commit 4adf332cc24ee2d46064aaafd8216169d29566d5 Author: Alex Deucher Date: Sun Nov 14 20:24:35 2010 -0500 drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie commit c37cb9e61dce7437f63280d9347a9ffdf4ec34e7 Author: Alex Deucher Date: Wed Oct 27 01:44:35 2010 -0400 drm/radeon/kms: fix tiled db height calculation on 6xx/7xx Calculate height based on the slice bitfield rather than the size. Same as Dave's CB fix. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie commit b80c8fdc2fadd8182b958e91a10f2fa287f993e4 Author: Alex Deucher Date: Tue Oct 26 20:22:42 2010 -0400 drm/radeon/kms: fix r6xx/7xx 1D tiling CS checker v2 broken by: drm/radeon/r600: fix tiling issues in CS checker. v2: only apply it to 1D tiling case. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie commit 4336ac5c0a4e5dfbb51631ad680d6a5d0b295cd3 Author: Alex Deucher Date: Mon Oct 18 23:45:39 2010 -0400 drm/radeon/kms: fix 2D tile height alignment in the r600 CS checker macro tile heights are aligned to num channels, not num banks. Noticed by Dave Airlie. Signed-off-by: Alex Deucher Cc: stable@kernel.org Signed-off-by: Dave Airlie commit 5c76976a1419a633f9f33c6547bae00348b855d2 Author: Dave Airlie Date: Thu Oct 21 13:55:40 2010 +1000 drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher Signed-off-by: Dave Airlie commit fa479e9df4558af6f091c45be37f713e64b836a1 Author: Alex Deucher Date: Tue Sep 14 10:10:47 2010 -0400 drm/radeon/kms: only warn on mipmap size checks in r600 cs checker (v2) The texture base address registers are in units of 256 bytes. The original CS checker treated these offsets as bytes, so the original check was wrong. I fixed the units in a patch during the 2.6.36 cycle, but this ended up breaking some existing userspace (probably due to a bug in either userspace texture allocation or the drm texture mipmap checker). So for now, until we come up with a better fix, just warn if the mipmap size it too large. This will keep existing userspace working and it should be just as safe as before when we were checking the wrong units. These are GPU MC addresses, so if they fall outside of the VRAM or GART apertures, they end up at the GPU default page, so this should be safe from a security perspective. v2: Just disable the warning. It just spams the log and there's nothing the user can do about it. Signed-off-by: Alex Deucher Cc: Jerome Glisse Signed-off-by: Dave Airlie commit 6e8df81d43d5c95fe37db7f0ef55332de1a4b698 Author: Dave Airlie Date: Thu Aug 12 09:40:05 2010 +1000 drm/radeon: drop old and broken mesa warning This never really got fixed in mesa, and the kernel deals with the problem just fine, so don't got reporting things that confuse people. Signed-off-by: Dave Airlie commit 23f012fb9a0633f2f8901440e314d6276255b1c0 Author: Alex Deucher Date: Wed Aug 11 11:54:25 2010 -0400 drm/radeon/kms: another r6xx/r7xx CS checker fix add default case for buffer formats Signed-off-by: Alex Deucher Cc: Andre Maasikas Signed-off-by: Dave Airlie commit 2c7e76decda2d437f0ca064fef1a2d5d8892288e Author: Alex Deucher Date: Fri Aug 6 02:54:05 2010 -0400 drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher Cc: Andre Maasikas Signed-off-by: Dave Airlie commit 85d1363c9f15b5d4303b635142cee0ba9d1473fc Author: Alex Deucher Date: Fri Jun 4 18:41:42 2010 -0400 drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie commit b956b6e7c7fb207daf32520c0a72c8c06ef1d5f5 Author: Alex Deucher Date: Thu May 20 12:43:52 2010 -0400 drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie drivers/gpu/drm/radeon/r600_cs.c | 391 ++++++++++++++++++++++++++++++-------- drivers/gpu/drm/radeon/r600d.h | 12 ++ 2 files changed, 324 insertions(+), 79 deletions(-) diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c index 144c32d..0f90fc3 100644 --- a/drivers/gpu/drm/radeon/r600_cs.c +++ b/drivers/gpu/drm/radeon/r600_cs.c @@ -25,6 +25,7 @@ * Alex Deucher * Jerome Glisse */ +#include #include "drmP.h" #include "radeon.h" #include "r600d.h" @@ -49,6 +50,7 @@ struct r600_cs_track { u32 nsamples; u32 cb_color_base_last[8]; struct radeon_bo *cb_color_bo[8]; + u64 cb_color_bo_mc[8]; u32 cb_color_bo_offset[8]; struct radeon_bo *cb_color_frag_bo[8]; struct radeon_bo *cb_color_tile_bo[8]; @@ -66,6 +68,7 @@ struct r600_cs_track { u32 db_depth_size; u32 db_offset; struct radeon_bo *db_bo; + u64 db_bo_mc; }; static inline int r600_bpe_from_format(u32 *bpe, u32 format) @@ -132,12 +135,75 @@ static inline int r600_bpe_from_format(u32 *bpe, u32 format) case V_038004_FMT_GB_GR: case V_038004_FMT_BG_RG: case V_038004_COLOR_INVALID: + default: *bpe = 16; return -EINVAL; } return 0; } +struct array_mode_checker { + int array_mode; + u32 group_size; + u32 nbanks; + u32 npipes; + u32 nsamples; + u32 bpe; +}; + +/* returns alignment in pixels for pitch/height/depth and bytes for base */ +static inline int r600_get_array_mode_alignment(struct array_mode_checker *values, + u32 *pitch_align, + u32 *height_align, + u32 *depth_align, + u64 *base_align) +{ + u32 tile_width = 8; + u32 tile_height = 8; + u32 macro_tile_width = values->nbanks; + u32 macro_tile_height = values->npipes; + u32 tile_bytes = tile_width * tile_height * values->bpe * values->nsamples; + u32 macro_tile_bytes = macro_tile_width * macro_tile_height * tile_bytes; + + switch (values->array_mode) { + case ARRAY_LINEAR_GENERAL: + /* technically tile_width/_height for pitch/height */ + *pitch_align = 1; /* tile_width */ + *height_align = 1; /* tile_height */ + *depth_align = 1; + *base_align = 1; + break; + case ARRAY_LINEAR_ALIGNED: + *pitch_align = max((u32)64, (u32)(values->group_size / values->bpe)); + *height_align = tile_height; + *depth_align = 1; + *base_align = values->group_size; + break; + case ARRAY_1D_TILED_THIN1: + *pitch_align = max((u32)tile_width, + (u32)(values->group_size / + (tile_height * values->bpe * values->nsamples))); + *height_align = tile_height; + *depth_align = 1; + *base_align = values->group_size; + break; + case ARRAY_2D_TILED_THIN1: + *pitch_align = max((u32)macro_tile_width, + (u32)(((values->group_size / tile_height) / + (values->bpe * values->nsamples)) * + values->nbanks)) * tile_width; + *height_align = macro_tile_height * tile_height; + *depth_align = 1; + *base_align = max(macro_tile_bytes, + (*pitch_align) * values->bpe * (*height_align) * values->nsamples); + break; + default: + return -EINVAL; + } + + return 0; +} + static void r600_cs_track_init(struct r600_cs_track *track) { int i; @@ -151,10 +217,12 @@ static void r600_cs_track_init(struct r600_cs_track *track) track->cb_color_info[i] = 0; track->cb_color_bo[i] = NULL; track->cb_color_bo_offset[i] = 0xFFFFFFFF; + track->cb_color_bo_mc[i] = 0xFFFFFFFF; } track->cb_target_mask = 0xFFFFFFFF; track->cb_shader_mask = 0xFFFFFFFF; track->db_bo = NULL; + track->db_bo_mc = 0xFFFFFFFF; /* assume the biggest format and that htile is enabled */ track->db_depth_info = 7 | (1 << 25); track->db_depth_view = 0xFFFFC000; @@ -166,70 +234,58 @@ static void r600_cs_track_init(struct r600_cs_track *track) static inline int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i) { struct r600_cs_track *track = p->track; - u32 bpe = 0, pitch, slice_tile_max, size, tmp, height; + u32 bpe = 0, slice_tile_max, size, tmp; + u32 height, height_align, pitch, pitch_align, depth_align; + u64 base_offset, base_align; + struct array_mode_checker array_check; volatile u32 *ib = p->ib->ptr; + unsigned array_mode; if (G_0280A0_TILE_MODE(track->cb_color_info[i])) { dev_warn(p->dev, "FMASK or CMASK buffer are not supported by this kernel\n"); return -EINVAL; } - size = radeon_bo_size(track->cb_color_bo[i]); + size = radeon_bo_size(track->cb_color_bo[i]) - track->cb_color_bo_offset[i]; if (r600_bpe_from_format(&bpe, G_0280A0_FORMAT(track->cb_color_info[i]))) { dev_warn(p->dev, "%s:%d cb invalid format %d for %d (0x%08X)\n", __func__, __LINE__, G_0280A0_FORMAT(track->cb_color_info[i]), i, track->cb_color_info[i]); return -EINVAL; } - pitch = (G_028060_PITCH_TILE_MAX(track->cb_color_size[i]) + 1) << 3; + /* pitch in pixels */ + pitch = (G_028060_PITCH_TILE_MAX(track->cb_color_size[i]) + 1) * 8; slice_tile_max = G_028060_SLICE_TILE_MAX(track->cb_color_size[i]) + 1; - if (!pitch) { - dev_warn(p->dev, "%s:%d cb pitch (%d) for %d invalid (0x%08X)\n", - __func__, __LINE__, pitch, i, track->cb_color_size[i]); - return -EINVAL; - } - height = size / (pitch * bpe); + slice_tile_max *= 64; + height = slice_tile_max / pitch; if (height > 8192) height = 8192; - switch (G_0280A0_ARRAY_MODE(track->cb_color_info[i])) { + array_mode = G_0280A0_ARRAY_MODE(track->cb_color_info[i]); + + base_offset = track->cb_color_bo_mc[i] + track->cb_color_bo_offset[i]; + 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.bpe = 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 for %d (0x%08X)\n", __func__, + G_0280A0_ARRAY_MODE(track->cb_color_info[i]), i, + track->cb_color_info[i]); + return -EINVAL; + } + switch (array_mode) { case V_0280A0_ARRAY_LINEAR_GENERAL: + break; case V_0280A0_ARRAY_LINEAR_ALIGNED: - if (pitch & 0x3f) { - dev_warn(p->dev, "%s:%d cb pitch (%d x %d = %d) invalid\n", - __func__, __LINE__, pitch, bpe, pitch * bpe); - return -EINVAL; - } - if ((pitch * bpe) & (track->group_size - 1)) { - dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n", - __func__, __LINE__, pitch); - return -EINVAL; - } break; case V_0280A0_ARRAY_1D_TILED_THIN1: - if ((pitch * 8 * bpe * track->nsamples) & (track->group_size - 1)) { - dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n", - __func__, __LINE__, pitch); - return -EINVAL; - } - height &= ~0x7; - if (!height) - height = 8; + /* avoid breaking userspace */ + if (height > 7) + height &= ~0x7; break; case V_0280A0_ARRAY_2D_TILED_THIN1: - if (pitch & ((8 * track->nbanks) - 1)) { - dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n", - __func__, __LINE__, pitch); - return -EINVAL; - } - tmp = pitch * 8 * bpe * track->nsamples; - tmp = tmp / track->nbanks; - if (tmp & (track->group_size - 1)) { - dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n", - __func__, __LINE__, pitch); - return -EINVAL; - } - height &= ~((16 * track->npipes) - 1); - if (!height) - height = 16 * track->npipes; break; default: dev_warn(p->dev, "%s invalid tiling %d for %d (0x%08X)\n", __func__, @@ -237,17 +293,43 @@ static inline int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i) track->cb_color_info[i]); return -EINVAL; } + + if (!IS_ALIGNED(pitch, pitch_align)) { + dev_warn(p->dev, "%s:%d cb pitch (%d) invalid\n", + __func__, __LINE__, pitch); + return -EINVAL; + } + if (!IS_ALIGNED(height, height_align)) { + dev_warn(p->dev, "%s:%d cb height (%d) invalid\n", + __func__, __LINE__, height); + return -EINVAL; + } + if (!IS_ALIGNED(base_offset, base_align)) { + dev_warn(p->dev, "%s offset[%d] 0x%llx not aligned\n", __func__, i, base_offset); + return -EINVAL; + } + /* check offset */ - tmp = height * pitch; + tmp = height * pitch * bpe; if ((tmp + track->cb_color_bo_offset[i]) > radeon_bo_size(track->cb_color_bo[i])) { - dev_warn(p->dev, "%s offset[%d] %d to big\n", __func__, i, track->cb_color_bo_offset[i]); - return -EINVAL; + if (array_mode == V_0280A0_ARRAY_LINEAR_GENERAL) { + /* the initial DDX does bad things with the CB size occasionally */ + /* it rounds up height too far for slice tile max but the BO is smaller */ + tmp = (height - 7) * 8 * bpe; + if ((tmp + track->cb_color_bo_offset[i]) > radeon_bo_size(track->cb_color_bo[i])) { + dev_warn(p->dev, "%s offset[%d] %d %d %lu too big\n", __func__, i, track->cb_color_bo_offset[i], tmp, radeon_bo_size(track->cb_color_bo[i])); + return -EINVAL; + } + } else { + dev_warn(p->dev, "%s offset[%d] %d %d %lu too big\n", __func__, i, track->cb_color_bo_offset[i], tmp, radeon_bo_size(track->cb_color_bo[i])); + return -EINVAL; + } } /* limit max tile */ tmp = (height * pitch) >> 6; if (tmp < slice_tile_max) slice_tile_max = tmp; - tmp = S_028060_PITCH_TILE_MAX((pitch >> 3) - 1) | + tmp = S_028060_PITCH_TILE_MAX((pitch / 8) - 1) | S_028060_SLICE_TILE_MAX(slice_tile_max - 1); ib[track->cb_color_size_idx[i]] = tmp; return 0; @@ -289,7 +371,12 @@ static int r600_cs_track_check(struct radeon_cs_parser *p) /* Check depth buffer */ if (G_028800_STENCIL_ENABLE(track->db_depth_control) || G_028800_Z_ENABLE(track->db_depth_control)) { - u32 nviews, bpe, ntiles; + 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; @@ -321,7 +408,6 @@ static int r600_cs_track_check(struct radeon_cs_parser *p) dev_warn(p->dev, "z/stencil buffer size not set\n"); return -EINVAL; } - printk_once(KERN_WARNING "You have old & broken userspace please consider updating mesa\n"); tmp = radeon_bo_size(track->db_bo) - track->db_offset; tmp = (tmp / bpe) >> 6; if (!tmp) { @@ -332,11 +418,63 @@ static int r600_cs_track_check(struct radeon_cs_parser *p) } 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.bpe = 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) invalid\n", + __func__, __LINE__, pitch); + return -EINVAL; + } + if (!IS_ALIGNED(height, height_align)) { + dev_warn(p->dev, "%s:%d db height (%d) invalid\n", + __func__, __LINE__, height); + return -EINVAL; + } + if (!IS_ALIGNED(base_offset, base_align)) { + dev_warn(p->dev, "%s offset[%d] 0x%llx not aligned\n", __func__, i, base_offset); + 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 too small (0x%08X %d %d %d -> %d have %ld)\n", + dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %d -> %u have %lu)\n", track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset, radeon_bo_size(track->db_bo)); return -EINVAL; @@ -724,7 +862,25 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx track->db_depth_control = radeon_get_ib_value(p, idx); break; case R_028010_DB_DEPTH_INFO: - track->db_depth_info = radeon_get_ib_value(p, idx); + if (r600_cs_packet_next_is_pkt3_nop(p)) { + 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->db_depth_info = radeon_get_ib_value(p, idx); + ib[idx] &= C_028010_ARRAY_MODE; + track->db_depth_info &= C_028010_ARRAY_MODE; + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) { + ib[idx] |= S_028010_ARRAY_MODE(V_028010_ARRAY_2D_TILED_THIN1); + track->db_depth_info |= S_028010_ARRAY_MODE(V_028010_ARRAY_2D_TILED_THIN1); + } else { + ib[idx] |= S_028010_ARRAY_MODE(V_028010_ARRAY_1D_TILED_THIN1); + track->db_depth_info |= S_028010_ARRAY_MODE(V_028010_ARRAY_1D_TILED_THIN1); + } + } else + track->db_depth_info = radeon_get_ib_value(p, idx); break; case R_028004_DB_DEPTH_VIEW: track->db_depth_view = radeon_get_ib_value(p, idx); @@ -757,8 +913,25 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx case R_0280B4_CB_COLOR5_INFO: case R_0280B8_CB_COLOR6_INFO: case R_0280BC_CB_COLOR7_INFO: - tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4; - track->cb_color_info[tmp] = radeon_get_ib_value(p, idx); + if (r600_cs_packet_next_is_pkt3_nop(p)) { + r = r600_cs_packet_next_reloc(p, &reloc); + if (r) { + dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg); + return -EINVAL; + } + tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4; + track->cb_color_info[tmp] = radeon_get_ib_value(p, idx); + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) { + ib[idx] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_2D_TILED_THIN1); + track->cb_color_info[tmp] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_2D_TILED_THIN1); + } else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) { + ib[idx] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_1D_TILED_THIN1); + track->cb_color_info[tmp] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_1D_TILED_THIN1); + } + } else { + tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4; + track->cb_color_info[tmp] = radeon_get_ib_value(p, idx); + } break; case R_028060_CB_COLOR0_SIZE: case R_028064_CB_COLOR1_SIZE: @@ -796,8 +969,6 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx return -EINVAL; } ib[idx] = track->cb_color_base_last[tmp]; - printk_once(KERN_WARNING "You have old & broken userspace " - "please consider updating mesa & xf86-video-ati\n"); track->cb_color_frag_bo[tmp] = track->cb_color_bo[tmp]; } else { r = r600_cs_packet_next_reloc(p, &reloc); @@ -824,8 +995,6 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx return -EINVAL; } ib[idx] = track->cb_color_base_last[tmp]; - printk_once(KERN_WARNING "You have old & broken userspace " - "please consider updating mesa & xf86-video-ati\n"); track->cb_color_tile_bo[tmp] = track->cb_color_bo[tmp]; } else { r = r600_cs_packet_next_reloc(p, &reloc); @@ -852,10 +1021,11 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx return -EINVAL; } tmp = (reg - CB_COLOR0_BASE) / 4; - track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx); + track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx) << 8; ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); track->cb_color_base_last[tmp] = ib[idx]; track->cb_color_bo[tmp] = reloc->robj; + track->cb_color_bo_mc[tmp] = reloc->lobj.gpu_offset; break; case DB_DEPTH_BASE: r = r600_cs_packet_next_reloc(p, &reloc); @@ -864,9 +1034,10 @@ static inline int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx "0x%04X\n", reg); return -EINVAL; } - track->db_offset = radeon_get_ib_value(p, idx); + track->db_offset = radeon_get_ib_value(p, idx) << 8; ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); track->db_bo = reloc->robj; + track->db_bo_mc = reloc->lobj.gpu_offset; break; case DB_HTILE_DATA_BASE: case SQ_PGM_START_FS: @@ -946,8 +1117,9 @@ static inline unsigned minify(unsigned size, unsigned levels) } static void r600_texture_size(unsigned nfaces, unsigned blevel, unsigned nlevels, - unsigned w0, unsigned h0, unsigned d0, unsigned bpe, - unsigned *l0_size, unsigned *mipmap_size) + unsigned w0, unsigned h0, unsigned d0, unsigned bpe, + unsigned pitch_align, + unsigned *l0_size, unsigned *mipmap_size) { unsigned offset, i, level, face; unsigned width, height, depth, rowstride, size; @@ -960,18 +1132,18 @@ static void r600_texture_size(unsigned nfaces, unsigned blevel, unsigned nlevels height = minify(h0, i); depth = minify(d0, i); for(face = 0; face < nfaces; face++) { - rowstride = ((width * bpe) + 255) & ~255; + rowstride = ALIGN((width * bpe), pitch_align); size = height * rowstride * depth; offset += size; offset = (offset + 0x1f) & ~0x1f; } } - *l0_size = (((w0 * bpe) + 255) & ~255) * h0 * d0; + *l0_size = ALIGN((w0 * bpe), pitch_align) * h0 * d0; *mipmap_size = offset; - if (!blevel) - *mipmap_size -= *l0_size; if (!nlevels) *mipmap_size = *l0_size; + if (!blevel) + *mipmap_size -= *l0_size; } /** @@ -985,16 +1157,32 @@ static void r600_texture_size(unsigned nfaces, unsigned blevel, unsigned nlevels * the texture and mipmap bo object are big enough to cover this resource. */ static inline int r600_check_texture_resource(struct radeon_cs_parser *p, u32 idx, - struct radeon_bo *texture, - struct radeon_bo *mipmap) + struct radeon_bo *texture, + struct radeon_bo *mipmap, + u64 base_offset, + u64 mip_offset, + u32 tiling_flags) { + struct r600_cs_track *track = p->track; u32 nfaces, nlevels, blevel, w0, h0, d0, bpe = 0; u32 word0, word1, l0_size, mipmap_size; + u32 height_align, pitch, pitch_align, depth_align; + u64 base_align; + struct array_mode_checker array_check; /* on legacy kernel we don't perform advanced check */ if (p->rdev == NULL) return 0; + + /* convert to bytes */ + base_offset <<= 8; + mip_offset <<= 8; + word0 = radeon_get_ib_value(p, idx + 0); + if (tiling_flags & RADEON_TILING_MACRO) + word0 |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1); + else if (tiling_flags & RADEON_TILING_MICRO) + word0 |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1); word1 = radeon_get_ib_value(p, idx + 1); w0 = G_038000_TEX_WIDTH(word0) + 1; h0 = G_038004_TEX_HEIGHT(word1) + 1; @@ -1021,24 +1209,59 @@ static inline int r600_check_texture_resource(struct radeon_cs_parser *p, u32 i __func__, __LINE__, G_038004_DATA_FORMAT(word1)); return -EINVAL; } + + /* pitch in texels */ + pitch = (G_038000_PITCH(word0) + 1) * 8; + array_check.array_mode = G_038000_TILE_MODE(word0); + array_check.group_size = track->group_size; + array_check.nbanks = track->nbanks; + array_check.npipes = track->npipes; + array_check.nsamples = 1; + array_check.bpe = bpe; + if (r600_get_array_mode_alignment(&array_check, + &pitch_align, &height_align, &depth_align, &base_align)) { + dev_warn(p->dev, "%s:%d tex array mode (%d) invalid\n", + __func__, __LINE__, G_038000_TILE_MODE(word0)); + return -EINVAL; + } + + /* XXX check height as well... */ + + if (!IS_ALIGNED(pitch, pitch_align)) { + dev_warn(p->dev, "%s:%d tex pitch (%d) invalid\n", + __func__, __LINE__, pitch); + return -EINVAL; + } + if (!IS_ALIGNED(base_offset, base_align)) { + dev_warn(p->dev, "%s:%d tex base offset (0x%llx) invalid\n", + __func__, __LINE__, base_offset); + return -EINVAL; + } + if (!IS_ALIGNED(mip_offset, base_align)) { + dev_warn(p->dev, "%s:%d tex mip offset (0x%llx) invalid\n", + __func__, __LINE__, mip_offset); + return -EINVAL; + } + word0 = radeon_get_ib_value(p, idx + 4); word1 = radeon_get_ib_value(p, idx + 5); blevel = G_038010_BASE_LEVEL(word0); nlevels = G_038014_LAST_LEVEL(word1); - r600_texture_size(nfaces, blevel, nlevels, w0, h0, d0, bpe, &l0_size, &mipmap_size); + r600_texture_size(nfaces, blevel, nlevels, w0, h0, d0, bpe, + (pitch_align * bpe), + &l0_size, &mipmap_size); /* using get ib will give us the offset into the texture bo */ - word0 = radeon_get_ib_value(p, idx + 2); + word0 = radeon_get_ib_value(p, idx + 2) << 8; if ((l0_size + word0) > radeon_bo_size(texture)) { dev_warn(p->dev, "texture bo too small (%d %d %d %d -> %d have %ld)\n", w0, h0, bpe, word0, l0_size, radeon_bo_size(texture)); return -EINVAL; } /* using get ib will give us the offset into the mipmap bo */ - word0 = radeon_get_ib_value(p, idx + 3); + word0 = radeon_get_ib_value(p, idx + 3) << 8; if ((mipmap_size + word0) > radeon_bo_size(mipmap)) { - dev_warn(p->dev, "mipmap bo too small (%d %d %d %d %d %d -> %d have %ld)\n", - w0, h0, bpe, blevel, nlevels, word0, mipmap_size, radeon_bo_size(texture)); - return -EINVAL; + /*dev_warn(p->dev, "mipmap bo too small (%d %d %d %d %d %d -> %d have %ld)\n", + w0, h0, bpe, blevel, nlevels, word0, mipmap_size, radeon_bo_size(texture));*/ } return 0; } @@ -1228,7 +1451,7 @@ static int r600_packet3_check(struct radeon_cs_parser *p, } for (i = 0; i < (pkt->count / 7); i++) { struct radeon_bo *texture, *mipmap; - u32 size, offset; + u32 size, offset, base_offset, mip_offset; switch (G__SQ_VTX_CONSTANT_TYPE(radeon_get_ib_value(p, idx+(i*7)+6+1))) { case SQ_TEX_VTX_VALID_TEXTURE: @@ -1238,7 +1461,11 @@ static int r600_packet3_check(struct radeon_cs_parser *p, DRM_ERROR("bad SET_RESOURCE\n"); return -EINVAL; } - ib[idx+1+(i*7)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); + base_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1); + else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) + ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1); texture = reloc->robj; /* tex mip base */ r = r600_cs_packet_next_reloc(p, &reloc); @@ -1246,12 +1473,17 @@ static int r600_packet3_check(struct radeon_cs_parser *p, DRM_ERROR("bad SET_RESOURCE\n"); return -EINVAL; } - ib[idx+1+(i*7)+3] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); + mip_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff); mipmap = reloc->robj; r = r600_check_texture_resource(p, idx+(i*7)+1, - texture, mipmap); + texture, mipmap, + base_offset + radeon_get_ib_value(p, idx+1+(i*7)+2), + mip_offset + radeon_get_ib_value(p, idx+1+(i*7)+3), + reloc->lobj.tiling_flags); if (r) return r; + ib[idx+1+(i*7)+2] += base_offset; + ib[idx+1+(i*7)+3] += mip_offset; break; case SQ_TEX_VTX_VALID_BUFFER: /* vtx base */ @@ -1261,10 +1493,11 @@ static int r600_packet3_check(struct radeon_cs_parser *p, return -EINVAL; } offset = radeon_get_ib_value(p, idx+1+(i*7)+0); - size = radeon_get_ib_value(p, idx+1+(i*7)+1); + size = radeon_get_ib_value(p, idx+1+(i*7)+1) + 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"); + dev_warn(p->dev, "vbo resource seems too big (%d) for the bo (%ld)\n", + size + offset, radeon_bo_size(reloc->robj)); ib[idx+1+(i*7)+1] = radeon_bo_size(reloc->robj); } ib[idx+1+(i*7)+0] += (u32)((reloc->lobj.gpu_offset) & 0xffffffff); diff --git a/drivers/gpu/drm/radeon/r600d.h b/drivers/gpu/drm/radeon/r600d.h index 84bc28e..9577945 100644 --- a/drivers/gpu/drm/radeon/r600d.h +++ b/drivers/gpu/drm/radeon/r600d.h @@ -51,6 +51,12 @@ #define PTE_READABLE (1 << 5) #define PTE_WRITEABLE (1 << 6) +/* tiling bits */ +#define ARRAY_LINEAR_GENERAL 0x00000000 +#define ARRAY_LINEAR_ALIGNED 0x00000001 +#define ARRAY_1D_TILED_THIN1 0x00000002 +#define ARRAY_2D_TILED_THIN1 0x00000004 + /* Registers */ #define ARB_POP 0x2418 #define ENABLE_TC128 (1 << 30) @@ -1155,6 +1161,10 @@ #define S_038000_TILE_MODE(x) (((x) & 0xF) << 3) #define G_038000_TILE_MODE(x) (((x) >> 3) & 0xF) #define C_038000_TILE_MODE 0xFFFFFF87 +#define V_038000_ARRAY_LINEAR_GENERAL 0x00000000 +#define V_038000_ARRAY_LINEAR_ALIGNED 0x00000001 +#define V_038000_ARRAY_1D_TILED_THIN1 0x00000002 +#define V_038000_ARRAY_2D_TILED_THIN1 0x00000004 #define S_038000_TILE_TYPE(x) (((x) & 0x1) << 7) #define G_038000_TILE_TYPE(x) (((x) >> 7) & 0x1) #define C_038000_TILE_TYPE 0xFFFFFF7F @@ -1358,6 +1368,8 @@ #define S_028010_ARRAY_MODE(x) (((x) & 0xF) << 15) #define G_028010_ARRAY_MODE(x) (((x) >> 15) & 0xF) #define C_028010_ARRAY_MODE 0xFFF87FFF +#define V_028010_ARRAY_1D_TILED_THIN1 0x00000002 +#define V_028010_ARRAY_2D_TILED_THIN1 0x00000004 #define S_028010_TILE_SURFACE_ENABLE(x) (((x) & 0x1) << 25) #define G_028010_TILE_SURFACE_ENABLE(x) (((x) >> 25) & 0x1) #define C_028010_TILE_SURFACE_ENABLE 0xFDFFFFFF