x7) != 0x3) {
DRM_ERROR("vline WAIT_REG_MEM function not equal\n");
return -EINVAL;
}
if ((radeon_get_ib_value(p, wait_reg_mem.idx + 2) << 2) != EVERGREEN_VLINE_STATUS) {
DRM_ERROR("vline WAIT_REG_MEM bad reg\n");
return -EINVAL;
}
if (radeon_get_ib_value(p, wait_reg_mem.idx + 5) != EVERGREEN_VLINE_STAT) {
DRM_ERROR("vline WAIT_REG_MEM bad bit mask\n");
return -EINVAL;
}
/* jump over the NOP */
r = evergreen_cs_packet_parse(p, &p3reloc, p->idx + wait_reg_mem.count + 2);
if (r)
return r;
h_idx = p->idx - 2;
p->idx += wait_reg_mem.count + 2;
p->idx += p3reloc.count + 2;
header = radeon_get_ib_value(p, h_idx);
crtc_id = radeon_get_ib_value(p, h_idx + 2 + 7 + 1);
reg = CP_PACKET0_GET_REG(header);
obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC);
if (!obj) {
DRM_ERROR("cannot find crtc %d\n", crtc_id);
return -EINVAL;
}
crtc = obj_to_crtc(obj);
radeon_crtc = to_radeon_crtc(crtc);
crtc_id = radeon_crtc->crtc_id;
if (!crtc->enabled) {
/* if the CRTC isn't enabled - we need to nop out the WAIT_REG_MEM */
ib[h_idx + 2] = PACKET2(0);
ib[h_idx + 3] = PACKET2(0);
ib[h_idx + 4] = PACKET2(0);
ib[h_idx + 5] = PACKET2(0);
ib[h_idx + 6] = PACKET2(0);
ib[h_idx + 7] = PACKET2(0);
ib[h_idx + 8] = PACKET2(0);
} else {
switch (reg) {
case EVERGREEN_VLINE_START_END:
header &= ~R600_CP_PACKET0_REG_MASK;
header |= (EVERGREEN_VLINE_START_END + radeon_crtc->crtc_offset) >> 2;
ib[h_idx] = header;
ib[h_idx + 4] = (EVERGREEN_VLINE_STATUS + radeon_crtc->crtc_offset) >> 2;
break;
default:
DRM_ERROR("unknown crtc reloc\n");
return -EINVAL;
}
}
return 0;
}
static int evergreen_packet0_check(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt,
unsigned idx, unsigned reg)
{
int r;
switch (reg) {
case EVERGREEN_VLINE_START_END:
r = evergreen_cs_packet_parse_vline(p);
if (r) {
DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
idx, reg);
return r;
}
break;
default:
printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
reg, idx);
return -EINVAL;
}
return 0;
}
static int evergreen_cs_parse_packet0(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt)
{
unsigned reg, i;
unsigned idx;
int r;
idx = pkt->idx + 1;
reg = pkt->reg;
for (i = 0; i <= pkt->count; i++, idx++, reg += 4) {
r = evergreen_packet0_check(p, pkt, idx, reg);
if (r) {
return r;
}
}
return 0;
}
/**
* evergreen_cs_check_reg() - check if register is authorized or not
* @parser: parser structure holding parsing context
* @reg: register we are testing
* @idx: index into the cs buffer
*
* This function will test against evergreen_reg_safe_bm and return 0
* if register is safe. If register is not flag as safe this function
* will test it against a list of register needind special handling.
*/
static int evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
{
struct evergreen_cs_track *track = (struct evergreen_cs_track *)p->track;
struct radeon_cs_reloc *reloc;
u32 last_reg;
u32 m, i, tmp, *ib;
int r;
if (p->rdev->family >= CHIP_CAYMAN)
last_reg = ARRAY_SIZE(cayman_reg_safe_bm);
else
last_reg = ARRAY_SIZE(evergreen_reg_safe_bm);
i = (reg >> 7);
if (i >= last_reg) {
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return -EINVAL;
}
m = 1 << ((reg >> 2) & 31);
if (p->rdev->family >= CHIP_CAYMAN) {
if (!(cayman_reg_safe_bm[i] & m))
return 0;
} else {
if (!(evergreen_reg_safe_bm[i] & m))
return 0;
}
ib = p->ib->ptr;
switch (reg) {
/* force following reg to 0 in an attempt to disable out buffer
* which will need us to better understand how it works to perform
* security check on it (Jerome)
*/
case SQ_ESGS_RING_SIZE:
case SQ_GSVS_RING_SIZE:
case SQ_ESTMP_RING_SIZE:
case SQ_GSTMP_RING_SIZE:
case SQ_HSTMP_RING_SIZE:
case SQ_LSTMP_RING_SIZE:
case SQ_PSTMP_RING_SIZE:
case SQ_VSTMP_RING_SIZE:
case SQ_ESGS_RING_ITEMSIZE:
case SQ_ESTMP_RING_ITEMSIZE:
case SQ_GSTMP_RING_ITEMSIZE:
case SQ_GSVS_RING_ITEMSIZE:
case SQ_GS_VERT_ITEMSIZE:
case SQ_GS_VERT_ITEMSIZE_1:
case SQ_GS_VERT_ITEMSIZE_2:
case SQ_GS_VERT_ITEMSIZE_3:
case SQ_GSVS_RING_OFFSET_1:
case SQ_GSVS_RING_OFFSET_2:
case SQ_GSVS_RING_OFFSET_3:
case SQ_HSTMP_RING_ITEMSIZE:
case SQ_LSTMP_RING_ITEMSIZE:
case SQ_PSTMP_RING_ITEMSIZE:
case SQ_VSTMP_RING_ITEMSIZE:
case VGT_TF_RING_SIZE:
/* get value to populate the IB don't remove */
/*tmp =radeon_get_ib_value(p, idx);
ib[idx] = 0;*/
break;
case SQ_ESGS_RING_BASE:
case SQ_GSVS_RING_BASE:
case SQ_ESTMP_RING_BASE:
case SQ_GSTMP_RING_BASE:
case SQ_HSTMP_RING_BASE:
case SQ_LSTMP_RING_BASE:
case SQ_PSTMP_RING_BASE:
case SQ_VSTMP_RING_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
case DB_DEPTH_CONTROL:
track->db_depth_control = radeon_get_ib_value(p, idx);
break;
case CAYMAN_DB_EQAA:
if (p->rdev->family < CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
break;
case CAYMAN_DB_DEPTH_INFO:
if (p->rdev->family < CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
break;
case DB_Z_INFO:
track->db_z_info = radeon_get_ib_value(p, idx);
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (!p->keep_tiling_flags) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] &= ~Z_ARRAY_MODE(0xf);
track->db_z_info &= ~Z_ARRAY_MODE(0xf);
ib[idx] |= Z_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
track->db_z_info |= Z_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
ib[idx] |= DB_NUM_BANKS(evergreen_cs_get_num_banks(track->nbanks));
ib[idx] |= DB_TILE_SPLIT(evergreen_cs_get_tile_split(track->row_size));
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
}
}
break;
case DB_STENCIL_INFO:
track->db_s_info = radeon_get_ib_value(p, idx);
break;
case DB_DEPTH_VIEW:
track->db_depth_view = radeon_get_ib_value(p, idx);
break;
case DB_DEPTH_SIZE:
track->db_depth_size = radeon_get_ib_value(p, idx);
track->db_depth_size_idx = idx;
break;
case DB_Z_READ_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->db_z_read_offset = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->db_z_read_bo = reloc->robj;
break;
case DB_Z_WRITE_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->db_z_write_offset = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->db_z_write_bo = reloc->robj;
break;
case DB_STENCIL_READ_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->db_s_read_offset = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->db_s_read_bo = reloc->robj;
break;
case DB_STENCIL_WRITE_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->db_s_write_offset = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->db_s_write_bo = reloc->robj;
break;
case VGT_STRMOUT_CONFIG:
track->vgt_strmout_config = radeon_get_ib_value(p, idx);
break;
case VGT_STRMOUT_BUFFER_CONFIG:
track->vgt_strmout_buffer_config = radeon_get_ib_value(p, idx);
break;
case CB_TARGET_MASK:
track->cb_target_mask = radeon_get_ib_value(p, idx);
break;
case CB_SHADER_MASK:
track->cb_shader_mask = radeon_get_ib_value(p, idx);
break;
case PA_SC_AA_CONFIG:
if (p->rdev->family >= CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
tmp = radeon_get_ib_value(p, idx) & MSAA_NUM_SAMPLES_MASK;
track->nsamples = 1 << tmp;
break;
case CAYMAN_PA_SC_AA_CONFIG:
if (p->rdev->family < CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
tmp = radeon_get_ib_value(p, idx) & CAYMAN_MSAA_NUM_SAMPLES_MASK;
track->nsamples = 1 << tmp;
break;
case CB_COLOR0_VIEW:
case CB_COLOR1_VIEW:
case CB_COLOR2_VIEW:
case CB_COLOR3_VIEW:
case CB_COLOR4_VIEW:
case CB_COLOR5_VIEW:
case CB_COLOR6_VIEW:
case CB_COLOR7_VIEW:
tmp = (reg - CB_COLOR0_VIEW) / 0x3c;
track->cb_color_view[tmp] = radeon_get_ib_value(p, idx);
break;
case CB_COLOR8_VIEW:
case CB_COLOR9_VIEW:
case CB_COLOR10_VIEW:
case CB_COLOR11_VIEW:
tmp = ((reg - CB_COLOR8_VIEW) / 0x1c) + 8;
track->cb_color_view[tmp] = radeon_get_ib_value(p, idx);
break;
case CB_COLOR0_INFO:
case CB_COLOR1_INFO:
case CB_COLOR2_INFO:
case CB_COLOR3_INFO:
case CB_COLOR4_INFO:
case CB_COLOR5_INFO:
case CB_COLOR6_INFO:
case CB_COLOR7_INFO:
tmp = (reg - CB_COLOR0_INFO) / 0x3c;
track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (!p->keep_tiling_flags) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] |= CB_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
track->cb_color_info[tmp] |= CB_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
}
break;
case CB_COLOR8_INFO:
case CB_COLOR9_INFO:
case CB_COLOR10_INFO:
case CB_COLOR11_INFO:
tmp = ((reg - CB_COLOR8_INFO) / 0x1c) + 8;
track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (!p->keep_tiling_flags) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] |= CB_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
track->cb_color_info[tmp] |= CB_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
}
break;
case CB_COLOR0_PITCH:
case CB_COLOR1_PITCH:
case CB_COLOR2_PITCH:
case CB_COLOR3_PITCH:
case CB_COLOR4_PITCH:
case CB_COLOR5_PITCH:
case CB_COLOR6_PITCH:
case CB_COLOR7_PITCH:
tmp = (reg - CB_COLOR0_PITCH) / 0x3c;
track->cb_color_pitch[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_pitch_idx[tmp] = idx;
break;
case CB_COLOR8_PITCH:
case CB_COLOR9_PITCH:
case CB_COLOR10_PITCH:
case CB_COLOR11_PITCH:
tmp = ((reg - CB_COLOR8_PITCH) / 0x1c) + 8;
track->cb_color_pitch[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_pitch_idx[tmp] = idx;
break;
case CB_COLOR0_SLICE:
case CB_COLOR1_SLICE:
case CB_COLOR2_SLICE:
case CB_COLOR3_SLICE:
case CB_COLOR4_SLICE:
case CB_COLOR5_SLICE:
case CB_COLOR6_SLICE:
case CB_COLOR7_SLICE:
tmp = (reg - CB_COLOR0_SLICE) / 0x3c;
track->cb_color_slice[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_slice_idx[tmp] = idx;
break;
case CB_COLOR8_SLICE:
case CB_COLOR9_SLICE:
case CB_COLOR10_SLICE:
case CB_COLOR11_SLICE:
tmp = ((reg - CB_COLOR8_SLICE) / 0x1c) + 8;
track->cb_color_slice[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_slice_idx[tmp] = idx;
break;
case CB_COLOR0_ATTRIB:
case CB_COLOR1_ATTRIB:
case CB_COLOR2_ATTRIB:
case CB_COLOR3_ATTRIB:
case CB_COLOR4_ATTRIB:
case CB_COLOR5_ATTRIB:
case CB_COLOR6_ATTRIB:
case CB_COLOR7_ATTRIB:
case CB_COLOR8_ATTRIB:
case CB_COLOR9_ATTRIB:
case CB_COLOR10_ATTRIB:
case CB_COLOR11_ATTRIB:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
ib[idx] |= CB_NUM_BANKS(evergreen_cs_get_num_banks(track->nbanks));
ib[idx] |= CB_TILE_SPLIT(evergreen_cs_get_tile_split(track->row_size));
}
break;
case CB_COLOR0_DIM:
case CB_COLOR1_DIM:
case CB_COLOR2_DIM:
case CB_COLOR3_DIM:
case CB_COLOR4_DIM:
case CB_COLOR5_DIM:
case CB_COLOR6_DIM:
case CB_COLOR7_DIM:
tmp = (reg - CB_COLOR0_DIM) / 0x3c;
track->cb_color_dim[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_dim_idx[tmp] = idx;
break;
case CB_COLOR8_DIM:
case CB_COLOR9_DIM:
case CB_COLOR10_DIM:
case CB_COLOR11_DIM:
tmp = ((reg - CB_COLOR8_DIM) / 0x1c) + 8;
track->cb_color_dim[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_dim_idx[tmp] = idx;
break;
case CB_COLOR0_FMASK:
case CB_COLOR1_FMASK:
case CB_COLOR2_FMASK:
case CB_COLOR3_FMASK:
case CB_COLOR4_FMASK:
case CB_COLOR5_FMASK:
case CB_COLOR6_FMASK:
case CB_COLOR7_FMASK:
tmp = (reg - CB_COLOR0_FMASK) / 0x3c;
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->cb_color_fmask_bo[tmp] = reloc->robj;
break;
case CB_COLOR0_CMASK:
case CB_COLOR1_CMASK:
case CB_COLOR2_CMASK:
case CB_COLOR3_CMASK:
case CB_COLOR4_CMASK:
case CB_COLOR5_CMASK:
case CB_COLOR6_CMASK:
case CB_COLOR7_CMASK:
tmp = (reg - CB_COLOR0_CMASK) / 0x3c;
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->cb_color_cmask_bo[tmp] = reloc->robj;
break;
case CB_COLOR0_FMASK_SLICE:
case CB_COLOR1_FMASK_SLICE:
case CB_COLOR2_FMASK_SLICE:
case CB_COLOR3_FMASK_SLICE:
case CB_COLOR4_FMASK_SLICE:
case CB_COLOR5_FMASK_SLICE:
case CB_COLOR6_FMASK_SLICE:
case CB_COLOR7_FMASK_SLICE:
tmp = (reg - CB_COLOR0_FMASK_SLICE) / 0x3c;
track->cb_color_fmask_slice[tmp] = radeon_get_ib_value(p, idx);
break;
case CB_COLOR0_CMASK_SLICE:
case CB_COLOR1_CMASK_SLICE:
case CB_COLOR2_CMASK_SLICE:
case CB_COLOR3_CMASK_SLICE:
case CB_COLOR4_CMASK_SLICE:
case CB_COLOR5_CMASK_SLICE:
case CB_COLOR6_CMASK_SLICE:
case CB_COLOR7_CMASK_SLICE:
tmp = (reg - CB_COLOR0_CMASK_SLICE) / 0x3c;
track->cb_color_cmask_slice[tmp] = radeon_get_ib_value(p, idx);
break;
case CB_COLOR0_BASE:
case CB_COLOR1_BASE:
case CB_COLOR2_BASE:
case CB_COLOR3_BASE:
case CB_COLOR4_BASE:
case CB_COLOR5_BASE:
case CB_COLOR6_BASE:
case CB_COLOR7_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
tmp = (reg - CB_COLOR0_BASE) / 0x3c;
track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->cb_color_base_last[tmp] = ib[idx];
track->cb_color_bo[tmp] = reloc->robj;
break;
case CB_COLOR8_BASE:
case CB_COLOR9_BASE:
case CB_COLOR10_BASE:
case CB_COLOR11_BASE:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
tmp = ((reg - CB_COLOR8_BASE) / 0x1c) + 8;
track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx);
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->cb_color_base_last[tmp] = ib[idx];
track->cb_color_bo[tmp] = reloc->robj;
break;
case CB_IMMED0_BASE:
case CB_IMMED1_BASE:
case CB_IMMED2_BASE:
case CB_IMMED3_BASE:
case CB_IMMED4_BASE:
case CB_IMMED5_BASE:
case CB_IMMED6_BASE:
case CB_IMMED7_BASE:
case CB_IMMED8_BASE:
case CB_IMMED9_BASE:
case CB_IMMED10_BASE:
case CB_IMMED11_BASE:
case DB_HTILE_DATA_BASE:
case SQ_PGM_START_FS:
case SQ_PGM_START_ES:
case SQ_PGM_START_VS:
case SQ_PGM_START_GS:
case SQ_PGM_START_PS:
case SQ_PGM_START_HS:
case SQ_PGM_START_LS:
case SQ_CONST_MEM_BASE:
case SQ_ALU_CONST_CACHE_GS_0:
case SQ_ALU_CONST_CACHE_GS_1:
case SQ_ALU_CONST_CACHE_GS_2:
case SQ_ALU_CONST_CACHE_GS_3:
case SQ_ALU_CONST_CACHE_GS_4:
case SQ_ALU_CONST_CACHE_GS_5:
case SQ_ALU_CONST_CACHE_GS_6:
case SQ_ALU_CONST_CACHE_GS_7:
case SQ_ALU_CONST_CACHE_GS_8:
case SQ_ALU_CONST_CACHE_GS_9:
case SQ_ALU_CONST_CACHE_GS_10:
case SQ_ALU_CONST_CACHE_GS_11:
case SQ_ALU_CONST_CACHE_GS_12:
case SQ_ALU_CONST_CACHE_GS_13:
case SQ_ALU_CONST_CACHE_GS_14:
case SQ_ALU_CONST_CACHE_GS_15:
case SQ_ALU_CONST_CACHE_PS_0:
case SQ_ALU_CONST_CACHE_PS_1:
case SQ_ALU_CONST_CACHE_PS_2:
case SQ_ALU_CONST_CACHE_PS_3:
case SQ_ALU_CONST_CACHE_PS_4:
case SQ_ALU_CONST_CACHE_PS_5:
case SQ_ALU_CONST_CACHE_PS_6:
case SQ_ALU_CONST_CACHE_PS_7:
case SQ_ALU_CONST_CACHE_PS_8:
case SQ_ALU_CONST_CACHE_PS_9:
case SQ_ALU_CONST_CACHE_PS_10:
case SQ_ALU_CONST_CACHE_PS_11:
case SQ_ALU_CONST_CACHE_PS_12:
case SQ_ALU_CONST_CACHE_PS_13:
case SQ_ALU_CONST_CACHE_PS_14:
case SQ_ALU_CONST_CACHE_PS_15:
case SQ_ALU_CONST_CACHE_VS_0:
case SQ_ALU_CONST_CACHE_VS_1:
case SQ_ALU_CONST_CACHE_VS_2:
case SQ_ALU_CONST_CACHE_VS_3:
case SQ_ALU_CONST_CACHE_VS_4:
case SQ_ALU_CONST_CACHE_VS_5:
case SQ_ALU_CONST_CACHE_VS_6:
case SQ_ALU_CONST_CACHE_VS_7:
case SQ_ALU_CONST_CACHE_VS_8:
case SQ_ALU_CONST_CACHE_VS_9:
case SQ_ALU_CONST_CACHE_VS_10:
case SQ_ALU_CONST_CACHE_VS_11:
case SQ_ALU_CONST_CACHE_VS_12:
case SQ_ALU_CONST_CACHE_VS_13:
case SQ_ALU_CONST_CACHE_VS_14:
case SQ_ALU_CONST_CACHE_VS_15:
case SQ_ALU_CONST_CACHE_HS_0:
case SQ_ALU_CONST_CACHE_HS_1:
case SQ_ALU_CONST_CACHE_HS_2:
case SQ_ALU_CONST_CACHE_HS_3:
case SQ_ALU_CONST_CACHE_HS_4:
case SQ_ALU_CONST_CACHE_HS_5:
case SQ_ALU_CONST_CACHE_HS_6:
case SQ_ALU_CONST_CACHE_HS_7:
case SQ_ALU_CONST_CACHE_HS_8:
case SQ_ALU_CONST_CACHE_HS_9:
case SQ_ALU_CONST_CACHE_HS_10:
case SQ_ALU_CONST_CACHE_HS_11:
case SQ_ALU_CONST_CACHE_HS_12:
case SQ_ALU_CONST_CACHE_HS_13:
case SQ_ALU_CONST_CACHE_HS_14:
case SQ_ALU_CONST_CACHE_HS_15:
case SQ_ALU_CONST_CACHE_LS_0:
case SQ_ALU_CONST_CACHE_LS_1:
case SQ_ALU_CONST_CACHE_LS_2:
case SQ_ALU_CONST_CACHE_LS_3:
case SQ_ALU_CONST_CACHE_LS_4:
case SQ_ALU_CONST_CACHE_LS_5:
case SQ_ALU_CONST_CACHE_LS_6:
case SQ_ALU_CONST_CACHE_LS_7:
case SQ_ALU_CONST_CACHE_LS_8:
case SQ_ALU_CONST_CACHE_LS_9:
case SQ_ALU_CONST_CACHE_LS_10:
case SQ_ALU_CONST_CACHE_LS_11:
case SQ_ALU_CONST_CACHE_LS_12:
case SQ_ALU_CONST_CACHE_LS_13:
case SQ_ALU_CONST_CACHE_LS_14:
case SQ_ALU_CONST_CACHE_LS_15:
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
case SX_MEMORY_EXPORT_BASE:
if (p->rdev->family >= CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONFIG_REG "
"0x%04X\n", reg);
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONFIG_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
case CAYMAN_SX_SCATTER_EXPORT_BASE:
if (p->rdev->family < CHIP_CAYMAN) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
default:
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return -EINVAL;
}
return 0;
}
/**
* evergreen_check_texture_resource() - check if register is authorized or not
* @p: parser structure holding parsing context
* @idx: index into the cs buffer
* @texture: texture's bo structure
* @mipmap: mipmap's bo structure
*
* This function will check that the resource has valid field and that
* the texture and mipmap bo object are big enough to cover this resource.
*/
static int evergreen_check_texture_resource(struct radeon_cs_parser *p, u32 idx,
struct radeon_bo *texture,
struct radeon_bo *mipmap)
{
/* XXX fill in */
return 0;
}
static int evergreen_packet3_check(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt)
{
struct radeon_cs_reloc *reloc;
struct evergreen_cs_track *track;
volatile u32 *ib;
unsigned idx;
unsigned i;
unsigned start_reg, end_reg, reg;
int r;
u32 idx_value;
track = (struct evergreen_cs_track *)p->track;
ib = p->ib->ptr;
idx = pkt->idx + 1;
idx_value = radeon_get_ib_value(p, idx);
switch (pkt->opcode) {
case PACKET3_SET_PREDICATION:
{
int pred_op;
int tmp;
if (pkt->count != 1) {
DRM_ERROR("bad SET PREDICATION\n");
return -EINVAL;
}
tmp = radeon_get_ib_value(p, idx + 1);
pred_op = (tmp >> 16) & 0x7;
/* for the clear predicate operation */
if (pred_op == 0)
return 0;
if (pred_op > 2) {
DRM_ERROR("bad SET PREDICATION operation %d\n", pred_op);
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET PREDICATION\n");
return -EINVAL;
}
ib[idx + 0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx + 1] = tmp + (upper_32_bits(reloc->lobj.gpu_offset) & 0xff);
}
break;
case PACKET3_CONTEXT_CONTROL:
if (pkt->count != 1) {
DRM_ERROR("bad CONTEXT_CONTROL\n");
return -EINVAL;
}
break;
case PACKET3_INDEX_TYPE:
case PACKET3_NUM_INSTANCES:
case PACKET3_CLEAR_STATE:
if (pkt->count) {
DRM_ERROR("bad INDEX_TYPE/NUM_INSTANCES/CLEAR_STATE\n");
return -EINVAL;
}
break;
case CAYMAN_PACKET3_DEALLOC_STATE:
if (p->rdev->family < CHIP_CAYMAN) {
DRM_ERROR("bad PACKET3_DEALLOC_STATE\n");
return -EINVAL;
}
if (pkt->count) {
DRM_ERROR("bad INDEX_TYPE/NUM_INSTANCES/CLEAR_STATE\n");
return -EINVAL;
}
break;
case PACKET3_INDEX_BASE:
if (pkt->count != 1) {
DRM_ERROR("bad INDEX_BASE\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad INDEX_BASE\n");
return -EINVAL;
}
ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DRAW_INDEX:
if (pkt->count != 3) {
DRM_ERROR("bad DRAW_INDEX\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad DRAW_INDEX\n");
return -EINVAL;
}
ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DRAW_INDEX_2:
if (pkt->count != 4) {
DRM_ERROR("bad DRAW_INDEX_2\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad DRAW_INDEX_2\n");
return -EINVAL;
}
ib[idx+1] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DRAW_INDEX_AUTO:
if (pkt->count != 1) {
DRM_ERROR("bad DRAW_INDEX_AUTO\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream %d\n", __func__, __LINE__, idx);
return r;
}
break;
case PACKET3_DRAW_INDEX_MULTI_AUTO:
if (pkt->count != 2) {
DRM_ERROR("bad DRAW_INDEX_MULTI_AUTO\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream %d\n", __func__, __LINE__, idx);
return r;
}
break;
case PACKET3_DRAW_INDEX_IMMD:
if (pkt->count < 2) {
DRM_ERROR("bad DRAW_INDEX_IMMD\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DRAW_INDEX_OFFSET:
if (pkt->count != 2) {
DRM_ERROR("bad DRAW_INDEX_OFFSET\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DRAW_INDEX_OFFSET_2:
if (pkt->count != 3) {
DRM_ERROR("bad DRAW_INDEX_OFFSET_2\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_DISPATCH_DIRECT:
if (pkt->count != 3) {
DRM_ERROR("bad DISPATCH_DIRECT\n");
return -EINVAL;
}
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream %d\n", __func__, __LINE__, idx);
return r;
}
break;
case PACKET3_DISPATCH_INDIRECT:
if (pkt->count != 1) {
DRM_ERROR("bad DISPATCH_INDIRECT\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad DISPATCH_INDIRECT\n");
return -EINVAL;
}
ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);
r = evergreen_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
break;
case PACKET3_WAIT_REG_MEM:
if (pkt->count != 5) {
DRM_ERROR("bad WAIT_REG_MEM\n");
return -EINVAL;
}
/* bit 4 is reg (0) or mem (1) */
if (idx_value & 0x10) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad WAIT_REG_MEM\n");
return -EINVAL;
}
ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
}
break;
case PACKET3_SURFACE_SYNC:
if (pkt->count != 3) {
DRM_ERROR("bad SURFACE_SYNC\n");
return -EINVAL;
}
/* 0xffffffff/0x0 is flush all cache flag */
if (radeon_get_ib_value(p, idx + 1) != 0xffffffff ||
radeon_get_ib_value(p, idx + 2) != 0) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SURFACE_SYNC\n");
return -EINVAL;
}
ib[idx+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
}
break;
case PACKET3_EVENT_WRITE:
if (pkt->count != 2 && pkt->count != 0) {
DRM_ERROR("bad EVENT_WRITE\n");
return -EINVAL;
}
if (pkt->count) {
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad EVENT_WRITE\n");
return -EINVAL;
}
ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
}
break;
case PACKET3_EVENT_WRITE_EOP:
if (pkt->count != 4) {
DRM_ERROR("bad EVENT_WRITE_EOP\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad EVENT_WRITE_EOP\n");
return -EINVAL;
}
ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
break;
case PACKET3_EVENT_WRITE_EOS:
if (pkt->count != 3) {
DRM_ERROR("bad EVENT_WRITE_EOS\n");
return -EINVAL;
}
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad EVENT_WRITE_EOS\n");
return -EINVAL;
}
ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);
ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
break;
case PACKET3_SET_CONFIG_REG:
start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CONFIG_REG_START) ||
(start_reg >= PACKET3_SET_CONFIG_REG_END) ||
(end_reg >= PACKET3_SET_CONFIG_REG_END)) {
DRM_ERROR("bad PACKET3_SET_CONFIG_REG\n");
return -EINVAL;
}
for (i = 0; i < pkt->count; i++) {
reg = start_reg + (4 * i);
r = evergreen_cs_check_reg(p, reg, idx+1+i);
if (r)
return r;
}
break;
case PACKET3_SET_CONTEXT_REG:
start_reg = (idx_value << 2) + PACKET3_SET_CONTEXT_REG_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CONTEXT_REG_START) ||
(start_reg >= PACKET3_SET_CONTEXT_REG_END) ||
(end_reg >= PACKET3_SET_CONTEXT_REG_END)) {
DRM_ERROR("bad PACKET3_SET_CONTEXT_REG\n");
return -EINVAL;
}
for (i = 0; i < pkt->count; i++) {
reg = start_reg + (4 * i);
r = evergreen_cs_check_reg(p, reg, idx+1+i);
if (r)
return r;
}
break;
case PACKET3_SET_RESOURCE:
if (pkt->count % 8) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
start_reg = (idx_value << 2) + PACKET3_SET_RESOURCE_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_RESOURCE_START) ||
(start_reg >= PACKET3_SET_RESOURCE_END) ||
(end_reg >= PACKET3_SET_RESOURCE_END)) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
for (i = 0; i < (pkt->count / 8); i++) {
struct radeon_bo *texture, *mipmap;
u32 size, offset;
switch (G__SQ_CONSTANT_TYPE(radeon_get_ib_value(p, idx+1+(i*8)+7))) {
case SQ_TEX_VTX_VALID_TEXTURE:
/* tex base */
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE (tex)\n");
return -EINVAL;
}
ib[idx+1+(i*8)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (!p->keep_tiling_flags) {
ib[idx+1+(i*8)+1] |=
TEX_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
ib[idx+1+(i*8)+6] |=
TEX_TILE_SPLIT(evergreen_cs_get_tile_split(track->row_size));
ib[idx+1+(i*8)+7] |=
TEX_NUM_BANKS(evergreen_cs_get_num_banks(track->nbanks));
}
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
}
texture = reloc->robj;
/* tex mip base */
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE (tex)\n");
return -EINVAL;
}
ib[idx+1+(i*8)+3] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
mipmap = reloc->robj;
r = evergreen_check_texture_resource(p, idx+1+(i*8),
texture, mipmap);
if (r)
return r;
break;
case SQ_TEX_VTX_VALID_BUFFER:
/* vtx base */
r = evergreen_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE (vtx)\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+1+(i*8)+0);
size = radeon_get_ib_value(p, idx+1+(i*8)+1);
if (p->rdev && (size + offset) > radeon_bo_size(reloc->robj)) {
/* force size to size of the buffer */
dev_warn(p->dev, "vbo resource seems too big for the bo\n");
ib[idx+1+(i*8)+1] = radeon_bo_size(reloc->robj);
}
ib[idx+1+(i*8)+0] += (u32)((reloc->lobj.gpu_offset) & 0xffffffff);
ib[idx+1+(i*8)+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;
break;
case SQ_TEX_VTX_INVALID_TEXTURE:
case SQ_TEX_VTX_INVALID_BUFFER:
default:
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
}
break;
case PACKET3_SET_ALU_CONST:
/* XXX fix me ALU const buffers only */
break;
case PACKET3_SET_BOOL_CONST:
start_reg = (idx_value << 2) + PACKET3_SET_BOOL_CONST_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_BOOL_CONST_START) ||
(start_reg >= PACKET3_SET_BOOL_CONST_END) ||
(end_reg >= PACKET3_SET_BOOL_CONST_END)) {
DRM_ERROR("bad SET_BOOL_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_LOOP_CONST:
start_reg = (idx_value << 2) + PACKET3_SET_LOOP_CONST_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_LOOP_CONST_START) ||
(start_reg >= PACKET3_SET_LOOP_CONST_END) ||
(end_reg >= PACKET3_SET_LOOP_CONST_END)) {
DRM_ERROR("bad SET_LOOP_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_CTL_CONST:
start_reg = (idx_value << 2) + PACKET3_SET_CTL_CONST_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CTL_CONST_START) ||
(start_reg >= PACKET3_SET_CTL_CONST_END) ||
(end_reg >= PACKET3_SET_CTL_CONST_END)) {
DRM_ERROR("bad SET_CTL_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_SAMPLER:
if (pkt->count % 3) {
DRM_ERROR("bad SET_SAMPLER\n");
return -EINVAL;
}
start_reg = (idx_value << 2) + PACKET3_SET_SAMPLER_START;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_SAMPLER_START) ||
(start_reg >= PACKET3_SET_SAMPLER_END) ||
(end_reg >= PACKET3_SET_SAMPLER_END)) {
DRM_ERROR("bad SET_SAMPLER\n");
return -EINVAL;
}
break;
case PACKET3_NOP:
break;
default:
DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode);
return -EINVAL;
}
return 0;
}
int evergreen_cs_parse(struct radeon_cs_parser *p)
{
struct radeon_cs_packet pkt;
struct evergreen_cs_track *track;
u32 tmp;
int r;
if (p->track == NULL) {
/* initialize tracker, we are in kms */
track = kzalloc(sizeof(*track), GFP_KERNEL);
if (track == NULL)
return -ENOMEM;
evergreen_cs_track_init(track);
if (p->rdev->family >= CHIP_CAYMAN)
tmp = p->rdev->config.cayman.tile_config;
else
tmp = p->rdev->config.evergreen.tile_config;
switch (tmp & 0xf) {
case 0:
track->npipes = 1;
break;
case 1:
default:
track->npipes = 2;
break;
case 2:
track->npipes = 4;
break;
case 3:
track->npipes = 8;
break;
}
switch ((tmp & 0xf0) >> 4) {
case 0:
track->nbanks = 4;
break;
case 1:
default:
track->nbanks = 8;
break;
case 2:
track->nbanks = 16;
break;
}
switch ((tmp & 0xf00) >> 8) {
case 0:
track->group_size = 256;
break;
case 1:
default:
track->group_size = 512;
break;
}
switch ((tmp & 0xf000) >> 12) {
case 0:
track->row_size = 1;
break;
case 1:
default:
track->row_size = 2;
break;
case 2:
track->row_size = 4;
break;
}
p->track = track;
}
do {
r = evergreen_cs_packet_parse(p, &pkt, p->idx);
if (r) {
kfree(p->track);
p->track = NULL;
return r;
}
p->idx += pkt.count + 2;
switch (pkt.type) {
case PACKET_TYPE0:
r = evergreen_cs_parse_packet0(p, &pkt);
break;
case PACKET_TYPE2:
break;
case PACKET_TYPE3:
r = evergreen_packet3_check(p, &pkt);
break;
default:
DRM_ERROR("Unknown packet type %d !\n", pkt.type);
kfree(p->track);
p->track = NULL;
return -EINVAL;
}
if (r) {
kfree(p->track);
p->track = NULL;
return r;
}
} while (p->idx < p->chunks[p->chunk_ib_idx].length_dw);
#if 0
for (r = 0; r < p->ib->length_dw; r++) {
printk(KERN_INFO "%05d 0x%08X\n", r, p->ib->ptr[r]);
mdelay(1);
}
#endif
kfree(p->track);
p->track = NULL;
return 0;
}