summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/radeon
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/radeon')
-rw-r--r--drivers/gpu/drm/radeon/atom.c15
-rw-r--r--drivers/gpu/drm/radeon/atom.h1
-rw-r--r--drivers/gpu/drm/radeon/atombios_crtc.c8
-rw-r--r--drivers/gpu/drm/radeon/atombios_dp.c3
-rw-r--r--drivers/gpu/drm/radeon/atombios_encoders.c4
-rw-r--r--drivers/gpu/drm/radeon/evergreen_cs.c98
-rw-r--r--drivers/gpu/drm/radeon/evergreend.h8
-rw-r--r--drivers/gpu/drm/radeon/r600_cs.c391
-rw-r--r--drivers/gpu/drm/radeon/r600d.h8
-rw-r--r--drivers/gpu/drm/radeon/radeon_object.c13
-rw-r--r--drivers/gpu/drm/radeon/reg_srcs/cayman1
-rw-r--r--drivers/gpu/drm/radeon/reg_srcs/evergreen1
-rw-r--r--drivers/gpu/drm/radeon/reg_srcs/r6001
13 files changed, 422 insertions, 130 deletions
diff --git a/drivers/gpu/drm/radeon/atom.c b/drivers/gpu/drm/radeon/atom.c
index d1bd239cd9e..5ce9bf51a8d 100644
--- a/drivers/gpu/drm/radeon/atom.c
+++ b/drivers/gpu/drm/radeon/atom.c
@@ -1306,8 +1306,11 @@ struct atom_context *atom_parse(struct card_info *card, void *bios)
int atom_asic_init(struct atom_context *ctx)
{
+ struct radeon_device *rdev = ctx->card->dev->dev_private;
int hwi = CU16(ctx->data_table + ATOM_DATA_FWI_PTR);
uint32_t ps[16];
+ int ret;
+
memset(ps, 0, 64);
ps[0] = cpu_to_le32(CU32(hwi + ATOM_FWI_DEFSCLK_PTR));
@@ -1317,7 +1320,17 @@ int atom_asic_init(struct atom_context *ctx)
if (!CU16(ctx->cmd_table + 4 + 2 * ATOM_CMD_INIT))
return 1;
- return atom_execute_table(ctx, ATOM_CMD_INIT, ps);
+ ret = atom_execute_table(ctx, ATOM_CMD_INIT, ps);
+ if (ret)
+ return ret;
+
+ memset(ps, 0, 64);
+
+ if (rdev->family < CHIP_R600) {
+ if (CU16(ctx->cmd_table + 4 + 2 * ATOM_CMD_SPDFANCNTL))
+ atom_execute_table(ctx, ATOM_CMD_SPDFANCNTL, ps);
+ }
+ return ret;
}
void atom_destroy(struct atom_context *ctx)
diff --git a/drivers/gpu/drm/radeon/atom.h b/drivers/gpu/drm/radeon/atom.h
index 93cfe2086ba..25fea631dad 100644
--- a/drivers/gpu/drm/radeon/atom.h
+++ b/drivers/gpu/drm/radeon/atom.h
@@ -44,6 +44,7 @@
#define ATOM_CMD_SETSCLK 0x0A
#define ATOM_CMD_SETMCLK 0x0B
#define ATOM_CMD_SETPCLK 0x0C
+#define ATOM_CMD_SPDFANCNTL 0x39
#define ATOM_DATA_FWI_PTR 0xC
#define ATOM_DATA_IIO_PTR 0x32
diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index 083b3eada00..b5ff1f7b6f7 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -588,8 +588,8 @@ static u32 atombios_adjust_pll(struct drm_crtc *crtc,
if (encoder->crtc == crtc) {
radeon_encoder = to_radeon_encoder(encoder);
connector = radeon_get_connector_for_encoder(encoder);
- if (connector && connector->display_info.bpc)
- bpc = connector->display_info.bpc;
+ /* if (connector && connector->display_info.bpc)
+ bpc = connector->display_info.bpc; */
encoder_mode = atombios_get_encoder_mode(encoder);
is_duallink = radeon_dig_monitor_is_duallink(encoder, mode->clock);
if ((radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT | ATOM_DEVICE_DFP_SUPPORT)) ||
@@ -965,7 +965,9 @@ static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode
struct radeon_connector_atom_dig *dig_connector =
radeon_connector->con_priv;
int dp_clock;
- bpc = connector->display_info.bpc;
+
+ /* if (connector->display_info.bpc)
+ bpc = connector->display_info.bpc; */
switch (encoder_mode) {
case ATOM_ENCODER_MODE_DP_MST:
diff --git a/drivers/gpu/drm/radeon/atombios_dp.c b/drivers/gpu/drm/radeon/atombios_dp.c
index 6c62be22680..c57d85664e7 100644
--- a/drivers/gpu/drm/radeon/atombios_dp.c
+++ b/drivers/gpu/drm/radeon/atombios_dp.c
@@ -405,10 +405,13 @@ static void dp_get_adjust_train(u8 link_status[DP_LINK_STATUS_SIZE],
/* get bpc from the EDID */
static int convert_bpc_to_bpp(int bpc)
{
+#if 0
if (bpc == 0)
return 24;
else
return bpc * 3;
+#endif
+ return 24;
}
/* get the max pix clock supported by the link rate and lane num */
diff --git a/drivers/gpu/drm/radeon/atombios_encoders.c b/drivers/gpu/drm/radeon/atombios_encoders.c
index 468b874336f..e607c4d7dd9 100644
--- a/drivers/gpu/drm/radeon/atombios_encoders.c
+++ b/drivers/gpu/drm/radeon/atombios_encoders.c
@@ -541,7 +541,7 @@ atombios_dig_encoder_setup(struct drm_encoder *encoder, int action, int panel_mo
dp_clock = dig_connector->dp_clock;
dp_lane_count = dig_connector->dp_lane_count;
hpd_id = radeon_connector->hpd.hpd;
- bpc = connector->display_info.bpc;
+ /* bpc = connector->display_info.bpc; */
}
/* no dig encoder assigned */
@@ -1159,7 +1159,7 @@ atombios_external_encoder_setup(struct drm_encoder *encoder,
dp_lane_count = dig_connector->dp_lane_count;
connector_object_id =
(radeon_connector->connector_object_id & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT;
- bpc = connector->display_info.bpc;
+ /* bpc = connector->display_info.bpc; */
}
memset(&args, 0, sizeof(args));
diff --git a/drivers/gpu/drm/radeon/evergreen_cs.c b/drivers/gpu/drm/radeon/evergreen_cs.c
index a58b37a2e65..70089d32b80 100644
--- a/drivers/gpu/drm/radeon/evergreen_cs.c
+++ b/drivers/gpu/drm/radeon/evergreen_cs.c
@@ -80,6 +80,9 @@ struct evergreen_cs_track {
bool cb_dirty;
bool db_dirty;
bool streamout_dirty;
+ u32 htile_offset;
+ u32 htile_surface;
+ struct radeon_bo *htile_bo;
};
static u32 evergreen_cs_get_aray_mode(u32 tiling_flags)
@@ -144,6 +147,9 @@ static void evergreen_cs_track_init(struct evergreen_cs_track *track)
track->db_s_read_bo = NULL;
track->db_s_write_bo = NULL;
track->db_dirty = true;
+ track->htile_bo = NULL;
+ track->htile_offset = 0xFFFFFFFF;
+ track->htile_surface = 0;
for (i = 0; i < 4; i++) {
track->vgt_strmout_size[i] = 0;
@@ -444,6 +450,62 @@ static int evergreen_cs_track_validate_cb(struct radeon_cs_parser *p, unsigned i
return 0;
}
+static int evergreen_cs_track_validate_htile(struct radeon_cs_parser *p,
+ unsigned nbx, unsigned nby)
+{
+ struct evergreen_cs_track *track = p->track;
+ unsigned long size;
+
+ if (track->htile_bo == NULL) {
+ dev_warn(p->dev, "%s:%d htile enabled without htile surface 0x%08x\n",
+ __func__, __LINE__, track->db_z_info);
+ return -EINVAL;
+ }
+
+ if (G_028ABC_LINEAR(track->htile_surface)) {
+ /* pitch must be 16 htiles aligned == 16 * 8 pixel aligned */
+ nbx = round_up(nbx, 16 * 8);
+ /* height is npipes htiles aligned == npipes * 8 pixel aligned */
+ nby = round_up(nby, track->npipes * 8);
+ } else {
+ switch (track->npipes) {
+ case 8:
+ nbx = round_up(nbx, 64 * 8);
+ nby = round_up(nby, 64 * 8);
+ break;
+ case 4:
+ nbx = round_up(nbx, 64 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 2:
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 1:
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ default:
+ dev_warn(p->dev, "%s:%d invalid num pipes %d\n",
+ __func__, __LINE__, track->npipes);
+ return -EINVAL;
+ }
+ }
+ /* compute number of htile */
+ nbx = nbx / 8;
+ nby = nby / 8;
+ size = nbx * nby * 4;
+ size += track->htile_offset;
+
+ if (size > radeon_bo_size(track->htile_bo)) {
+ dev_warn(p->dev, "%s:%d htile surface too small %ld for %ld (%d %d)\n",
+ __func__, __LINE__, radeon_bo_size(track->htile_bo),
+ size, nbx, nby);
+ return -EINVAL;
+ }
+ return 0;
+}
+
static int evergreen_cs_track_validate_stencil(struct radeon_cs_parser *p)
{
struct evergreen_cs_track *track = p->track;
@@ -530,6 +592,14 @@ static int evergreen_cs_track_validate_stencil(struct radeon_cs_parser *p)
return -EINVAL;
}
+ /* hyperz */
+ if (G_028040_TILE_SURFACE_ENABLE(track->db_z_info)) {
+ r = evergreen_cs_track_validate_htile(p, surf.nbx, surf.nby);
+ if (r) {
+ return r;
+ }
+ }
+
return 0;
}
@@ -617,6 +687,14 @@ static int evergreen_cs_track_validate_depth(struct radeon_cs_parser *p)
return -EINVAL;
}
+ /* hyperz */
+ if (G_028040_TILE_SURFACE_ENABLE(track->db_z_info)) {
+ r = evergreen_cs_track_validate_htile(p, surf.nbx, surf.nby);
+ if (r) {
+ return r;
+ }
+ }
+
return 0;
}
@@ -850,7 +928,7 @@ static int evergreen_cs_track_check(struct radeon_cs_parser *p)
return r;
}
/* Check depth buffer */
- if (G_028800_Z_WRITE_ENABLE(track->db_depth_control)) {
+ if (G_028800_Z_ENABLE(track->db_depth_control)) {
r = evergreen_cs_track_validate_depth(p);
if (r)
return r;
@@ -1616,6 +1694,23 @@ static int evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
track->cb_color_bo[tmp] = reloc->robj;
track->cb_dirty = true;
break;
+ case DB_HTILE_DATA_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->htile_offset = radeon_get_ib_value(p, idx);
+ ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+ track->htile_bo = reloc->robj;
+ track->db_dirty = true;
+ break;
+ case DB_HTILE_SURFACE:
+ /* 8x8 only */
+ track->htile_surface = radeon_get_ib_value(p, idx);
+ track->db_dirty = true;
+ break;
case CB_IMMED0_BASE:
case CB_IMMED1_BASE:
case CB_IMMED2_BASE:
@@ -1628,7 +1723,6 @@ static int evergreen_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
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:
diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h
index eb5708c7159..b4eefc355f1 100644
--- a/drivers/gpu/drm/radeon/evergreend.h
+++ b/drivers/gpu/drm/radeon/evergreend.h
@@ -991,6 +991,14 @@
#define G_028008_SLICE_MAX(x) (((x) >> 13) & 0x7FF)
#define C_028008_SLICE_MAX 0xFF001FFF
#define DB_HTILE_DATA_BASE 0x28014
+#define DB_HTILE_SURFACE 0x28abc
+#define S_028ABC_HTILE_WIDTH(x) (((x) & 0x1) << 0)
+#define G_028ABC_HTILE_WIDTH(x) (((x) >> 0) & 0x1)
+#define C_028ABC_HTILE_WIDTH 0xFFFFFFFE
+#define S_028ABC_HTILE_HEIGHT(x) (((x) & 0x1) << 1)
+#define G_028ABC_HTILE_HEIGHT(x) (((x) >> 1) & 0x1)
+#define C_028ABC_HTILE_HEIGHT 0xFFFFFFFD
+#define G_028ABC_LINEAR(x) (((x) >> 2) & 0x1)
#define DB_Z_INFO 0x28040
# define Z_ARRAY_MODE(x) ((x) << 4)
# define DB_TILE_SPLIT(x) (((x) & 0x7) << 8)
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index 0ec3f205f9c..b8e12af304a 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -78,6 +78,9 @@ struct r600_cs_track {
bool cb_dirty;
bool db_dirty;
bool streamout_dirty;
+ struct radeon_bo *htile_bo;
+ u64 htile_offset;
+ u32 htile_surface;
};
#define FMT_8_BIT(fmt, vc) [fmt] = { 1, 1, 1, vc, CHIP_R600 }
@@ -321,6 +324,9 @@ static void r600_cs_track_init(struct r600_cs_track *track)
track->db_depth_size_idx = 0;
track->db_depth_control = 0xFFFFFFFF;
track->db_dirty = true;
+ track->htile_bo = NULL;
+ track->htile_offset = 0xFFFFFFFF;
+ track->htile_surface = 0;
for (i = 0; i < 4; i++) {
track->vgt_strmout_size[i] = 0;
@@ -455,12 +461,256 @@ static int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
return 0;
}
+static int r600_cs_track_validate_db(struct radeon_cs_parser *p)
+{
+ struct r600_cs_track *track = p->track;
+ u32 nviews, bpe, ntiles, size, slice_tile_max, tmp;
+ u32 height_align, pitch_align, depth_align;
+ u32 pitch = 8192;
+ u32 height = 8192;
+ u64 base_offset, base_align;
+ struct array_mode_checker array_check;
+ int array_mode;
+ volatile u32 *ib = p->ib->ptr;
+
+
+ if (track->db_bo == NULL) {
+ dev_warn(p->dev, "z/stencil with no depth buffer\n");
+ return -EINVAL;
+ }
+ switch (G_028010_FORMAT(track->db_depth_info)) {
+ case V_028010_DEPTH_16:
+ bpe = 2;
+ break;
+ case V_028010_DEPTH_X8_24:
+ case V_028010_DEPTH_8_24:
+ case V_028010_DEPTH_X8_24_FLOAT:
+ case V_028010_DEPTH_8_24_FLOAT:
+ case V_028010_DEPTH_32_FLOAT:
+ bpe = 4;
+ break;
+ case V_028010_DEPTH_X24_8_32_FLOAT:
+ bpe = 8;
+ break;
+ default:
+ dev_warn(p->dev, "z/stencil with invalid format %d\n", G_028010_FORMAT(track->db_depth_info));
+ return -EINVAL;
+ }
+ if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
+ if (!track->db_depth_size_idx) {
+ dev_warn(p->dev, "z/stencil buffer size not set\n");
+ return -EINVAL;
+ }
+ tmp = radeon_bo_size(track->db_bo) - track->db_offset;
+ tmp = (tmp / bpe) >> 6;
+ if (!tmp) {
+ dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %ld)\n",
+ track->db_depth_size, bpe, track->db_offset,
+ radeon_bo_size(track->db_bo));
+ return -EINVAL;
+ }
+ ib[track->db_depth_size_idx] = S_028000_SLICE_TILE_MAX(tmp - 1) | (track->db_depth_size & 0x3FF);
+ } else {
+ size = radeon_bo_size(track->db_bo);
+ /* pitch in pixels */
+ pitch = (G_028000_PITCH_TILE_MAX(track->db_depth_size) + 1) * 8;
+ slice_tile_max = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
+ slice_tile_max *= 64;
+ height = slice_tile_max / pitch;
+ if (height > 8192)
+ height = 8192;
+ base_offset = track->db_bo_mc + track->db_offset;
+ array_mode = G_028010_ARRAY_MODE(track->db_depth_info);
+ array_check.array_mode = array_mode;
+ array_check.group_size = track->group_size;
+ array_check.nbanks = track->nbanks;
+ array_check.npipes = track->npipes;
+ array_check.nsamples = track->nsamples;
+ array_check.blocksize = bpe;
+ if (r600_get_array_mode_alignment(&array_check,
+ &pitch_align, &height_align, &depth_align, &base_align)) {
+ dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
+ G_028010_ARRAY_MODE(track->db_depth_info),
+ track->db_depth_info);
+ return -EINVAL;
+ }
+ switch (array_mode) {
+ case V_028010_ARRAY_1D_TILED_THIN1:
+ /* don't break userspace */
+ height &= ~0x7;
+ break;
+ case V_028010_ARRAY_2D_TILED_THIN1:
+ break;
+ default:
+ dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
+ G_028010_ARRAY_MODE(track->db_depth_info),
+ track->db_depth_info);
+ return -EINVAL;
+ }
+
+ if (!IS_ALIGNED(pitch, pitch_align)) {
+ dev_warn(p->dev, "%s:%d db pitch (%d, 0x%x, %d) invalid\n",
+ __func__, __LINE__, pitch, pitch_align, array_mode);
+ return -EINVAL;
+ }
+ if (!IS_ALIGNED(height, height_align)) {
+ dev_warn(p->dev, "%s:%d db height (%d, 0x%x, %d) invalid\n",
+ __func__, __LINE__, height, height_align, array_mode);
+ return -EINVAL;
+ }
+ if (!IS_ALIGNED(base_offset, base_align)) {
+ dev_warn(p->dev, "%s offset 0x%llx, 0x%llx, %d not aligned\n", __func__,
+ base_offset, base_align, array_mode);
+ return -EINVAL;
+ }
+
+ ntiles = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
+ nviews = G_028004_SLICE_MAX(track->db_depth_view) + 1;
+ tmp = ntiles * bpe * 64 * nviews;
+ if ((tmp + track->db_offset) > radeon_bo_size(track->db_bo)) {
+ dev_warn(p->dev, "z/stencil buffer (%d) too small (0x%08X %d %d %d -> %u have %lu)\n",
+ array_mode,
+ track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
+ radeon_bo_size(track->db_bo));
+ return -EINVAL;
+ }
+ }
+
+ /* hyperz */
+ if (G_028010_TILE_SURFACE_ENABLE(track->db_depth_info)) {
+ unsigned long size;
+ unsigned nbx, nby;
+
+ if (track->htile_bo == NULL) {
+ dev_warn(p->dev, "%s:%d htile enabled without htile surface 0x%08x\n",
+ __func__, __LINE__, track->db_depth_info);
+ return -EINVAL;
+ }
+ if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
+ dev_warn(p->dev, "%s:%d htile can't be enabled with bogus db_depth_size 0x%08x\n",
+ __func__, __LINE__, track->db_depth_size);
+ return -EINVAL;
+ }
+
+ nbx = pitch;
+ nby = height;
+ if (G_028D24_LINEAR(track->htile_surface)) {
+ /* nbx must be 16 htiles aligned == 16 * 8 pixel aligned */
+ nbx = round_up(nbx, 16 * 8);
+ /* nby is npipes htiles aligned == npipes * 8 pixel aligned */
+ nby = round_up(nby, track->npipes * 8);
+ } else {
+ /* htile widht & nby (8 or 4) make 2 bits number */
+ tmp = track->htile_surface & 3;
+ /* align is htile align * 8, htile align vary according to
+ * number of pipe and tile width and nby
+ */
+ switch (track->npipes) {
+ case 8:
+ switch (tmp) {
+ case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
+ nbx = round_up(nbx, 64 * 8);
+ nby = round_up(nby, 64 * 8);
+ break;
+ case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
+ case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 64 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ default:
+ return -EINVAL;
+ }
+ break;
+ case 4:
+ switch (tmp) {
+ case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
+ nbx = round_up(nbx, 64 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
+ case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ default:
+ return -EINVAL;
+ }
+ break;
+ case 2:
+ switch (tmp) {
+ case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 32 * 8);
+ break;
+ case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
+ case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 16 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ default:
+ return -EINVAL;
+ }
+ break;
+ case 1:
+ switch (tmp) {
+ case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
+ nbx = round_up(nbx, 32 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
+ case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 16 * 8);
+ nby = round_up(nby, 16 * 8);
+ break;
+ case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
+ nbx = round_up(nbx, 16 * 8);
+ nby = round_up(nby, 8 * 8);
+ break;
+ default:
+ return -EINVAL;
+ }
+ break;
+ default:
+ dev_warn(p->dev, "%s:%d invalid num pipes %d\n",
+ __func__, __LINE__, track->npipes);
+ return -EINVAL;
+ }
+ }
+ /* compute number of htile */
+ nbx = G_028D24_HTILE_WIDTH(track->htile_surface) ? nbx / 8 : nbx / 4;
+ nby = G_028D24_HTILE_HEIGHT(track->htile_surface) ? nby / 8 : nby / 4;
+ size = nbx * nby * 4;
+ size += track->htile_offset;
+
+ if (size > radeon_bo_size(track->htile_bo)) {
+ dev_warn(p->dev, "%s:%d htile surface too small %ld for %ld (%d %d)\n",
+ __func__, __LINE__, radeon_bo_size(track->htile_bo),
+ size, nbx, nby);
+ return -EINVAL;
+ }
+ }
+
+ track->db_dirty = false;
+ return 0;
+}
+
static int r600_cs_track_check(struct radeon_cs_parser *p)
{
struct r600_cs_track *track = p->track;
u32 tmp;
int r, i;
- volatile u32 *ib = p->ib->ptr;
/* on legacy kernel we don't perform advanced check */
if (p->rdev == NULL)
@@ -513,124 +763,14 @@ static int r600_cs_track_check(struct radeon_cs_parser *p)
track->cb_dirty = false;
}
- if (track->db_dirty) {
- /* Check depth buffer */
- if (G_028800_STENCIL_ENABLE(track->db_depth_control) ||
- G_028800_Z_ENABLE(track->db_depth_control)) {
- u32 nviews, bpe, ntiles, size, slice_tile_max;
- u32 height, height_align, pitch, pitch_align, depth_align;
- u64 base_offset, base_align;
- struct array_mode_checker array_check;
- int array_mode;
-
- if (track->db_bo == NULL) {
- dev_warn(p->dev, "z/stencil with no depth buffer\n");
- return -EINVAL;
- }
- if (G_028010_TILE_SURFACE_ENABLE(track->db_depth_info)) {
- dev_warn(p->dev, "this kernel doesn't support z/stencil htile\n");
- return -EINVAL;
- }
- switch (G_028010_FORMAT(track->db_depth_info)) {
- case V_028010_DEPTH_16:
- bpe = 2;
- break;
- case V_028010_DEPTH_X8_24:
- case V_028010_DEPTH_8_24:
- case V_028010_DEPTH_X8_24_FLOAT:
- case V_028010_DEPTH_8_24_FLOAT:
- case V_028010_DEPTH_32_FLOAT:
- bpe = 4;
- break;
- case V_028010_DEPTH_X24_8_32_FLOAT:
- bpe = 8;
- break;
- default:
- dev_warn(p->dev, "z/stencil with invalid format %d\n", G_028010_FORMAT(track->db_depth_info));
- return -EINVAL;
- }
- if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
- if (!track->db_depth_size_idx) {
- dev_warn(p->dev, "z/stencil buffer size not set\n");
- return -EINVAL;
- }
- tmp = radeon_bo_size(track->db_bo) - track->db_offset;
- tmp = (tmp / bpe) >> 6;
- if (!tmp) {
- dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %ld)\n",
- track->db_depth_size, bpe, track->db_offset,
- radeon_bo_size(track->db_bo));
- return -EINVAL;
- }
- ib[track->db_depth_size_idx] = S_028000_SLICE_TILE_MAX(tmp - 1) | (track->db_depth_size & 0x3FF);
- } else {
- size = radeon_bo_size(track->db_bo);
- /* pitch in pixels */
- pitch = (G_028000_PITCH_TILE_MAX(track->db_depth_size) + 1) * 8;
- slice_tile_max = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
- slice_tile_max *= 64;
- height = slice_tile_max / pitch;
- if (height > 8192)
- height = 8192;
- base_offset = track->db_bo_mc + track->db_offset;
- array_mode = G_028010_ARRAY_MODE(track->db_depth_info);
- array_check.array_mode = array_mode;
- array_check.group_size = track->group_size;
- array_check.nbanks = track->nbanks;
- array_check.npipes = track->npipes;
- array_check.nsamples = track->nsamples;
- array_check.blocksize = bpe;
- if (r600_get_array_mode_alignment(&array_check,
- &pitch_align, &height_align, &depth_align, &base_align)) {
- dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
- G_028010_ARRAY_MODE(track->db_depth_info),
- track->db_depth_info);
- return -EINVAL;
- }
- switch (array_mode) {
- case V_028010_ARRAY_1D_TILED_THIN1:
- /* don't break userspace */
- height &= ~0x7;
- break;
- case V_028010_ARRAY_2D_TILED_THIN1:
- break;
- default:
- dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
- G_028010_ARRAY_MODE(track->db_depth_info),
- track->db_depth_info);
- return -EINVAL;
- }
-
- if (!IS_ALIGNED(pitch, pitch_align)) {
- dev_warn(p->dev, "%s:%d db pitch (%d, 0x%x, %d) invalid\n",
- __func__, __LINE__, pitch, pitch_align, array_mode);
- return -EINVAL;
- }
- if (!IS_ALIGNED(height, height_align)) {
- dev_warn(p->dev, "%s:%d db height (%d, 0x%x, %d) invalid\n",
- __func__, __LINE__, height, height_align, array_mode);
- return -EINVAL;
- }
- if (!IS_ALIGNED(base_offset, base_align)) {
- dev_warn(p->dev, "%s offset[%d] 0x%llx, 0x%llx, %d not aligned\n", __func__, i,
- base_offset, base_align, array_mode);
- return -EINVAL;
- }
-
- ntiles = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
- nviews = G_028004_SLICE_MAX(track->db_depth_view) + 1;
- tmp = ntiles * bpe * 64 * nviews;
- if ((tmp + track->db_offset) > radeon_bo_size(track->db_bo)) {
- dev_warn(p->dev, "z/stencil buffer (%d) too small (0x%08X %d %d %d -> %u have %lu)\n",
- array_mode,
- track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
- radeon_bo_size(track->db_bo));
- return -EINVAL;
- }
- }
- }
- track->db_dirty = false;
+ /* Check depth buffer */
+ if (track->db_dirty && (G_028800_STENCIL_ENABLE(track->db_depth_control) ||
+ G_028800_Z_ENABLE(track->db_depth_control))) {
+ r = r600_cs_track_validate_db(p);
+ if (r)
+ return r;
}
+
return 0;
}
@@ -1244,6 +1384,21 @@ static int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
track->db_dirty = true;
break;
case DB_HTILE_DATA_BASE:
+ r = r600_cs_packet_next_reloc(p, &reloc);
+ if (r) {
+ dev_warn(p->dev, "bad SET_CONTEXT_REG "
+ "0x%04X\n", reg);
+ return -EINVAL;
+ }
+ track->htile_offset = radeon_get_ib_value(p, idx) << 8;
+ ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
+ track->htile_bo = reloc->robj;
+ track->db_dirty = true;
+ break;
+ case DB_HTILE_SURFACE:
+ track->htile_surface = radeon_get_ib_value(p, idx);
+ track->db_dirty = true;
+ break;
case SQ_PGM_START_FS:
case SQ_PGM_START_ES:
case SQ_PGM_START_VS:
diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h
index 3568a2e345f..59f9c993cc3 100644
--- a/drivers/gpu/drm/radeon/r600d.h
+++ b/drivers/gpu/drm/radeon/r600d.h
@@ -195,6 +195,14 @@
#define PREZ_MUST_WAIT_FOR_POSTZ_DONE (1 << 31)
#define DB_DEPTH_BASE 0x2800C
#define DB_HTILE_DATA_BASE 0x28014
+#define DB_HTILE_SURFACE 0x28D24
+#define S_028D24_HTILE_WIDTH(x) (((x) & 0x1) << 0)
+#define G_028D24_HTILE_WIDTH(x) (((x) >> 0) & 0x1)
+#define C_028D24_HTILE_WIDTH 0xFFFFFFFE
+#define S_028D24_HTILE_HEIGHT(x) (((x) & 0x1) << 1)
+#define G_028D24_HTILE_HEIGHT(x) (((x) >> 1) & 0x1)
+#define C_028D24_HTILE_HEIGHT 0xFFFFFFFD
+#define G_028D24_LINEAR(x) (((x) >> 2) & 0x1)
#define DB_WATERMARKS 0x9838
#define DEPTH_FREE(x) ((x) << 0)
#define DEPTH_FLUSH(x) ((x) << 5)
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 91541e63d58..df6a4dbd93f 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -233,7 +233,18 @@ int radeon_bo_pin_restricted(struct radeon_bo *bo, u32 domain, u64 max_offset,
bo->pin_count++;
if (gpu_addr)
*gpu_addr = radeon_bo_gpu_offset(bo);
- WARN_ON_ONCE(max_offset != 0);
+
+ if (max_offset != 0) {
+ u64 domain_start;
+
+ if (domain == RADEON_GEM_DOMAIN_VRAM)
+ domain_start = bo->rdev->mc.vram_start;
+ else
+ domain_start = bo->rdev->mc.gtt_start;
+ WARN_ON_ONCE(max_offset <
+ (radeon_bo_gpu_offset(bo) - domain_start));
+ }
+
return 0;
}
radeon_ttm_placement_from_domain(bo, domain);
diff --git a/drivers/gpu/drm/radeon/reg_srcs/cayman b/drivers/gpu/drm/radeon/reg_srcs/cayman
index aea63c41585..0f656b111c1 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/cayman
+++ b/drivers/gpu/drm/radeon/reg_srcs/cayman
@@ -509,7 +509,6 @@ cayman 0x9400
0x00028AA8 IA_MULTI_VGT_PARAM
0x00028AB4 VGT_REUSE_OFF
0x00028AB8 VGT_VTX_CNT_EN
-0x00028ABC DB_HTILE_SURFACE
0x00028AC0 DB_SRESULTS_COMPARE_STATE0
0x00028AC4 DB_SRESULTS_COMPARE_STATE1
0x00028AC8 DB_PRELOAD_CONTROL
diff --git a/drivers/gpu/drm/radeon/reg_srcs/evergreen b/drivers/gpu/drm/radeon/reg_srcs/evergreen
index 77c37202376..b912a37689b 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/evergreen
+++ b/drivers/gpu/drm/radeon/reg_srcs/evergreen
@@ -519,7 +519,6 @@ evergreen 0x9400
0x00028AA4 VGT_INSTANCE_STEP_RATE_1
0x00028AB4 VGT_REUSE_OFF
0x00028AB8 VGT_VTX_CNT_EN
-0x00028ABC DB_HTILE_SURFACE
0x00028AC0 DB_SRESULTS_COMPARE_STATE0
0x00028AC4 DB_SRESULTS_COMPARE_STATE1
0x00028AC8 DB_PRELOAD_CONTROL
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r600 b/drivers/gpu/drm/radeon/reg_srcs/r600
index 626c24ea0b5..5e659b034d9 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/r600
+++ b/drivers/gpu/drm/radeon/reg_srcs/r600
@@ -713,7 +713,6 @@ r600 0x9400
0x0000A710 TD_VS_SAMPLER17_BORDER_RED
0x00009508 TA_CNTL_AUX
0x0002802C DB_DEPTH_CLEAR
-0x00028D24 DB_HTILE_SURFACE
0x00028D34 DB_PREFETCH_LIMIT
0x00028D30 DB_PRELOAD_CONTROL
0x00028D0C DB_RENDER_CONTROL
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2024-11-13 16:06:30 +0000