diff options
46 files changed, 3576 insertions, 1529 deletions
diff --git a/arch/x86/mm/highmem_32.c b/arch/x86/mm/highmem_32.c index 58f621e..2112ed5 100644 --- a/arch/x86/mm/highmem_32.c +++ b/arch/x86/mm/highmem_32.c @@ -103,6 +103,7 @@ EXPORT_SYMBOL(kmap); EXPORT_SYMBOL(kunmap); EXPORT_SYMBOL(kmap_atomic); EXPORT_SYMBOL(kunmap_atomic); +EXPORT_SYMBOL(kmap_atomic_prot); void __init set_highmem_pages_init(void) { diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index 5fae1e0..013d380 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -13,7 +13,8 @@ radeon-$(CONFIG_DRM_RADEON_KMS) += radeon_device.o radeon_kms.o \ radeon_encoders.o radeon_display.o radeon_cursor.o radeon_i2c.o \ radeon_clocks.o radeon_fb.o radeon_gem.o radeon_ring.o radeon_irq_kms.o \ radeon_cs.o radeon_bios.o radeon_benchmark.o r100.o r300.o r420.o \ - rs400.o rs600.o rs690.o rv515.o r520.o r600.o rs780.o rv770.o + rs400.o rs600.o rs690.o rv515.o r520.o r600.o rs780.o rv770.o \ + radeon_test.o radeon-$(CONFIG_COMPAT) += radeon_ioc32.o diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c index c0080cc..74d034f 100644 --- a/drivers/gpu/drm/radeon/atombios_crtc.c +++ b/drivers/gpu/drm/radeon/atombios_crtc.c @@ -31,6 +31,132 @@ #include "atom.h" #include "atom-bits.h" +static void atombios_overscan_setup(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + struct drm_device *dev = crtc->dev; + struct radeon_device *rdev = dev->dev_private; + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + SET_CRTC_OVERSCAN_PS_ALLOCATION args; + int index = GetIndexIntoMasterTable(COMMAND, SetCRTC_OverScan); + int a1, a2; + + memset(&args, 0, sizeof(args)); + + args.usOverscanRight = 0; + args.usOverscanLeft = 0; + args.usOverscanBottom = 0; + args.usOverscanTop = 0; + args.ucCRTC = radeon_crtc->crtc_id; + + switch (radeon_crtc->rmx_type) { + case RMX_CENTER: + args.usOverscanTop = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2; + args.usOverscanBottom = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2; + args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2; + args.usOverscanRight = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2; + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + break; + case RMX_ASPECT: + a1 = mode->crtc_vdisplay * adjusted_mode->crtc_hdisplay; + a2 = adjusted_mode->crtc_vdisplay * mode->crtc_hdisplay; + + if (a1 > a2) { + args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2; + args.usOverscanRight = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2; + } else if (a2 > a1) { + args.usOverscanLeft = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2; + args.usOverscanRight = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2; + } + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + break; + case RMX_FULL: + default: + args.usOverscanRight = 0; + args.usOverscanLeft = 0; + args.usOverscanBottom = 0; + args.usOverscanTop = 0; + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + break; + } +} + +static void atombios_scaler_setup(struct drm_crtc *crtc) +{ + struct drm_device *dev = crtc->dev; + struct radeon_device *rdev = dev->dev_private; + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + ENABLE_SCALER_PS_ALLOCATION args; + int index = GetIndexIntoMasterTable(COMMAND, EnableScaler); + /* fixme - fill in enc_priv for atom dac */ + enum radeon_tv_std tv_std = TV_STD_NTSC; + + if (!ASIC_IS_AVIVO(rdev) && radeon_crtc->crtc_id) + return; + + memset(&args, 0, sizeof(args)); + + args.ucScaler = radeon_crtc->crtc_id; + + if (radeon_crtc->devices & (ATOM_DEVICE_TV_SUPPORT)) { + switch (tv_std) { + case TV_STD_NTSC: + default: + args.ucTVStandard = ATOM_TV_NTSC; + break; + case TV_STD_PAL: + args.ucTVStandard = ATOM_TV_PAL; + break; + case TV_STD_PAL_M: + args.ucTVStandard = ATOM_TV_PALM; + break; + case TV_STD_PAL_60: + args.ucTVStandard = ATOM_TV_PAL60; + break; + case TV_STD_NTSC_J: + args.ucTVStandard = ATOM_TV_NTSCJ; + break; + case TV_STD_SCART_PAL: + args.ucTVStandard = ATOM_TV_PAL; /* ??? */ + break; + case TV_STD_SECAM: + args.ucTVStandard = ATOM_TV_SECAM; + break; + case TV_STD_PAL_CN: + args.ucTVStandard = ATOM_TV_PALCN; + break; + } + args.ucEnable = SCALER_ENABLE_MULTITAP_MODE; + } else if (radeon_crtc->devices & (ATOM_DEVICE_CV_SUPPORT)) { + args.ucTVStandard = ATOM_TV_CV; + args.ucEnable = SCALER_ENABLE_MULTITAP_MODE; + } else { + switch (radeon_crtc->rmx_type) { + case RMX_FULL: + args.ucEnable = ATOM_SCALER_EXPANSION; + break; + case RMX_CENTER: + args.ucEnable = ATOM_SCALER_CENTER; + break; + case RMX_ASPECT: + args.ucEnable = ATOM_SCALER_EXPANSION; + break; + default: + if (ASIC_IS_AVIVO(rdev)) + args.ucEnable = ATOM_SCALER_DISABLE; + else + args.ucEnable = ATOM_SCALER_CENTER; + break; + } + } + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + if (radeon_crtc->devices & (ATOM_DEVICE_CV_SUPPORT | ATOM_DEVICE_TV_SUPPORT) + && rdev->family >= CHIP_RV515 && rdev->family <= CHIP_RV570) { + atom_rv515_force_tv_scaler(rdev); + } +} + static void atombios_lock_crtc(struct drm_crtc *crtc, int lock) { struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); @@ -203,6 +329,12 @@ void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode) if (ASIC_IS_AVIVO(rdev)) { uint32_t ss_cntl; + if ((rdev->family == CHIP_RS600) || + (rdev->family == CHIP_RS690) || + (rdev->family == CHIP_RS740)) + pll_flags |= (RADEON_PLL_USE_FRAC_FB_DIV | + RADEON_PLL_PREFER_CLOSEST_LOWER); + if (ASIC_IS_DCE32(rdev) && mode->clock > 200000) /* range limits??? */ pll_flags |= RADEON_PLL_PREFER_HIGH_FB_DIV; else @@ -321,7 +453,7 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y, struct drm_gem_object *obj; struct drm_radeon_gem_object *obj_priv; uint64_t fb_location; - uint32_t fb_format, fb_pitch_pixels; + uint32_t fb_format, fb_pitch_pixels, tiling_flags; if (!crtc->fb) return -EINVAL; @@ -358,7 +490,14 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y, return -EINVAL; } - /* TODO tiling */ + radeon_object_get_tiling_flags(obj->driver_private, + &tiling_flags, NULL); + if (tiling_flags & RADEON_TILING_MACRO) + fb_format |= AVIVO_D1GRPH_MACRO_ADDRESS_MODE; + + if (tiling_flags & RADEON_TILING_MICRO) + fb_format |= AVIVO_D1GRPH_TILED; + if (radeon_crtc->crtc_id == 0) WREG32(AVIVO_D1VGA_CONTROL, 0); else @@ -509,6 +648,9 @@ int atombios_crtc_mode_set(struct drm_crtc *crtc, radeon_crtc_set_base(crtc, x, y, old_fb); radeon_legacy_atom_set_surface(crtc); } + atombios_overscan_setup(crtc, mode, adjusted_mode); + atombios_scaler_setup(crtc); + radeon_bandwidth_update(rdev); return 0; } @@ -516,6 +658,8 @@ static bool atombios_crtc_mode_fixup(struct drm_crtc *crtc, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) { + if (!radeon_crtc_scaling_mode_fixup(crtc, mode, adjusted_mode)) + return false; return true; } @@ -548,148 +692,3 @@ void radeon_atombios_init_crtc(struct drm_device *dev, AVIVO_D2CRTC_H_TOTAL - AVIVO_D1CRTC_H_TOTAL; drm_crtc_helper_add(&radeon_crtc->base, &atombios_helper_funcs); } - -void radeon_init_disp_bw_avivo(struct drm_device *dev, - struct drm_display_mode *mode1, - uint32_t pixel_bytes1, - struct drm_display_mode *mode2, - uint32_t pixel_bytes2) -{ - struct radeon_device *rdev = dev->dev_private; - fixed20_12 min_mem_eff; - fixed20_12 peak_disp_bw, mem_bw, pix_clk, pix_clk2, temp_ff; - fixed20_12 sclk_ff, mclk_ff; - uint32_t dc_lb_memory_split, temp; - - min_mem_eff.full = rfixed_const_8(0); - if (rdev->disp_priority == 2) { - uint32_t mc_init_misc_lat_timer = 0; - if (rdev->family == CHIP_RV515) - mc_init_misc_lat_timer = - RREG32_MC(RV515_MC_INIT_MISC_LAT_TIMER); - else if (rdev->family == CHIP_RS690) - mc_init_misc_lat_timer = - RREG32_MC(RS690_MC_INIT_MISC_LAT_TIMER); - - mc_init_misc_lat_timer &= - ~(R300_MC_DISP1R_INIT_LAT_MASK << - R300_MC_DISP1R_INIT_LAT_SHIFT); - mc_init_misc_lat_timer &= - ~(R300_MC_DISP0R_INIT_LAT_MASK << - R300_MC_DISP0R_INIT_LAT_SHIFT); - - if (mode2) - mc_init_misc_lat_timer |= - (1 << R300_MC_DISP1R_INIT_LAT_SHIFT); - if (mode1) - mc_init_misc_lat_timer |= - (1 << R300_MC_DISP0R_INIT_LAT_SHIFT); - - if (rdev->family == CHIP_RV515) - WREG32_MC(RV515_MC_INIT_MISC_LAT_TIMER, - mc_init_misc_lat_timer); - else if (rdev->family == CHIP_RS690) - WREG32_MC(RS690_MC_INIT_MISC_LAT_TIMER, - mc_init_misc_lat_timer); - } - - /* - * determine is there is enough bw for current mode - */ - temp_ff.full = rfixed_const(100); - mclk_ff.full = rfixed_const(rdev->clock.default_mclk); - mclk_ff.full = rfixed_div(mclk_ff, temp_ff); - sclk_ff.full = rfixed_const(rdev->clock.default_sclk); - sclk_ff.full = rfixed_div(sclk_ff, temp_ff); - - temp = (rdev->mc.vram_width / 8) * (rdev->mc.vram_is_ddr ? 2 : 1); - temp_ff.full = rfixed_const(temp); - mem_bw.full = rfixed_mul(mclk_ff, temp_ff); - mem_bw.full = rfixed_mul(mem_bw, min_mem_eff); - - pix_clk.full = 0; - pix_clk2.full = 0; - peak_disp_bw.full = 0; - if (mode1) { - temp_ff.full = rfixed_const(1000); - pix_clk.full = rfixed_const(mode1->clock); /* convert to fixed point */ - pix_clk.full = rfixed_div(pix_clk, temp_ff); - temp_ff.full = rfixed_const(pixel_bytes1); - peak_disp_bw.full += rfixed_mul(pix_clk, temp_ff); - } - if (mode2) { - temp_ff.full = rfixed_const(1000); - pix_clk2.full = rfixed_const(mode2->clock); /* convert to fixed point */ - pix_clk2.full = rfixed_div(pix_clk2, temp_ff); - temp_ff.full = rfixed_const(pixel_bytes2); - peak_disp_bw.full += rfixed_mul(pix_clk2, temp_ff); - } - - if (peak_disp_bw.full >= mem_bw.full) { - DRM_ERROR - ("You may not have enough display bandwidth for current mode\n" - "If you have flickering problem, try to lower resolution, refresh rate, or color depth\n"); - printk("peak disp bw %d, mem_bw %d\n", - rfixed_trunc(peak_disp_bw), rfixed_trunc(mem_bw)); - } - - /* - * Line Buffer Setup - * There is a single line buffer shared by both display controllers. - * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between the display - * controllers. The paritioning can either be done manually or via one of four - * preset allocations specified in bits 1:0: - * 0 - line buffer is divided in half and shared between each display controller - * 1 - D1 gets 3/4 of the line buffer, D2 gets 1/4 - * 2 - D1 gets the whole buffer - * 3 - D1 gets 1/4 of the line buffer, D2 gets 3/4 - * Setting bit 2 of DC_LB_MEMORY_SPLIT controls switches to manual allocation mode. - * In manual allocation mode, D1 always starts at 0, D1 end/2 is specified in bits - * 14:4; D2 allocation follows D1. - */ - - /* is auto or manual better ? */ - dc_lb_memory_split = - RREG32(AVIVO_DC_LB_MEMORY_SPLIT) & ~AVIVO_DC_LB_MEMORY_SPLIT_MASK; - dc_lb_memory_split &= ~AVIVO_DC_LB_MEMORY_SPLIT_SHIFT_MODE; -#if 1 - /* auto */ - if (mode1 && mode2) { - if (mode1->hdisplay > mode2->hdisplay) { - if (mode1->hdisplay > 2560) - dc_lb_memory_split |= - AVIVO_DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q; - else - dc_lb_memory_split |= - AVIVO_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; - } else if (mode2->hdisplay > mode1->hdisplay) { - if (mode2->hdisplay > 2560) - dc_lb_memory_split |= - AVIVO_DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q; - else - dc_lb_memory_split |= - AVIVO_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; - } else - dc_lb_memory_split |= - AVIVO_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; - } else if (mode1) { - dc_lb_memory_split |= AVIVO_DC_LB_MEMORY_SPLIT_D1_ONLY; - } else if (mode2) { - dc_lb_memory_split |= AVIVO_DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q; - } -#else - /* manual */ - dc_lb_memory_split |= AVIVO_DC_LB_MEMORY_SPLIT_SHIFT_MODE; - dc_lb_memory_split &= - ~(AVIVO_DC_LB_DISP1_END_ADR_MASK << - AVIVO_DC_LB_DISP1_END_ADR_SHIFT); - if (mode1) { - dc_lb_memory_split |= - ((((mode1->hdisplay / 2) + 64) & AVIVO_DC_LB_DISP1_END_ADR_MASK) - << AVIVO_DC_LB_DISP1_END_ADR_SHIFT); - } else if (mode2) { - dc_lb_memory_split |= (0 << AVIVO_DC_LB_DISP1_END_ADR_SHIFT); - } -#endif - WREG32(AVIVO_DC_LB_MEMORY_SPLIT, dc_lb_memory_split); -} diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index c550932..05a4489 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -110,7 +110,7 @@ int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; } - rdev->gart.table.ram.ptr[i] = cpu_to_le32((uint32_t)addr); + rdev->gart.table.ram.ptr[i] = cpu_to_le32(lower_32_bits(addr)); return 0; } @@ -173,8 +173,12 @@ void r100_mc_setup(struct radeon_device *rdev) DRM_ERROR("Failed to register debugfs file for R100 MC !\n"); } /* Write VRAM size in case we are limiting it */ - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + /* Novell bug 204882 for RN50/M6/M7 with 8/16/32MB VRAM, + * if the aperture is 64MB but we have 32MB VRAM + * we report only 32MB VRAM but we have to set MC_FB_LOCATION + * to 64MB, otherwise the gpu accidentially dies */ + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(RADEON_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(RADEON_MC_FB_START, rdev->mc.vram_location >> 16); WREG32(RADEON_MC_FB_LOCATION, tmp); @@ -215,7 +219,6 @@ int r100_mc_init(struct radeon_device *rdev) r100_pci_gart_disable(rdev); /* Setup GPU memory space */ - rdev->mc.vram_location = 0xFFFFFFFFUL; rdev->mc.gtt_location = 0xFFFFFFFFUL; if (rdev->flags & RADEON_IS_AGP) { r = radeon_agp_init(rdev); @@ -753,6 +756,102 @@ int r100_cs_packet_parse(struct radeon_cs_parser *p, } /** + * r100_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 + * PACKET0 - WAIT_UNTIL +_value + * RELOC (P3) - crtc_id in reloc. + * + * This function parses this and relocates the VLINE START END + * and WAIT UNTIL packets to the correct crtc. + * It also detects a switched off crtc and nulls out the + * wait in that case. + */ +int r100_cs_packet_parse_vline(struct radeon_cs_parser *p) +{ + struct radeon_cs_chunk *ib_chunk; + struct drm_mode_object *obj; + struct drm_crtc *crtc; + struct radeon_crtc *radeon_crtc; + struct radeon_cs_packet p3reloc, waitreloc; + int crtc_id; + int r; + uint32_t header, h_idx, reg; + + ib_chunk = &p->chunks[p->chunk_ib_idx]; + + /* parse the wait until */ + r = r100_cs_packet_parse(p, &waitreloc, p->idx); + if (r) + return r; + + /* check its a wait until and only 1 count */ + if (waitreloc.reg != RADEON_WAIT_UNTIL || + waitreloc.count != 0) { + DRM_ERROR("vline wait had illegal wait until segment\n"); + r = -EINVAL; + return r; + } + + if (ib_chunk->kdata[waitreloc.idx + 1] != RADEON_WAIT_CRTC_VLINE) { + DRM_ERROR("vline wait had illegal wait until\n"); + r = -EINVAL; + return r; + } + + /* jump over the NOP */ + r = r100_cs_packet_parse(p, &p3reloc, p->idx); + if (r) + return r; + + h_idx = p->idx - 2; + p->idx += waitreloc.count; + p->idx += p3reloc.count; + + header = ib_chunk->kdata[h_idx]; + crtc_id = ib_chunk->kdata[h_idx + 5]; + reg = ib_chunk->kdata[h_idx] >> 2; + mutex_lock(&p->rdev->ddev->mode_config.mutex); + 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); + r = -EINVAL; + goto out; + } + 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 until */ + ib_chunk->kdata[h_idx + 2] = PACKET2(0); + ib_chunk->kdata[h_idx + 3] = PACKET2(0); + } else if (crtc_id == 1) { + switch (reg) { + case AVIVO_D1MODE_VLINE_START_END: + header &= R300_CP_PACKET0_REG_MASK; + header |= AVIVO_D2MODE_VLINE_START_END >> 2; + break; + case RADEON_CRTC_GUI_TRIG_VLINE: + header &= R300_CP_PACKET0_REG_MASK; + header |= RADEON_CRTC2_GUI_TRIG_VLINE >> 2; + break; + default: + DRM_ERROR("unknown crtc reloc\n"); + r = -EINVAL; + goto out; + } + ib_chunk->kdata[h_idx] = header; + ib_chunk->kdata[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1; + } +out: + mutex_unlock(&p->rdev->ddev->mode_config.mutex); + return r; +} + +/** * r100_cs_packet_next_reloc() - parse next packet which should be reloc packet3 * @parser: parser structure holding parsing context. * @data: pointer to relocation data @@ -814,6 +913,7 @@ static int r100_packet0_check(struct radeon_cs_parser *p, unsigned idx; bool onereg; int r; + u32 tile_flags = 0; ib = p->ib->ptr; ib_chunk = &p->chunks[p->chunk_ib_idx]; @@ -825,6 +925,15 @@ static int r100_packet0_check(struct radeon_cs_parser *p, } for (i = 0; i <= pkt->count; i++, idx++, reg += 4) { switch (reg) { + case RADEON_CRTC_GUI_TRIG_VLINE: + r = r100_cs_packet_parse_vline(p); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + break; /* FIXME: only allow PACKET3 blit? easier to check for out of * range access */ case RADEON_DST_PITCH_OFFSET: @@ -838,7 +947,20 @@ static int r100_packet0_check(struct radeon_cs_parser *p, } tmp = ib_chunk->kdata[idx] & 0x003fffff; tmp += (((u32)reloc->lobj.gpu_offset) >> 10); - ib[idx] = (ib_chunk->kdata[idx] & 0xffc00000) | tmp; + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= RADEON_DST_TILE_MACRO; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) { + if (reg == RADEON_SRC_PITCH_OFFSET) { + DRM_ERROR("Cannot src blit from microtiled surface\n"); + r100_cs_dump_packet(p, pkt); + return -EINVAL; + } + tile_flags |= RADEON_DST_TILE_MICRO; + } + + tmp |= tile_flags; + ib[idx] = (ib_chunk->kdata[idx] & 0x3fc00000) | tmp; break; case RADEON_RB3D_DEPTHOFFSET: case RADEON_RB3D_COLOROFFSET: @@ -869,6 +991,11 @@ static int r100_packet0_check(struct radeon_cs_parser *p, case R300_TX_OFFSET_0+52: case R300_TX_OFFSET_0+56: case R300_TX_OFFSET_0+60: + /* rn50 has no 3D engine so fail on any 3d setup */ + if (ASIC_IS_RN50(p->rdev)) { + DRM_ERROR("attempt to use RN50 3D engine failed\n"); + return -EINVAL; + } r = r100_cs_packet_next_reloc(p, &reloc); if (r) { DRM_ERROR("No reloc for ib[%d]=0x%04X\n", @@ -878,6 +1005,25 @@ static int r100_packet0_check(struct radeon_cs_parser *p, } ib[idx] = ib_chunk->kdata[idx] + ((u32)reloc->lobj.gpu_offset); break; + case R300_RB3D_COLORPITCH0: + case RADEON_RB3D_COLORPITCH: + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= RADEON_COLOR_TILE_ENABLE; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) + tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; + + tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); + tmp |= tile_flags; + ib[idx] = tmp; + break; default: /* FIXME: we don't want to allow anyothers packet */ break; @@ -1256,29 +1402,100 @@ static void r100_vram_get_type(struct radeon_device *rdev) } } -void r100_vram_info(struct radeon_device *rdev) +static u32 r100_get_accessible_vram(struct radeon_device *rdev) { - r100_vram_get_type(rdev); + u32 aper_size; + u8 byte; + + aper_size = RREG32(RADEON_CONFIG_APER_SIZE); + + /* Set HDP_APER_CNTL only on cards that are known not to be broken, + * that is has the 2nd generation multifunction PCI interface + */ + if (rdev->family == CHIP_RV280 || + rdev->family >= CHIP_RV350) { + WREG32_P(RADEON_HOST_PATH_CNTL, RADEON_HDP_APER_CNTL, + ~RADEON_HDP_APER_CNTL); + DRM_INFO("Generation 2 PCI interface, using max accessible memory\n"); + return aper_size * 2; + } + + /* Older cards have all sorts of funny issues to deal with. First + * check if it's a multifunction card by reading the PCI config + * header type... Limit those to one aperture size + */ + pci_read_config_byte(rdev->pdev, 0xe, &byte); + if (byte & 0x80) { + DRM_INFO("Generation 1 PCI interface in multifunction mode\n"); + DRM_INFO("Limiting VRAM to one aperture\n"); + return aper_size; + } + + /* Single function older card. We read HDP_APER_CNTL to see how the BIOS + * have set it up. We don't write this as it's broken on some ASICs but + * we expect the BIOS to have done the right thing (might be too optimistic...) + */ + if (RREG32(RADEON_HOST_PATH_CNTL) & RADEON_HDP_APER_CNTL) + return aper_size * 2; + return aper_size; +} + +void r100_vram_init_sizes(struct radeon_device *rdev) +{ + u64 config_aper_size; + u32 accessible; + + config_aper_size = RREG32(RADEON_CONFIG_APER_SIZE); if (rdev->flags & RADEON_IS_IGP) { uint32_t tom; /* read NB_TOM to get the amount of ram stolen for the GPU */ tom = RREG32(RADEON_NB_TOM); - rdev->mc.vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); + rdev->mc.real_vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); + /* for IGPs we need to keep VRAM where it was put by the BIOS */ + rdev->mc.vram_location = (tom & 0xffff) << 16; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; } else { - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); + rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE); /* Some production boards of m6 will report 0 * if it's 8 MB */ - if (rdev->mc.vram_size == 0) { - rdev->mc.vram_size = 8192 * 1024; - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); + if (rdev->mc.real_vram_size == 0) { + rdev->mc.real_vram_size = 8192 * 1024; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); } + /* let driver place VRAM */ + rdev->mc.vram_location = 0xFFFFFFFFUL; + /* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM - + * Novell bug 204882 + along with lots of ubuntu ones */ + if (config_aper_size > rdev->mc.real_vram_size) + rdev->mc.mc_vram_size = config_aper_size; + else + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; } + /* work out accessible VRAM */ + accessible = r100_get_accessible_vram(rdev); + rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + + if (accessible > rdev->mc.aper_size) + accessible = rdev->mc.aper_size; + + if (rdev->mc.mc_vram_size > rdev->mc.aper_size) + rdev->mc.mc_vram_size = rdev->mc.aper_size; + + if (rdev->mc.real_vram_size > rdev->mc.aper_size) + rdev->mc.real_vram_size = rdev->mc.aper_size; +} + +void r100_vram_info(struct radeon_device *rdev) +{ + r100_vram_get_type(rdev); + + r100_vram_init_sizes(rdev); } @@ -1533,3 +1750,530 @@ int r100_debugfs_mc_info_init(struct radeon_device *rdev) return 0; #endif } + +int r100_set_surface_reg(struct radeon_device *rdev, int reg, + uint32_t tiling_flags, uint32_t pitch, + uint32_t offset, uint32_t obj_size) +{ + int surf_index = reg * 16; + int flags = 0; + + /* r100/r200 divide by 16 */ + if (rdev->family < CHIP_R300) + flags = pitch / 16; + else + flags = pitch / 8; + + if (rdev->family <= CHIP_RS200) { + if ((tiling_flags & (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) + == (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) + flags |= RADEON_SURF_TILE_COLOR_BOTH; + if (tiling_flags & RADEON_TILING_MACRO) + flags |= RADEON_SURF_TILE_COLOR_MACRO; + } else if (rdev->family <= CHIP_RV280) { + if (tiling_flags & (RADEON_TILING_MACRO)) + flags |= R200_SURF_TILE_COLOR_MACRO; + if (tiling_flags & RADEON_TILING_MICRO) + flags |= R200_SURF_TILE_COLOR_MICRO; + } else { + if (tiling_flags & RADEON_TILING_MACRO) + flags |= R300_SURF_TILE_MACRO; + if (tiling_flags & RADEON_TILING_MICRO) + flags |= R300_SURF_TILE_MICRO; + } + + DRM_DEBUG("writing surface %d %d %x %x\n", reg, flags, offset, offset+obj_size-1); + WREG32(RADEON_SURFACE0_INFO + surf_index, flags); + WREG32(RADEON_SURFACE0_LOWER_BOUND + surf_index, offset); + WREG32(RADEON_SURFACE0_UPPER_BOUND + surf_index, offset + obj_size - 1); + return 0; +} + +void r100_clear_surface_reg(struct radeon_device *rdev, int reg) +{ + int surf_index = reg * 16; + WREG32(RADEON_SURFACE0_INFO + surf_index, 0); +} + +void r100_bandwidth_update(struct radeon_device *rdev) +{ + fixed20_12 trcd_ff, trp_ff, tras_ff, trbs_ff, tcas_ff; + fixed20_12 sclk_ff, mclk_ff, sclk_eff_ff, sclk_delay_ff; + fixed20_12 peak_disp_bw, mem_bw, pix_clk, pix_clk2, temp_ff, crit_point_ff; + uint32_t temp, data, mem_trcd, mem_trp, mem_tras; + fixed20_12 memtcas_ff[8] = { + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(0), + fixed_init_half(1), + fixed_init_half(2), + fixed_init(0), + }; + fixed20_12 memtcas_rs480_ff[8] = { + fixed_init(0), + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(0), + fixed_init_half(1), + fixed_init_half(2), + fixed_init_half(3), + }; + fixed20_12 memtcas2_ff[8] = { + fixed_init(0), + fixed_init(1), + fixed_init(2), + fixed_init(3), + fixed_init(4), + fixed_init(5), + fixed_init(6), + fixed_init(7), + }; + fixed20_12 memtrbs[8] = { + fixed_init(1), + fixed_init_half(1), + fixed_init(2), + fixed_init_half(2), + fixed_init(3), + fixed_init_half(3), + fixed_init(4), + fixed_init_half(4) + }; + fixed20_12 memtrbs_r4xx[8] = { + fixed_init(4), + fixed_init(5), + fixed_init(6), + fixed_init(7), + fixed_init(8), + fixed_init(9), + fixed_init(10), + fixed_init(11) + }; + fixed20_12 min_mem_eff; + fixed20_12 mc_latency_sclk, mc_latency_mclk, k1; + fixed20_12 cur_latency_mclk, cur_latency_sclk; + fixed20_12 disp_latency, disp_latency_overhead, disp_drain_rate, + disp_drain_rate2, read_return_rate; + fixed20_12 time_disp1_drop_priority; + int c; + int cur_size = 16; /* in octawords */ + int critical_point = 0, critical_point2; +/* uint32_t read_return_rate, time_disp1_drop_priority; */ + int stop_req, max_stop_req; + struct drm_display_mode *mode1 = NULL; + struct drm_display_mode *mode2 = NULL; + uint32_t pixel_bytes1 = 0; + uint32_t pixel_bytes2 = 0; + + if (rdev->mode_info.crtcs[0]->base.enabled) { + mode1 = &rdev->mode_info.crtcs[0]->base.mode; + pixel_bytes1 = rdev->mode_info.crtcs[0]->base.fb->bits_per_pixel / 8; + } + if (rdev->mode_info.crtcs[1]->base.enabled) { + mode2 = &rdev->mode_info.crtcs[1]->base.mode; + pixel_bytes2 = rdev->mode_info.crtcs[1]->base.fb->bits_per_pixel / 8; + } + + min_mem_eff.full = rfixed_const_8(0); + /* get modes */ + if ((rdev->disp_priority == 2) && ASIC_IS_R300(rdev)) { + uint32_t mc_init_misc_lat_timer = RREG32(R300_MC_INIT_MISC_LAT_TIMER); + mc_init_misc_lat_timer &= ~(R300_MC_DISP1R_INIT_LAT_MASK << R300_MC_DISP1R_INIT_LAT_SHIFT); + mc_init_misc_lat_timer &= ~(R300_MC_DISP0R_INIT_LAT_MASK << R300_MC_DISP0R_INIT_LAT_SHIFT); + /* check crtc enables */ + if (mode2) + mc_init_misc_lat_timer |= (1 << R300_MC_DISP1R_INIT_LAT_SHIFT); + if (mode1) + mc_init_misc_lat_timer |= (1 << R300_MC_DISP0R_INIT_LAT_SHIFT); + WREG32(R300_MC_INIT_MISC_LAT_TIMER, mc_init_misc_lat_timer); + } + + /* + * determine is there is enough bw for current mode + */ + mclk_ff.full = rfixed_const(rdev->clock.default_mclk); + temp_ff.full = rfixed_const(100); + mclk_ff.full = rfixed_div(mclk_ff, temp_ff); + sclk_ff.full = rfixed_const(rdev->clock.default_sclk); + sclk_ff.full = rfixed_div(sclk_ff, temp_ff); + + temp = (rdev->mc.vram_width / 8) * (rdev->mc.vram_is_ddr ? 2 : 1); + temp_ff.full = rfixed_const(temp); + mem_bw.full = rfixed_mul(mclk_ff, temp_ff); + + pix_clk.full = 0; + pix_clk2.full = 0; + peak_disp_bw.full = 0; + if (mode1) { + temp_ff.full = rfixed_const(1000); + pix_clk.full = rfixed_const(mode1->clock); /* convert to fixed point */ + pix_clk.full = rfixed_div(pix_clk, temp_ff); + temp_ff.full = rfixed_const(pixel_bytes1); + peak_disp_bw.full += rfixed_mul(pix_clk, temp_ff); + } + if (mode2) { + temp_ff.full = rfixed_const(1000); + pix_clk2.full = rfixed_const(mode2->clock); /* convert to fixed point */ + pix_clk2.full = rfixed_div(pix_clk2, temp_ff); + temp_ff.full = rfixed_const(pixel_bytes2); + peak_disp_bw.full += rfixed_mul(pix_clk2, temp_ff); + } + + mem_bw.full = rfixed_mul(mem_bw, min_mem_eff); + if (peak_disp_bw.full >= mem_bw.full) { + DRM_ERROR("You may not have enough display bandwidth for current mode\n" + "If you have flickering problem, try to lower resolution, refresh rate, or color depth\n"); + } + + /* Get values from the EXT_MEM_CNTL register...converting its contents. */ + temp = RREG32(RADEON_MEM_TIMING_CNTL); + if ((rdev->family == CHIP_RV100) || (rdev->flags & RADEON_IS_IGP)) { /* RV100, M6, IGPs */ + mem_trcd = ((temp >> 2) & 0x3) + 1; + mem_trp = ((temp & 0x3)) + 1; + mem_tras = ((temp & 0x70) >> 4) + 1; + } else if (rdev->family == CHIP_R300 || + rdev->family == CHIP_R350) { /* r300, r350 */ + mem_trcd = (temp & 0x7) + 1; + mem_trp = ((temp >> 8) & 0x7) + 1; + mem_tras = ((temp >> 11) & 0xf) + 4; + } else if (rdev->family == CHIP_RV350 || + rdev->family <= CHIP_RV380) { + /* rv3x0 */ + mem_trcd = (temp & 0x7) + 3; + mem_trp = ((temp >> 8) & 0x7) + 3; + mem_tras = ((temp >> 11) & 0xf) + 6; + } else if (rdev->family == CHIP_R420 || + rdev->family == CHIP_R423 || + rdev->family == CHIP_RV410) { + /* r4xx */ + mem_trcd = (temp & 0xf) + 3; + if (mem_trcd > 15) + mem_trcd = 15; + mem_trp = ((temp >> 8) & 0xf) + 3; + if (mem_trp > 15) + mem_trp = 15; + mem_tras = ((temp >> 12) & 0x1f) + 6; + if (mem_tras > 31) + mem_tras = 31; + } else { /* RV200, R200 */ + mem_trcd = (temp & 0x7) + 1; + mem_trp = ((temp >> 8) & 0x7) + 1; + mem_tras = ((temp >> 12) & 0xf) + 4; + } + /* convert to FF */ + trcd_ff.full = rfixed_const(mem_trcd); + trp_ff.full = rfixed_const(mem_trp); + tras_ff.full = rfixed_const(mem_tras); + + /* Get values from the MEM_SDRAM_MODE_REG register...converting its */ + temp = RREG32(RADEON_MEM_SDRAM_MODE_REG); + data = (temp & (7 << 20)) >> 20; + if ((rdev->family == CHIP_RV100) || rdev->flags & RADEON_IS_IGP) { + if (rdev->family == CHIP_RS480) /* don't think rs400 */ + tcas_ff = memtcas_rs480_ff[data]; + else + tcas_ff = memtcas_ff[data]; + } else + tcas_ff = memtcas2_ff[data]; + + if (rdev->family == CHIP_RS400 || + rdev->family == CHIP_RS480) { + /* extra cas latency stored in bits 23-25 0-4 clocks */ + data = (temp >> 23) & 0x7; + if (data < 5) + tcas_ff.full += rfixed_const(data); + } + + if (ASIC_IS_R300(rdev) && !(rdev->flags & RADEON_IS_IGP)) { + /* on the R300, Tcas is included in Trbs. + */ + temp = RREG32(RADEON_MEM_CNTL); + data = (R300_MEM_NUM_CHANNELS_MASK & temp); + if (data == 1) { + if (R300_MEM_USE_CD_CH_ONLY & temp) { + temp = RREG32(R300_MC_IND_INDEX); + temp &= ~R300_MC_IND_ADDR_MASK; + temp |= R300_MC_READ_CNTL_CD_mcind; + WREG32(R300_MC_IND_INDEX, temp); + temp = RREG32(R300_MC_IND_DATA); + data = (R300_MEM_RBS_POSITION_C_MASK & temp); + } else { + temp = RREG32(R300_MC_READ_CNTL_AB); + data = (R300_MEM_RBS_POSITION_A_MASK & temp); + } + } else { + temp = RREG32(R300_MC_READ_CNTL_AB); + data = (R300_MEM_RBS_POSITION_A_MASK & temp); + } + if (rdev->family == CHIP_RV410 || + rdev->family == CHIP_R420 || + rdev->family == CHIP_R423) + trbs_ff = memtrbs_r4xx[data]; + else + trbs_ff = memtrbs[data]; + tcas_ff.full += trbs_ff.full; + } + + sclk_eff_ff.full = sclk_ff.full; + + if (rdev->flags & RADEON_IS_AGP) { + fixed20_12 agpmode_ff; + agpmode_ff.full = rfixed_const(radeon_agpmode); + temp_ff.full = rfixed_const_666(16); + sclk_eff_ff.full -= rfixed_mul(agpmode_ff, temp_ff); + } + /* TODO PCIE lanes may affect this - agpmode == 16?? */ + + if (ASIC_IS_R300(rdev)) { + sclk_delay_ff.full = rfixed_const(250); + } else { + if ((rdev->family == CHIP_RV100) || + rdev->flags & RADEON_IS_IGP) { + if (rdev->mc.vram_is_ddr) + sclk_delay_ff.full = rfixed_const(41); + else + sclk_delay_ff.full = rfixed_const(33); + } else { + if (rdev->mc.vram_width == 128) + sclk_delay_ff.full = rfixed_const(57); + else + sclk_delay_ff.full = rfixed_const(41); + } + } + + mc_latency_sclk.full = rfixed_div(sclk_delay_ff, sclk_eff_ff); + + if (rdev->mc.vram_is_ddr) { + if (rdev->mc.vram_width == 32) { + k1.full = rfixed_const(40); + c = 3; + } else { + k1.full = rfixed_const(20); + c = 1; + } + } else { + k1.full = rfixed_const(40); + c = 3; + } + + temp_ff.full = rfixed_const(2); + mc_latency_mclk.full = rfixed_mul(trcd_ff, temp_ff); + temp_ff.full = rfixed_const(c); + mc_latency_mclk.full += rfixed_mul(tcas_ff, temp_ff); + temp_ff.full = rfixed_const(4); + mc_latency_mclk.full += rfixed_mul(tras_ff, temp_ff); + mc_latency_mclk.full += rfixed_mul(trp_ff, temp_ff); + mc_latency_mclk.full += k1.full; + + mc_latency_mclk.full = rfixed_div(mc_latency_mclk, mclk_ff); + mc_latency_mclk.full += rfixed_div(temp_ff, sclk_eff_ff); + + /* + HW cursor time assuming worst case of full size colour cursor. + */ + temp_ff.full = rfixed_const((2 * (cur_size - (rdev->mc.vram_is_ddr + 1)))); + temp_ff.full += trcd_ff.full; + if (temp_ff.full < tras_ff.full) + temp_ff.full = tras_ff.full; + cur_latency_mclk.full = rfixed_div(temp_ff, mclk_ff); + + temp_ff.full = rfixed_const(cur_size); + cur_latency_sclk.full = rfixed_div(temp_ff, sclk_eff_ff); + /* + Find the total latency for the display data. + */ + disp_latency_overhead.full = rfixed_const(80); + disp_latency_overhead.full = rfixed_div(disp_latency_overhead, sclk_ff); + mc_latency_mclk.full += disp_latency_overhead.full + cur_latency_mclk.full; + mc_latency_sclk.full += disp_latency_overhead.full + cur_latency_sclk.full; + + if (mc_latency_mclk.full > mc_latency_sclk.full) + disp_latency.full = mc_latency_mclk.full; + else + disp_latency.full = mc_latency_sclk.full; + + /* setup Max GRPH_STOP_REQ default value */ + if (ASIC_IS_RV100(rdev)) + max_stop_req = 0x5c; + else + max_stop_req = 0x7c; + + if (mode1) { + /* CRTC1 + Set GRPH_BUFFER_CNTL register using h/w defined optimal values. + GRPH_STOP_REQ <= MIN[ 0x7C, (CRTC_H_DISP + 1) * (bit depth) / 0x10 ] + */ + stop_req = mode1->hdisplay * pixel_bytes1 / 16; + + if (stop_req > max_stop_req) + stop_req = max_stop_req; + + /* + Find the drain rate of the display buffer. + */ + temp_ff.full = rfixed_const((16/pixel_bytes1)); + disp_drain_rate.full = rfixed_div(pix_clk, temp_ff); + + /* + Find the critical point of the display buffer. + */ + crit_point_ff.full = rfixed_mul(disp_drain_rate, disp_latency); + crit_point_ff.full += rfixed_const_half(0); + + critical_point = rfixed_trunc(crit_point_ff); + + if (rdev->disp_priority == 2) { + critical_point = 0; + } + + /* + The critical point should never be above max_stop_req-4. Setting + GRPH_CRITICAL_CNTL = 0 will thus force high priority all the time. + */ + if (max_stop_req - critical_point < 4) + critical_point = 0; + + if (critical_point == 0 && mode2 && rdev->family == CHIP_R300) { + /* some R300 cards have problem with this set to 0, when CRTC2 is enabled.*/ + critical_point = 0x10; + } + + temp = RREG32(RADEON_GRPH_BUFFER_CNTL); + temp &= ~(RADEON_GRPH_STOP_REQ_MASK); + temp |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); + temp &= ~(RADEON_GRPH_START_REQ_MASK); + if ((rdev->family == CHIP_R350) && + (stop_req > 0x15)) { + stop_req -= 0x10; + } + temp |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); + temp |= RADEON_GRPH_BUFFER_SIZE; + temp &= ~(RADEON_GRPH_CRITICAL_CNTL | + RADEON_GRPH_CRITICAL_AT_SOF | + RADEON_GRPH_STOP_CNTL); + /* + Write the result into the register. + */ + WREG32(RADEON_GRPH_BUFFER_CNTL, ((temp & ~RADEON_GRPH_CRITICAL_POINT_MASK) | + (critical_point << RADEON_GRPH_CRITICAL_POINT_SHIFT))); + +#if 0 + if ((rdev->family == CHIP_RS400) || + (rdev->family == CHIP_RS480)) { + /* attempt to program RS400 disp regs correctly ??? */ + temp = RREG32(RS400_DISP1_REG_CNTL); + temp &= ~(RS400_DISP1_START_REQ_LEVEL_MASK | + RS400_DISP1_STOP_REQ_LEVEL_MASK); + WREG32(RS400_DISP1_REQ_CNTL1, (temp | + (critical_point << RS400_DISP1_START_REQ_LEVEL_SHIFT) | + (critical_point << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); + temp = RREG32(RS400_DMIF_MEM_CNTL1); + temp &= ~(RS400_DISP1_CRITICAL_POINT_START_MASK | + RS400_DISP1_CRITICAL_POINT_STOP_MASK); + WREG32(RS400_DMIF_MEM_CNTL1, (temp | + (critical_point << RS400_DISP1_CRITICAL_POINT_START_SHIFT) | + (critical_point << RS400_DISP1_CRITICAL_POINT_STOP_SHIFT))); + } +#endif + + DRM_DEBUG("GRPH_BUFFER_CNTL from to %x\n", + /* (unsigned int)info->SavedReg->grph_buffer_cntl, */ + (unsigned int)RREG32(RADEON_GRPH_BUFFER_CNTL)); + } + + if (mode2) { + u32 grph2_cntl; + stop_req = mode2->hdisplay * pixel_bytes2 / 16; + + if (stop_req > max_stop_req) + stop_req = max_stop_req; + + /* + Find the drain rate of the display buffer. + */ + temp_ff.full = rfixed_const((16/pixel_bytes2)); + disp_drain_rate2.full = rfixed_div(pix_clk2, temp_ff); + + grph2_cntl = RREG32(RADEON_GRPH2_BUFFER_CNTL); + grph2_cntl &= ~(RADEON_GRPH_STOP_REQ_MASK); + grph2_cntl |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); + grph2_cntl &= ~(RADEON_GRPH_START_REQ_MASK); + if ((rdev->family == CHIP_R350) && + (stop_req > 0x15)) { + stop_req -= 0x10; + } + grph2_cntl |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); + grph2_cntl |= RADEON_GRPH_BUFFER_SIZE; + grph2_cntl &= ~(RADEON_GRPH_CRITICAL_CNTL | + RADEON_GRPH_CRITICAL_AT_SOF | + RADEON_GRPH_STOP_CNTL); + + if ((rdev->family == CHIP_RS100) || + (rdev->family == CHIP_RS200)) + critical_point2 = 0; + else { + temp = (rdev->mc.vram_width * rdev->mc.vram_is_ddr + 1)/128; + temp_ff.full = rfixed_const(temp); + temp_ff.full = rfixed_mul(mclk_ff, temp_ff); + if (sclk_ff.full < temp_ff.full) + temp_ff.full = sclk_ff.full; + + read_return_rate.full = temp_ff.full; + + if (mode1) { + temp_ff.full = read_return_rate.full - disp_drain_rate.full; + time_disp1_drop_priority.full = rfixed_div(crit_point_ff, temp_ff); + } else { + time_disp1_drop_priority.full = 0; + } + crit_point_ff.full = disp_latency.full + time_disp1_drop_priority.full + disp_latency.full; + crit_point_ff.full = rfixed_mul(crit_point_ff, disp_drain_rate2); + crit_point_ff.full += rfixed_const_half(0); + + critical_point2 = rfixed_trunc(crit_point_ff); + + if (rdev->disp_priority == 2) { + critical_point2 = 0; + } + + if (max_stop_req - critical_point2 < 4) + critical_point2 = 0; + + } + + if (critical_point2 == 0 && rdev->family == CHIP_R300) { + /* some R300 cards have problem with this set to 0 */ + critical_point2 = 0x10; + } + + WREG32(RADEON_GRPH2_BUFFER_CNTL, ((grph2_cntl & ~RADEON_GRPH_CRITICAL_POINT_MASK) | + (critical_point2 << RADEON_GRPH_CRITICAL_POINT_SHIFT))); + + if ((rdev->family == CHIP_RS400) || + (rdev->family == CHIP_RS480)) { +#if 0 + /* attempt to program RS400 disp2 regs correctly ??? */ + temp = RREG32(RS400_DISP2_REQ_CNTL1); + temp &= ~(RS400_DISP2_START_REQ_LEVEL_MASK | + RS400_DISP2_STOP_REQ_LEVEL_MASK); + WREG32(RS400_DISP2_REQ_CNTL1, (temp | + (critical_point2 << RS400_DISP1_START_REQ_LEVEL_SHIFT) | + (critical_point2 << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); + temp = RREG32(RS400_DISP2_REQ_CNTL2); + temp &= ~(RS400_DISP2_CRITICAL_POINT_START_MASK | + RS400_DISP2_CRITICAL_POINT_STOP_MASK); + WREG32(RS400_DISP2_REQ_CNTL2, (temp | + (critical_point2 << RS400_DISP2_CRITICAL_POINT_START_SHIFT) | + (critical_point2 << RS400_DISP2_CRITICAL_POINT_STOP_SHIFT))); +#endif + WREG32(RS400_DISP2_REQ_CNTL1, 0x105DC1CC); + WREG32(RS400_DISP2_REQ_CNTL2, 0x2749D000); + WREG32(RS400_DMIF_MEM_CNTL1, 0x29CA71DC); + WREG32(RS400_DISP1_REQ_CNTL1, 0x28FBC3AC); + } + + DRM_DEBUG("GRPH2_BUFFER_CNTL from to %x\n", + (unsigned int)RREG32(RADEON_GRPH2_BUFFER_CNTL)); + } +} diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index e2ed5bc..9c8d415 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -30,6 +30,8 @@ #include "drm.h" #include "radeon_reg.h" #include "radeon.h" +#include "radeon_drm.h" +#include "radeon_share.h" /* r300,r350,rv350,rv370,rv380 depends on : */ void r100_hdp_reset(struct radeon_device *rdev); @@ -44,6 +46,7 @@ int r100_gui_wait_for_idle(struct radeon_device *rdev); int r100_cs_packet_parse(struct radeon_cs_parser *p, struct radeon_cs_packet *pkt, unsigned idx); +int r100_cs_packet_parse_vline(struct radeon_cs_parser *p); int r100_cs_packet_next_reloc(struct radeon_cs_parser *p, struct radeon_cs_reloc **cs_reloc); int r100_cs_parse_packet0(struct radeon_cs_parser *p, @@ -150,8 +153,13 @@ int rv370_pcie_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; } - addr = (((u32)addr) >> 8) | ((upper_32_bits(addr) & 0xff) << 4) | 0xC; - writel(cpu_to_le32(addr), ((void __iomem *)ptr) + (i * 4)); + addr = (lower_32_bits(addr) >> 8) | + ((upper_32_bits(addr) & 0xff) << 24) | + 0xc; + /* on x86 we want this to be CPU endian, on powerpc + * on powerpc without HW swappers, it'll get swapped on way + * into VRAM - so no need for cpu_to_le32 on VRAM tables */ + writel(addr, ((void __iomem *)ptr) + (i * 4)); return 0; } @@ -579,10 +587,8 @@ void r300_vram_info(struct radeon_device *rdev) } else { rdev->mc.vram_width = 64; } - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); - rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); - rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + r100_vram_init_sizes(rdev); } @@ -970,7 +976,7 @@ static inline void r300_cs_track_clear(struct r300_cs_track *track) static const unsigned r300_reg_safe_bm[159] = { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, - 0xFFFFFFBF, 0xFFFFFFFF, 0xFFFFFFBF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, @@ -1019,7 +1025,7 @@ static int r300_packet0_check(struct radeon_cs_parser *p, struct radeon_cs_reloc *reloc; struct r300_cs_track *track; volatile uint32_t *ib; - uint32_t tmp; + uint32_t tmp, tile_flags = 0; unsigned i; int r; @@ -1027,6 +1033,16 @@ static int r300_packet0_check(struct radeon_cs_parser *p, ib_chunk = &p->chunks[p->chunk_ib_idx]; track = (struct r300_cs_track*)p->track; switch(reg) { + case AVIVO_D1MODE_VLINE_START_END: + case RADEON_CRTC_GUI_TRIG_VLINE: + r = r100_cs_packet_parse_vline(p); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + break; case RADEON_DST_PITCH_OFFSET: case RADEON_SRC_PITCH_OFFSET: r = r100_cs_packet_next_reloc(p, &reloc); @@ -1038,7 +1054,19 @@ static int r300_packet0_check(struct radeon_cs_parser *p, } tmp = ib_chunk->kdata[idx] & 0x003fffff; tmp += (((u32)reloc->lobj.gpu_offset) >> 10); - ib[idx] = (ib_chunk->kdata[idx] & 0xffc00000) | tmp; + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= RADEON_DST_TILE_MACRO; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) { + if (reg == RADEON_SRC_PITCH_OFFSET) { + DRM_ERROR("Cannot src blit from microtiled surface\n"); + r100_cs_dump_packet(p, pkt); + return -EINVAL; + } + tile_flags |= RADEON_DST_TILE_MICRO; + } + tmp |= tile_flags; + ib[idx] = (ib_chunk->kdata[idx] & 0x3fc00000) | tmp; break; case R300_RB3D_COLOROFFSET0: case R300_RB3D_COLOROFFSET1: @@ -1127,6 +1155,23 @@ static int r300_packet0_check(struct radeon_cs_parser *p, /* RB3D_COLORPITCH1 */ /* RB3D_COLORPITCH2 */ /* RB3D_COLORPITCH3 */ + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= R300_COLOR_TILE_ENABLE; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) + tile_flags |= R300_COLOR_MICROTILE_ENABLE; + + tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); + tmp |= tile_flags; + ib[idx] = tmp; + i = (reg - 0x4E38) >> 2; track->cb[i].pitch = ib_chunk->kdata[idx] & 0x3FFE; switch (((ib_chunk->kdata[idx] >> 21) & 0xF)) { @@ -1182,6 +1227,23 @@ static int r300_packet0_check(struct radeon_cs_parser *p, break; case 0x4F24: /* ZB_DEPTHPITCH */ + r = r100_cs_packet_next_reloc(p, &reloc); + if (r) { + DRM_ERROR("No reloc for ib[%d]=0x%04X\n", + idx, reg); + r100_cs_dump_packet(p, pkt); + return r; + } + + if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) + tile_flags |= R300_DEPTHMACROTILE_ENABLE; + if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) + tile_flags |= R300_DEPTHMICROTILE_TILED;; + + tmp = ib_chunk->kdata[idx] & ~(0x7 << 16); + tmp |= tile_flags; + ib[idx] = tmp; + track->zb.pitch = ib_chunk->kdata[idx] & 0x3FFC; break; case 0x4104: diff --git a/drivers/gpu/drm/radeon/r300_reg.h b/drivers/gpu/drm/radeon/r300_reg.h index 70f4860..4b7afef 100644 --- a/drivers/gpu/drm/radeon/r300_reg.h +++ b/drivers/gpu/drm/radeon/r300_reg.h @@ -27,7 +27,9 @@ #ifndef _R300_REG_H_ #define _R300_REG_H_ - +#define R300_SURF_TILE_MACRO (1<<16) +#define R300_SURF_TILE_MICRO (2<<16) +#define R300_SURF_TILE_BOTH (3<<16) #define R300_MC_INIT_MISC_LAT_TIMER 0x180 diff --git a/drivers/gpu/drm/radeon/r500_reg.h b/drivers/gpu/drm/radeon/r500_reg.h index 9070a1c..036691b 100644 --- a/drivers/gpu/drm/radeon/r500_reg.h +++ b/drivers/gpu/drm/radeon/r500_reg.h @@ -445,6 +445,7 @@ #define AVIVO_D1MODE_DATA_FORMAT 0x6528 # define AVIVO_D1MODE_INTERLEAVE_EN (1 << 0) #define AVIVO_D1MODE_DESKTOP_HEIGHT 0x652C +#define AVIVO_D1MODE_VLINE_START_END 0x6538 #define AVIVO_D1MODE_VIEWPORT_START 0x6580 #define AVIVO_D1MODE_VIEWPORT_SIZE 0x6584 #define AVIVO_D1MODE_EXT_OVERSCAN_LEFT_RIGHT 0x6588 @@ -496,6 +497,7 @@ #define AVIVO_D2CUR_SIZE 0x6c10 #define AVIVO_D2CUR_POSITION 0x6c14 +#define AVIVO_D2MODE_VLINE_START_END 0x6d38 #define AVIVO_D2MODE_VIEWPORT_START 0x6d80 #define AVIVO_D2MODE_VIEWPORT_SIZE 0x6d84 #define AVIVO_D2MODE_EXT_OVERSCAN_LEFT_RIGHT 0x6d88 diff --git a/drivers/gpu/drm/radeon/r520.c b/drivers/gpu/drm/radeon/r520.c index 570a244..09fb0b6 100644 --- a/drivers/gpu/drm/radeon/r520.c +++ b/drivers/gpu/drm/radeon/r520.c @@ -28,6 +28,7 @@ #include "drmP.h" #include "radeon_reg.h" #include "radeon.h" +#include "radeon_share.h" /* r520,rv530,rv560,rv570,r580 depends on : */ void r100_hdp_reset(struct radeon_device *rdev); @@ -94,8 +95,8 @@ int r520_mc_init(struct radeon_device *rdev) "programming pipes. Bad things might happen.\n"); } /* Write VRAM size in case we are limiting it */ - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(R520_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(R520_MC_FB_START, rdev->mc.vram_location >> 16); WREG32_MC(R520_MC_FB_LOCATION, tmp); @@ -226,9 +227,20 @@ static void r520_vram_get_type(struct radeon_device *rdev) void r520_vram_info(struct radeon_device *rdev) { + fixed20_12 a; + r520_vram_get_type(rdev); - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); - rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); - rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + r100_vram_init_sizes(rdev); + /* FIXME: we should enforce default clock in case GPU is not in + * default setup + */ + a.full = rfixed_const(100); + rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk); + rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a); +} + +void r520_bandwidth_update(struct radeon_device *rdev) +{ + rv515_bandwidth_avivo_update(rdev); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index c45559f..538cd90 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -67,7 +67,7 @@ int r600_mc_init(struct radeon_device *rdev) "programming pipes. Bad things might happen.\n"); } - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(R600_MC_FB_TOP, tmp >> 24); tmp |= REG_SET(R600_MC_FB_BASE, rdev->mc.vram_location >> 24); WREG32(R600_MC_VM_FB_LOCATION, tmp); @@ -140,7 +140,8 @@ void r600_vram_get_type(struct radeon_device *rdev) void r600_vram_info(struct radeon_device *rdev) { r600_vram_get_type(rdev); - rdev->mc.vram_size = RREG32(R600_CONFIG_MEMSIZE); + rdev->mc.real_vram_size = RREG32(R600_CONFIG_MEMSIZE); + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; /* Could aper size report 0 ? */ rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index d61f2fc..b1d945b 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -64,6 +64,7 @@ extern int radeon_agpmode; extern int radeon_vram_limit; extern int radeon_gart_size; extern int radeon_benchmarking; +extern int radeon_testing; extern int radeon_connector_table; /* @@ -113,6 +114,7 @@ enum radeon_family { CHIP_RV770, CHIP_RV730, CHIP_RV710, + CHIP_RS880, CHIP_LAST, }; @@ -201,6 +203,14 @@ int radeon_fence_wait_last(struct radeon_device *rdev); struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence); void radeon_fence_unref(struct radeon_fence **fence); +/* + * Tiling registers + */ +struct radeon_surface_reg { + struct radeon_object *robj; +}; + +#define RADEON_GEM_MAX_SURFACES 8 /* * Radeon buffer. @@ -213,6 +223,7 @@ struct radeon_object_list { uint64_t gpu_offset; unsigned rdomain; unsigned wdomain; + uint32_t tiling_flags; }; int radeon_object_init(struct radeon_device *rdev); @@ -242,8 +253,15 @@ void radeon_object_list_clean(struct list_head *head); int radeon_object_fbdev_mmap(struct radeon_object *robj, struct vm_area_struct *vma); unsigned long radeon_object_size(struct radeon_object *robj); - - +void radeon_object_clear_surface_reg(struct radeon_object *robj); +int radeon_object_check_tiling(struct radeon_object *robj, bool has_moved, + bool force_drop); +void radeon_object_set_tiling_flags(struct radeon_object *robj, + uint32_t tiling_flags, uint32_t pitch); +void radeon_object_get_tiling_flags(struct radeon_object *robj, uint32_t *tiling_flags, uint32_t *pitch); +void radeon_bo_move_notify(struct ttm_buffer_object *bo, + struct ttm_mem_reg *mem); +void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo); /* * GEM objects. */ @@ -315,8 +333,11 @@ struct radeon_mc { unsigned gtt_location; unsigned gtt_size; unsigned vram_location; - unsigned vram_size; + /* for some chips with <= 32MB we need to lie + * about vram size near mc fb location */ + unsigned mc_vram_size; unsigned vram_width; + unsigned real_vram_size; int vram_mtrr; bool vram_is_ddr; }; @@ -474,6 +495,39 @@ struct radeon_wb { uint64_t gpu_addr; }; +/** + * struct radeon_pm - power management datas + * @max_bandwidth: maximum bandwidth the gpu has (MByte/s) + * @igp_sideport_mclk: sideport memory clock Mhz (rs690,rs740,rs780,rs880) + * @igp_system_mclk: system clock Mhz (rs690,rs740,rs780,rs880) + * @igp_ht_link_clk: ht link clock Mhz (rs690,rs740,rs780,rs880) + * @igp_ht_link_width: ht link width in bits (rs690,rs740,rs780,rs880) + * @k8_bandwidth: k8 bandwidth the gpu has (MByte/s) (IGP) + * @sideport_bandwidth: sideport bandwidth the gpu has (MByte/s) (IGP) + * @ht_bandwidth: ht bandwidth the gpu has (MByte/s) (IGP) + * @core_bandwidth: core GPU bandwidth the gpu has (MByte/s) (IGP) + * @sclk: GPU clock Mhz (core bandwith depends of this clock) + * @needed_bandwidth: current bandwidth needs + * + * It keeps track of various data needed to take powermanagement decision. + * Bandwith need is used to determine minimun clock of the GPU and memory. + * Equation between gpu/memory clock and available bandwidth is hw dependent + * (type of memory, bus size, efficiency, ...) + */ +struct radeon_pm { + fixed20_12 max_bandwidth; + fixed20_12 igp_sideport_mclk; + fixed20_12 igp_system_mclk; + fixed20_12 igp_ht_link_clk; + fixed20_12 igp_ht_link_width; + fixed20_12 k8_bandwidth; + fixed20_12 sideport_bandwidth; + fixed20_12 ht_bandwidth; + fixed20_12 core_bandwidth; + fixed20_12 sclk; + fixed20_12 needed_bandwidth; +}; + /* * Benchmarking @@ -482,6 +536,12 @@ void radeon_benchmark(struct radeon_device *rdev); /* + * Testing + */ +void radeon_test_moves(struct radeon_device *rdev); + + +/* * Debugfs */ int radeon_debugfs_add_files(struct radeon_device *rdev, @@ -535,6 +595,11 @@ struct radeon_asic { void (*set_memory_clock)(struct radeon_device *rdev, uint32_t mem_clock); void (*set_pcie_lanes)(struct radeon_device *rdev, int lanes); void (*set_clock_gating)(struct radeon_device *rdev, int enable); + int (*set_surface_reg)(struct radeon_device *rdev, int reg, + uint32_t tiling_flags, uint32_t pitch, + uint32_t offset, uint32_t obj_size); + int (*clear_surface_reg)(struct radeon_device *rdev, int reg); + void (*bandwidth_update)(struct radeon_device *rdev); }; union radeon_asic_config { @@ -566,6 +631,10 @@ int radeon_gem_busy_ioctl(struct drm_device *dev, void *data, int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); +int radeon_gem_set_tiling_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int radeon_gem_get_tiling_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); /* @@ -594,8 +663,8 @@ struct radeon_device { struct radeon_object *fbdev_robj; struct radeon_framebuffer *fbdev_rfb; /* Register mmio */ - unsigned long rmmio_base; - unsigned long rmmio_size; + resource_size_t rmmio_base; + resource_size_t rmmio_size; void *rmmio; radeon_rreg_t mm_rreg; radeon_wreg_t mm_wreg; @@ -619,11 +688,14 @@ struct radeon_device { struct radeon_irq irq; struct radeon_asic *asic; struct radeon_gem gem; + struct radeon_pm pm; struct mutex cs_mutex; struct radeon_wb wb; bool gpu_lockup; bool shutdown; bool suspend; + bool need_dma32; + struct radeon_surface_reg surface_regs[RADEON_GEM_MAX_SURFACES]; }; int radeon_device_init(struct radeon_device *rdev, @@ -670,6 +742,8 @@ void r100_pll_errata_after_index(struct radeon_device *rdev); /* * ASICs helpers. */ +#define ASIC_IS_RN50(rdev) ((rdev->pdev->device == 0x515e) || \ + (rdev->pdev->device == 0x5969)) #define ASIC_IS_RV100(rdev) ((rdev->family == CHIP_RV100) || \ (rdev->family == CHIP_RV200) || \ (rdev->family == CHIP_RS100) || \ @@ -796,5 +870,8 @@ static inline void radeon_ring_write(struct radeon_device *rdev, uint32_t v) #define radeon_set_memory_clock(rdev, e) (rdev)->asic->set_engine_clock((rdev), (e)) #define radeon_set_pcie_lanes(rdev, l) (rdev)->asic->set_pcie_lanes((rdev), (l)) #define radeon_set_clock_gating(rdev, e) (rdev)->asic->set_clock_gating((rdev), (e)) +#define radeon_set_surface_reg(rdev, r, f, p, o, s) ((rdev)->asic->set_surface_reg((rdev), (r), (f), (p), (o), (s))) +#define radeon_clear_surface_reg(rdev, r) ((rdev)->asic->clear_surface_reg((rdev), (r))) +#define radeon_bandwidth_update(rdev) (rdev)->asic->bandwidth_update((rdev)) #endif diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index e2e5673..9a75876 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -71,6 +71,11 @@ int r100_copy_blit(struct radeon_device *rdev, uint64_t dst_offset, unsigned num_pages, struct radeon_fence *fence); +int r100_set_surface_reg(struct radeon_device *rdev, int reg, + uint32_t tiling_flags, uint32_t pitch, + uint32_t offset, uint32_t obj_size); +int r100_clear_surface_reg(struct radeon_device *rdev, int reg); +void r100_bandwidth_update(struct radeon_device *rdev); static struct radeon_asic r100_asic = { .init = &r100_init, @@ -100,6 +105,9 @@ static struct radeon_asic r100_asic = { .set_memory_clock = NULL, .set_pcie_lanes = NULL, .set_clock_gating = &radeon_legacy_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &r100_bandwidth_update, }; @@ -128,6 +136,7 @@ int r300_copy_dma(struct radeon_device *rdev, uint64_t dst_offset, unsigned num_pages, struct radeon_fence *fence); + static struct radeon_asic r300_asic = { .init = &r300_init, .errata = &r300_errata, @@ -156,6 +165,9 @@ static struct radeon_asic r300_asic = { .set_memory_clock = NULL, .set_pcie_lanes = &rv370_set_pcie_lanes, .set_clock_gating = &radeon_legacy_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &r100_bandwidth_update, }; /* @@ -193,6 +205,9 @@ static struct radeon_asic r420_asic = { .set_memory_clock = &radeon_atom_set_memory_clock, .set_pcie_lanes = &rv370_set_pcie_lanes, .set_clock_gating = &radeon_atom_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &r100_bandwidth_update, }; @@ -237,6 +252,9 @@ static struct radeon_asic rs400_asic = { .set_memory_clock = NULL, .set_pcie_lanes = NULL, .set_clock_gating = &radeon_legacy_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &r100_bandwidth_update, }; @@ -254,6 +272,7 @@ void rs600_gart_tlb_flush(struct radeon_device *rdev); int rs600_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr); uint32_t rs600_mc_rreg(struct radeon_device *rdev, uint32_t reg); void rs600_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); +void rs600_bandwidth_update(struct radeon_device *rdev); static struct radeon_asic rs600_asic = { .init = &r300_init, .errata = &rs600_errata, @@ -282,6 +301,7 @@ static struct radeon_asic rs600_asic = { .set_memory_clock = &radeon_atom_set_memory_clock, .set_pcie_lanes = NULL, .set_clock_gating = &radeon_atom_set_clock_gating, + .bandwidth_update = &rs600_bandwidth_update, }; @@ -294,6 +314,7 @@ int rs690_mc_init(struct radeon_device *rdev); void rs690_mc_fini(struct radeon_device *rdev); uint32_t rs690_mc_rreg(struct radeon_device *rdev, uint32_t reg); void rs690_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); +void rs690_bandwidth_update(struct radeon_device *rdev); static struct radeon_asic rs690_asic = { .init = &r300_init, .errata = &rs690_errata, @@ -322,6 +343,9 @@ static struct radeon_asic rs690_asic = { .set_memory_clock = &radeon_atom_set_memory_clock, .set_pcie_lanes = NULL, .set_clock_gating = &radeon_atom_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &rs690_bandwidth_update, }; @@ -339,6 +363,7 @@ void rv515_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); void rv515_ring_start(struct radeon_device *rdev); uint32_t rv515_pcie_rreg(struct radeon_device *rdev, uint32_t reg); void rv515_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); +void rv515_bandwidth_update(struct radeon_device *rdev); static struct radeon_asic rv515_asic = { .init = &rv515_init, .errata = &rv515_errata, @@ -367,6 +392,9 @@ static struct radeon_asic rv515_asic = { .set_memory_clock = &radeon_atom_set_memory_clock, .set_pcie_lanes = &rv370_set_pcie_lanes, .set_clock_gating = &radeon_atom_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &rv515_bandwidth_update, }; @@ -377,6 +405,7 @@ void r520_errata(struct radeon_device *rdev); void r520_vram_info(struct radeon_device *rdev); int r520_mc_init(struct radeon_device *rdev); void r520_mc_fini(struct radeon_device *rdev); +void r520_bandwidth_update(struct radeon_device *rdev); static struct radeon_asic r520_asic = { .init = &rv515_init, .errata = &r520_errata, @@ -405,6 +434,9 @@ static struct radeon_asic r520_asic = { .set_memory_clock = &radeon_atom_set_memory_clock, .set_pcie_lanes = &rv370_set_pcie_lanes, .set_clock_gating = &radeon_atom_set_clock_gating, + .set_surface_reg = r100_set_surface_reg, + .clear_surface_reg = r100_clear_surface_reg, + .bandwidth_update = &r520_bandwidth_update, }; /* diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c index 1f5a1a4..fcfe5c0 100644 --- a/drivers/gpu/drm/radeon/radeon_atombios.c +++ b/drivers/gpu/drm/radeon/radeon_atombios.c @@ -103,7 +103,8 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_gpio(struct drm_device static bool radeon_atom_apply_quirks(struct drm_device *dev, uint32_t supported_device, int *connector_type, - struct radeon_i2c_bus_rec *i2c_bus) + struct radeon_i2c_bus_rec *i2c_bus, + uint8_t *line_mux) { /* Asus M2A-VM HDMI board lists the DVI port as HDMI */ @@ -127,8 +128,10 @@ static bool radeon_atom_apply_quirks(struct drm_device *dev, if ((dev->pdev->device == 0x5653) && (dev->pdev->subsystem_vendor == 0x1462) && (dev->pdev->subsystem_device == 0x0291)) { - if (*connector_type == DRM_MODE_CONNECTOR_LVDS) + if (*connector_type == DRM_MODE_CONNECTOR_LVDS) { i2c_bus->valid = false; + *line_mux = 53; + } } /* Funky macbooks */ @@ -526,7 +529,7 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct if (!radeon_atom_apply_quirks (dev, (1 << i), &bios_connectors[i].connector_type, - &bios_connectors[i].ddc_bus)) + &bios_connectors[i].ddc_bus, &bios_connectors[i].line_mux)) continue; bios_connectors[i].valid = true; diff --git a/drivers/gpu/drm/radeon/radeon_benchmark.c b/drivers/gpu/drm/radeon/radeon_benchmark.c index c44403a..2e938f7 100644 --- a/drivers/gpu/drm/radeon/radeon_benchmark.c +++ b/drivers/gpu/drm/radeon/radeon_benchmark.c @@ -63,7 +63,7 @@ void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize, if (r) { goto out_cleanup; } - r = radeon_copy_dma(rdev, saddr, daddr, size >> 14, fence); + r = radeon_copy_dma(rdev, saddr, daddr, size / 4096, fence); if (r) { goto out_cleanup; } @@ -88,7 +88,7 @@ void radeon_benchmark_move(struct radeon_device *rdev, unsigned bsize, if (r) { goto out_cleanup; } - r = radeon_copy_blit(rdev, saddr, daddr, size >> 14, fence); + r = radeon_copy_blit(rdev, saddr, daddr, size / 4096, fence); if (r) { goto out_cleanup; } diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index b843f9b..a169067 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -127,17 +127,23 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data) sizeof(struct drm_radeon_cs_chunk))) { return -EFAULT; } + p->chunks[i].length_dw = user_chunk.length_dw; + p->chunks[i].kdata = NULL; p->chunks[i].chunk_id = user_chunk.chunk_id; + if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) { p->chunk_relocs_idx = i; } if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_IB) { p->chunk_ib_idx = i; + /* zero length IB isn't useful */ + if (p->chunks[i].length_dw == 0) + return -EINVAL; } + p->chunks[i].length_dw = user_chunk.length_dw; cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data; - p->chunks[i].kdata = NULL; size = p->chunks[i].length_dw * sizeof(uint32_t); p->chunks[i].kdata = kzalloc(size, GFP_KERNEL); if (p->chunks[i].kdata == NULL) { diff --git a/drivers/gpu/drm/radeon/radeon_cursor.c b/drivers/gpu/drm/radeon/radeon_cursor.c index 5232441..b13c79e 100644 --- a/drivers/gpu/drm/radeon/radeon_cursor.c +++ b/drivers/gpu/drm/radeon/radeon_cursor.c @@ -111,9 +111,11 @@ static void radeon_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj, if (ASIC_IS_AVIVO(rdev)) WREG32(AVIVO_D1CUR_SURFACE_ADDRESS + radeon_crtc->crtc_offset, gpu_addr); - else + else { + radeon_crtc->legacy_cursor_offset = gpu_addr - radeon_crtc->legacy_display_base_addr; /* offset is from DISP(2)_BASE_ADDRESS */ - WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, gpu_addr); + WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, radeon_crtc->legacy_cursor_offset); + } } int radeon_crtc_cursor_set(struct drm_crtc *crtc, @@ -245,6 +247,9 @@ int radeon_crtc_cursor_move(struct drm_crtc *crtc, (RADEON_CUR_LOCK | ((xorigin ? 0 : x) << 16) | (yorigin ? 0 : y))); + /* offset is from DISP(2)_BASE_ADDRESS */ + WREG32(RADEON_CUR_OFFSET + radeon_crtc->crtc_offset, (radeon_crtc->legacy_cursor_offset + + (yorigin * 256))); } radeon_lock_cursor(crtc, false); diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index f97563d..a162ade 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -48,6 +48,8 @@ static void radeon_surface_init(struct radeon_device *rdev) i * (RADEON_SURFACE1_INFO - RADEON_SURFACE0_INFO), 0); } + /* enable surfaces */ + WREG32(RADEON_SURFACE_CNTL, 0); } } @@ -119,7 +121,7 @@ int radeon_mc_setup(struct radeon_device *rdev) if (rdev->mc.vram_location != 0xFFFFFFFFUL) { /* vram location was already setup try to put gtt after * if it fits */ - tmp = rdev->mc.vram_location + rdev->mc.vram_size; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size; tmp = (tmp + rdev->mc.gtt_size - 1) & ~(rdev->mc.gtt_size - 1); if ((0xFFFFFFFFUL - tmp) >= rdev->mc.gtt_size) { rdev->mc.gtt_location = tmp; @@ -134,13 +136,13 @@ int radeon_mc_setup(struct radeon_device *rdev) } else if (rdev->mc.gtt_location != 0xFFFFFFFFUL) { /* gtt location was already setup try to put vram before * if it fits */ - if (rdev->mc.vram_size < rdev->mc.gtt_location) { + if (rdev->mc.mc_vram_size < rdev->mc.gtt_location) { rdev->mc.vram_location = 0; } else { tmp = rdev->mc.gtt_location + rdev->mc.gtt_size; - tmp += (rdev->mc.vram_size - 1); - tmp &= ~(rdev->mc.vram_size - 1); - if ((0xFFFFFFFFUL - tmp) >= rdev->mc.vram_size) { + tmp += (rdev->mc.mc_vram_size - 1); + tmp &= ~(rdev->mc.mc_vram_size - 1); + if ((0xFFFFFFFFUL - tmp) >= rdev->mc.mc_vram_size) { rdev->mc.vram_location = tmp; } else { printk(KERN_ERR "[drm] vram too big to fit " @@ -150,12 +152,14 @@ int radeon_mc_setup(struct radeon_device *rdev) } } else { rdev->mc.vram_location = 0; - rdev->mc.gtt_location = rdev->mc.vram_size; + rdev->mc.gtt_location = rdev->mc.mc_vram_size; } - DRM_INFO("radeon: VRAM %uM\n", rdev->mc.vram_size >> 20); + DRM_INFO("radeon: VRAM %uM\n", rdev->mc.real_vram_size >> 20); DRM_INFO("radeon: VRAM from 0x%08X to 0x%08X\n", rdev->mc.vram_location, - rdev->mc.vram_location + rdev->mc.vram_size - 1); + rdev->mc.vram_location + rdev->mc.mc_vram_size - 1); + if (rdev->mc.real_vram_size != rdev->mc.mc_vram_size) + DRM_INFO("radeon: VRAM less than aperture workaround enabled\n"); DRM_INFO("radeon: GTT %uM\n", rdev->mc.gtt_size >> 20); DRM_INFO("radeon: GTT from 0x%08X to 0x%08X\n", rdev->mc.gtt_location, @@ -450,6 +454,7 @@ int radeon_device_init(struct radeon_device *rdev, uint32_t flags) { int r, ret; + int dma_bits; DRM_INFO("radeon: Initializing kernel modesetting.\n"); rdev->shutdown = false; @@ -492,8 +497,20 @@ int radeon_device_init(struct radeon_device *rdev, return r; } - /* Report DMA addressing limitation */ - r = pci_set_dma_mask(rdev->pdev, DMA_BIT_MASK(32)); + /* set DMA mask + need_dma32 flags. + * PCIE - can handle 40-bits. + * IGP - can handle 40-bits (in theory) + * AGP - generally dma32 is safest + * PCI - only dma32 + */ + rdev->need_dma32 = false; + if (rdev->flags & RADEON_IS_AGP) + rdev->need_dma32 = true; + if (rdev->flags & RADEON_IS_PCI) + rdev->need_dma32 = true; + + dma_bits = rdev->need_dma32 ? 32 : 40; + r = pci_set_dma_mask(rdev->pdev, DMA_BIT_MASK(dma_bits)); if (r) { printk(KERN_WARNING "radeon: No suitable DMA available.\n"); } @@ -546,27 +563,22 @@ int radeon_device_init(struct radeon_device *rdev, radeon_combios_asic_init(rdev->ddev); } } + /* Initialize clocks */ + r = radeon_clocks_init(rdev); + if (r) { + return r; + } /* Get vram informations */ radeon_vram_info(rdev); - /* Device is severly broken if aper size > vram size. - * for RN50/M6/M7 - Novell bug 204882 ? - */ - if (rdev->mc.vram_size < rdev->mc.aper_size) { - rdev->mc.aper_size = rdev->mc.vram_size; - } + /* Add an MTRR for the VRAM */ rdev->mc.vram_mtrr = mtrr_add(rdev->mc.aper_base, rdev->mc.aper_size, MTRR_TYPE_WRCOMB, 1); DRM_INFO("Detected VRAM RAM=%uM, BAR=%uM\n", - rdev->mc.vram_size >> 20, + rdev->mc.real_vram_size >> 20, (unsigned)rdev->mc.aper_size >> 20); DRM_INFO("RAM width %dbits %cDR\n", rdev->mc.vram_width, rdev->mc.vram_is_ddr ? 'D' : 'S'); - /* Initialize clocks */ - r = radeon_clocks_init(rdev); - if (r) { - return r; - } /* Initialize memory controller (also test AGP) */ r = radeon_mc_init(rdev); if (r) { @@ -626,6 +638,9 @@ int radeon_device_init(struct radeon_device *rdev, if (!ret) { DRM_INFO("radeon: kernel modesetting successfully initialized.\n"); } + if (radeon_testing) { + radeon_test_moves(rdev); + } if (radeon_benchmarking) { radeon_benchmark(rdev); } diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c index 3efcf1a..a8fa1bb 100644 --- a/drivers/gpu/drm/radeon/radeon_display.c +++ b/drivers/gpu/drm/radeon/radeon_display.c @@ -187,6 +187,7 @@ static void radeon_crtc_init(struct drm_device *dev, int index) drm_mode_crtc_set_gamma_size(&radeon_crtc->base, 256); radeon_crtc->crtc_id = index; + rdev->mode_info.crtcs[index] = radeon_crtc; radeon_crtc->mode_set.crtc = &radeon_crtc->base; radeon_crtc->mode_set.connectors = (struct drm_connector **)(radeon_crtc + 1); @@ -491,7 +492,11 @@ void radeon_compute_pll(struct radeon_pll *pll, tmp += (uint64_t)pll->reference_freq * 1000 * frac_feedback_div; current_freq = radeon_div(tmp, ref_div * post_div); - error = abs(current_freq - freq); + if (flags & RADEON_PLL_PREFER_CLOSEST_LOWER) { + error = freq - current_freq; + error = error < 0 ? 0xffffffff : error; + } else + error = abs(current_freq - freq); vco_diff = abs(vco - best_vco); if ((best_vco == 0 && error < best_error) || @@ -657,36 +662,51 @@ void radeon_modeset_fini(struct radeon_device *rdev) } } -void radeon_init_disp_bandwidth(struct drm_device *dev) +bool radeon_crtc_scaling_mode_fixup(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) { - struct radeon_device *rdev = dev->dev_private; - struct drm_display_mode *modes[2]; - int pixel_bytes[2]; - struct drm_crtc *crtc; - - pixel_bytes[0] = pixel_bytes[1] = 0; - modes[0] = modes[1] = NULL; - - list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { - struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + struct drm_device *dev = crtc->dev; + struct drm_encoder *encoder; + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + struct radeon_encoder *radeon_encoder; + bool first = true; - if (crtc->enabled && crtc->fb) { - modes[radeon_crtc->crtc_id] = &crtc->mode; - pixel_bytes[radeon_crtc->crtc_id] = crtc->fb->bits_per_pixel / 8; + list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { + radeon_encoder = to_radeon_encoder(encoder); + if (encoder->crtc != crtc) + continue; + if (first) { + radeon_crtc->rmx_type = radeon_encoder->rmx_type; + radeon_crtc->devices = radeon_encoder->devices; + memcpy(&radeon_crtc->native_mode, + &radeon_encoder->native_mode, + sizeof(struct radeon_native_mode)); + first = false; + } else { + if (radeon_crtc->rmx_type != radeon_encoder->rmx_type) { + /* WARNING: Right now this can't happen but + * in the future we need to check that scaling + * are consistent accross different encoder + * (ie all encoder can work with the same + * scaling). + */ + DRM_ERROR("Scaling not consistent accross encoder.\n"); + return false; + } } } - - if (ASIC_IS_AVIVO(rdev)) { - radeon_init_disp_bw_avivo(dev, - modes[0], - pixel_bytes[0], - modes[1], - pixel_bytes[1]); + if (radeon_crtc->rmx_type != RMX_OFF) { + fixed20_12 a, b; + a.full = rfixed_const(crtc->mode.vdisplay); + b.full = rfixed_const(radeon_crtc->native_mode.panel_xres); + radeon_crtc->vsc.full = rfixed_div(a, b); + a.full = rfixed_const(crtc->mode.hdisplay); + b.full = rfixed_const(radeon_crtc->native_mode.panel_yres); + radeon_crtc->hsc.full = rfixed_div(a, b); } else { - radeon_init_disp_bw_legacy(dev, - modes[0], - pixel_bytes[0], - modes[1], - pixel_bytes[1]); + radeon_crtc->vsc.full = rfixed_const(1); + radeon_crtc->hsc.full = rfixed_const(1); } + return true; } diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index 84ba69f..3cfcee1 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -89,6 +89,7 @@ int radeon_agpmode = 0; int radeon_vram_limit = 0; int radeon_gart_size = 512; /* default gart size */ int radeon_benchmarking = 0; +int radeon_testing = 0; int radeon_connector_table = 0; #endif @@ -117,6 +118,9 @@ module_param_named(gartsize, radeon_gart_size, int, 0600); MODULE_PARM_DESC(benchmark, "Run benchmark"); module_param_named(benchmark, radeon_benchmarking, int, 0444); +MODULE_PARM_DESC(test, "Run tests"); +module_param_named(test, radeon_testing, int, 0444); + MODULE_PARM_DESC(connector_table, "Force connector table"); module_param_named(connector_table, radeon_connector_table, int, 0444); #endif diff --git a/drivers/gpu/drm/radeon/radeon_encoders.c b/drivers/gpu/drm/radeon/radeon_encoders.c index c8ef0d1..0a92706 100644 --- a/drivers/gpu/drm/radeon/radeon_encoders.c +++ b/drivers/gpu/drm/radeon/radeon_encoders.c @@ -154,7 +154,6 @@ void radeon_rmx_mode_fixup(struct drm_encoder *encoder, if (mode->hdisplay < native_mode->panel_xres || mode->vdisplay < native_mode->panel_yres) { - radeon_encoder->flags |= RADEON_USE_RMX; if (ASIC_IS_AVIVO(rdev)) { adjusted_mode->hdisplay = native_mode->panel_xres; adjusted_mode->vdisplay = native_mode->panel_yres; @@ -197,15 +196,13 @@ void radeon_rmx_mode_fixup(struct drm_encoder *encoder, } } + static bool radeon_atom_mode_fixup(struct drm_encoder *encoder, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) { - struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); - radeon_encoder->flags &= ~RADEON_USE_RMX; - drm_mode_set_crtcinfo(adjusted_mode, 0); if (radeon_encoder->rmx_type != RMX_OFF) @@ -808,234 +805,6 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action) } -static void atom_rv515_force_tv_scaler(struct radeon_device *rdev) -{ - - WREG32(0x659C, 0x0); - WREG32(0x6594, 0x705); - WREG32(0x65A4, 0x10001); - WREG32(0x65D8, 0x0); - WREG32(0x65B0, 0x0); - WREG32(0x65C0, 0x0); - WREG32(0x65D4, 0x0); - WREG32(0x6578, 0x0); - WREG32(0x657C, 0x841880A8); - WREG32(0x6578, 0x1); - WREG32(0x657C, 0x84208680); - WREG32(0x6578, 0x2); - WREG32(0x657C, 0xBFF880B0); - WREG32(0x6578, 0x100); - WREG32(0x657C, 0x83D88088); - WREG32(0x6578, 0x101); - WREG32(0x657C, 0x84608680); - WREG32(0x6578, 0x102); - WREG32(0x657C, 0xBFF080D0); - WREG32(0x6578, 0x200); - WREG32(0x657C, 0x83988068); - WREG32(0x6578, 0x201); - WREG32(0x657C, 0x84A08680); - WREG32(0x6578, 0x202); - WREG32(0x657C, 0xBFF080F8); - WREG32(0x6578, 0x300); - WREG32(0x657C, 0x83588058); - WREG32(0x6578, 0x301); - WREG32(0x657C, 0x84E08660); - WREG32(0x6578, 0x302); - WREG32(0x657C, 0xBFF88120); - WREG32(0x6578, 0x400); - WREG32(0x657C, 0x83188040); - WREG32(0x6578, 0x401); - WREG32(0x657C, 0x85008660); - WREG32(0x6578, 0x402); - WREG32(0x657C, 0xBFF88150); - WREG32(0x6578, 0x500); - WREG32(0x657C, 0x82D88030); - WREG32(0x6578, 0x501); - WREG32(0x657C, 0x85408640); - WREG32(0x6578, 0x502); - WREG32(0x657C, 0xBFF88180); - WREG32(0x6578, 0x600); - WREG32(0x657C, 0x82A08018); - WREG32(0x6578, 0x601); - WREG32(0x657C, 0x85808620); - WREG32(0x6578, 0x602); - WREG32(0x657C, 0xBFF081B8); - WREG32(0x6578, 0x700); - WREG32(0x657C, 0x82608010); - WREG32(0x6578, 0x701); - WREG32(0x657C, 0x85A08600); - WREG32(0x6578, 0x702); - WREG32(0x657C, 0x800081F0); - WREG32(0x6578, 0x800); - WREG32(0x657C, 0x8228BFF8); - WREG32(0x6578, 0x801); - WREG32(0x657C, 0x85E085E0); - WREG32(0x6578, 0x802); - WREG32(0x657C, 0xBFF88228); - WREG32(0x6578, 0x10000); - WREG32(0x657C, 0x82A8BF00); - WREG32(0x6578, 0x10001); - WREG32(0x657C, 0x82A08CC0); - WREG32(0x6578, 0x10002); - WREG32(0x657C, 0x8008BEF8); - WREG32(0x6578, 0x10100); - WREG32(0x657C, 0x81F0BF28); - WREG32(0x6578, 0x10101); - WREG32(0x657C, 0x83608CA0); - WREG32(0x6578, 0x10102); - WREG32(0x657C, 0x8018BED0); - WREG32(0x6578, 0x10200); - WREG32(0x657C, 0x8148BF38); - WREG32(0x6578, 0x10201); - WREG32(0x657C, 0x84408C80); - WREG32(0x6578, 0x10202); - WREG32(0x657C, 0x8008BEB8); - WREG32(0x6578, 0x10300); - WREG32(0x657C, 0x80B0BF78); - WREG32(0x6578, 0x10301); - WREG32(0x657C, 0x85008C20); - WREG32(0x6578, 0x10302); - WREG32(0x657C, 0x8020BEA0); - WREG32(0x6578, 0x10400); - WREG32(0x657C, 0x8028BF90); - WREG32(0x6578, 0x10401); - WREG32(0x657C, 0x85E08BC0); - WREG32(0x6578, 0x10402); - WREG32(0x657C, 0x8018BE90); - WREG32(0x6578, 0x10500); - WREG32(0x657C, 0xBFB8BFB0); - WREG32(0x6578, 0x10501); - WREG32(0x657C, 0x86C08B40); - WREG32(0x6578, 0x10502); - WREG32(0x657C, 0x8010BE90); - WREG32(0x6578, 0x10600); - WREG32(0x657C, 0xBF58BFC8); - WREG32(0x6578, 0x10601); - WREG32(0x657C, 0x87A08AA0); - WREG32(0x6578, 0x10602); - WREG32(0x657C, 0x8010BE98); - WREG32(0x6578, 0x10700); - WREG32(0x657C, 0xBF10BFF0); - WREG32(0x6578, 0x10701); - WREG32(0x657C, 0x886089E0); - WREG32(0x6578, 0x10702); - WREG32(0x657C, 0x8018BEB0); - WREG32(0x6578, 0x10800); - WREG32(0x657C, 0xBED8BFE8); - WREG32(0x6578, 0x10801); - WREG32(0x657C, 0x89408940); - WREG32(0x6578, 0x10802); - WREG32(0x657C, 0xBFE8BED8); - WREG32(0x6578, 0x20000); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20001); - WREG32(0x657C, 0x90008000); - WREG32(0x6578, 0x20002); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20003); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20100); - WREG32(0x657C, 0x80108000); - WREG32(0x6578, 0x20101); - WREG32(0x657C, 0x8FE0BF70); - WREG32(0x6578, 0x20102); - WREG32(0x657C, 0xBFE880C0); - WREG32(0x6578, 0x20103); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20200); - WREG32(0x657C, 0x8018BFF8); - WREG32(0x6578, 0x20201); - WREG32(0x657C, 0x8F80BF08); - WREG32(0x6578, 0x20202); - WREG32(0x657C, 0xBFD081A0); - WREG32(0x6578, 0x20203); - WREG32(0x657C, 0xBFF88000); - WREG32(0x6578, 0x20300); - WREG32(0x657C, 0x80188000); - WREG32(0x6578, 0x20301); - WREG32(0x657C, 0x8EE0BEC0); - WREG32(0x6578, 0x20302); - WREG32(0x657C, 0xBFB082A0); - WREG32(0x6578, 0x20303); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20400); - WREG32(0x657C, 0x80188000); - WREG32(0x6578, 0x20401); - WREG32(0x657C, 0x8E00BEA0); - WREG32(0x6578, 0x20402); - WREG32(0x657C, 0xBF8883C0); - WREG32(0x6578, 0x20403); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x20500); - WREG32(0x657C, 0x80188000); - WREG32(0x6578, 0x20501); - WREG32(0x657C, 0x8D00BE90); - WREG32(0x6578, 0x20502); - WREG32(0x657C, 0xBF588500); - WREG32(0x6578, 0x20503); - WREG32(0x657C, 0x80008008); - WREG32(0x6578, 0x20600); - WREG32(0x657C, 0x80188000); - WREG32(0x6578, 0x20601); - WREG32(0x657C, 0x8BC0BE98); - WREG32(0x6578, 0x20602); - WREG32(0x657C, 0xBF308660); - WREG32(0x6578, 0x20603); - WREG32(0x657C, 0x80008008); - WREG32(0x6578, 0x20700); - WREG32(0x657C, 0x80108000); - WREG32(0x6578, 0x20701); - WREG32(0x657C, 0x8A80BEB0); - WREG32(0x6578, 0x20702); - WREG32(0x657C, 0xBF0087C0); - WREG32(0x6578, 0x20703); - WREG32(0x657C, 0x80008008); - WREG32(0x6578, 0x20800); - WREG32(0x657C, 0x80108000); - WREG32(0x6578, 0x20801); - WREG32(0x657C, 0x8920BED0); - WREG32(0x6578, 0x20802); - WREG32(0x657C, 0xBED08920); - WREG32(0x6578, 0x20803); - WREG32(0x657C, 0x80008010); - WREG32(0x6578, 0x30000); - WREG32(0x657C, 0x90008000); - WREG32(0x6578, 0x30001); - WREG32(0x657C, 0x80008000); - WREG32(0x6578, 0x30100); - WREG32(0x657C, 0x8FE0BF90); - WREG32(0x6578, 0x30101); - WREG32(0x657C, 0xBFF880A0); - WREG32(0x6578, 0x30200); - WREG32(0x657C, 0x8F60BF40); - WREG32(0x6578, 0x30201); - WREG32(0x657C, 0xBFE88180); - WREG32(0x6578, 0x30300); - WREG32(0x657C, 0x8EC0BF00); - WREG32(0x6578, 0x30301); - WREG32(0x657C, 0xBFC88280); - WREG32(0x6578, 0x30400); - WREG32(0x657C, 0x8DE0BEE0); - WREG32(0x6578, 0x30401); - WREG32(0x657C, 0xBFA083A0); - WREG32(0x6578, 0x30500); - WREG32(0x657C, 0x8CE0BED0); - WREG32(0x6578, 0x30501); - WREG32(0x657C, 0xBF7884E0); - WREG32(0x6578, 0x30600); - WREG32(0x657C, 0x8BA0BED8); - WREG32(0x6578, 0x30601); - WREG32(0x657C, 0xBF508640); - WREG32(0x6578, 0x30700); - WREG32(0x657C, 0x8A60BEE8); - WREG32(0x6578, 0x30701); - WREG32(0x657C, 0xBF2087A0); - WREG32(0x6578, 0x30800); - WREG32(0x657C, 0x8900BF00); - WREG32(0x6578, 0x30801); - WREG32(0x657C, 0xBF008900); -} - static void atombios_yuv_setup(struct drm_encoder *encoder, bool enable) { @@ -1074,129 +843,6 @@ atombios_yuv_setup(struct drm_encoder *encoder, bool enable) } static void -atombios_overscan_setup(struct drm_encoder *encoder, - struct drm_display_mode *mode, - struct drm_display_mode *adjusted_mode) -{ - struct drm_device *dev = encoder->dev; - struct radeon_device *rdev = dev->dev_private; - struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); - struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc); - SET_CRTC_OVERSCAN_PS_ALLOCATION args; - int index = GetIndexIntoMasterTable(COMMAND, SetCRTC_OverScan); - - memset(&args, 0, sizeof(args)); - - args.usOverscanRight = 0; - args.usOverscanLeft = 0; - args.usOverscanBottom = 0; - args.usOverscanTop = 0; - args.ucCRTC = radeon_crtc->crtc_id; - - if (radeon_encoder->flags & RADEON_USE_RMX) { - if (radeon_encoder->rmx_type == RMX_FULL) { - args.usOverscanRight = 0; - args.usOverscanLeft = 0; - args.usOverscanBottom = 0; - args.usOverscanTop = 0; - } else if (radeon_encoder->rmx_type == RMX_CENTER) { - args.usOverscanTop = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2; - args.usOverscanBottom = (adjusted_mode->crtc_vdisplay - mode->crtc_vdisplay) / 2; - args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2; - args.usOverscanRight = (adjusted_mode->crtc_hdisplay - mode->crtc_hdisplay) / 2; - } else if (radeon_encoder->rmx_type == RMX_ASPECT) { - int a1 = mode->crtc_vdisplay * adjusted_mode->crtc_hdisplay; - int a2 = adjusted_mode->crtc_vdisplay * mode->crtc_hdisplay; - - if (a1 > a2) { - args.usOverscanLeft = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2; - args.usOverscanRight = (adjusted_mode->crtc_hdisplay - (a2 / mode->crtc_vdisplay)) / 2; - } else if (a2 > a1) { - args.usOverscanLeft = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2; - args.usOverscanRight = (adjusted_mode->crtc_vdisplay - (a1 / mode->crtc_hdisplay)) / 2; - } - } - } - - atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); - -} - -static void -atombios_scaler_setup(struct drm_encoder *encoder) -{ - struct drm_device *dev = encoder->dev; - struct radeon_device *rdev = dev->dev_private; - struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); - struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc); - ENABLE_SCALER_PS_ALLOCATION args; - int index = GetIndexIntoMasterTable(COMMAND, EnableScaler); - /* fixme - fill in enc_priv for atom dac */ - enum radeon_tv_std tv_std = TV_STD_NTSC; - - if (!ASIC_IS_AVIVO(rdev) && radeon_crtc->crtc_id) - return; - - memset(&args, 0, sizeof(args)); - - args.ucScaler = radeon_crtc->crtc_id; - - if (radeon_encoder->devices & (ATOM_DEVICE_TV_SUPPORT)) { - switch (tv_std) { - case TV_STD_NTSC: - default: - args.ucTVStandard = ATOM_TV_NTSC; - break; - case TV_STD_PAL: - args.ucTVStandard = ATOM_TV_PAL; - break; - case TV_STD_PAL_M: - args.ucTVStandard = ATOM_TV_PALM; - break; - case TV_STD_PAL_60: - args.ucTVStandard = ATOM_TV_PAL60; - break; - case TV_STD_NTSC_J: - args.ucTVStandard = ATOM_TV_NTSCJ; - break; - case TV_STD_SCART_PAL: - args.ucTVStandard = ATOM_TV_PAL; /* ??? */ - break; - case TV_STD_SECAM: - args.ucTVStandard = ATOM_TV_SECAM; - break; - case TV_STD_PAL_CN: - args.ucTVStandard = ATOM_TV_PALCN; - break; - } - args.ucEnable = SCALER_ENABLE_MULTITAP_MODE; - } else if (radeon_encoder->devices & (ATOM_DEVICE_CV_SUPPORT)) { - args.ucTVStandard = ATOM_TV_CV; - args.ucEnable = SCALER_ENABLE_MULTITAP_MODE; - } else if (radeon_encoder->flags & RADEON_USE_RMX) { - if (radeon_encoder->rmx_type == RMX_FULL) - args.ucEnable = ATOM_SCALER_EXPANSION; - else if (radeon_encoder->rmx_type == RMX_CENTER) - args.ucEnable = ATOM_SCALER_CENTER; - else if (radeon_encoder->rmx_type == RMX_ASPECT) - args.ucEnable = ATOM_SCALER_EXPANSION; - } else { - if (ASIC_IS_AVIVO(rdev)) - args.ucEnable = ATOM_SCALER_DISABLE; - else - args.ucEnable = ATOM_SCALER_CENTER; - } - - atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); - - if (radeon_encoder->devices & (ATOM_DEVICE_CV_SUPPORT | ATOM_DEVICE_TV_SUPPORT) - && rdev->family >= CHIP_RV515 && rdev->family <= CHIP_RV570) { - atom_rv515_force_tv_scaler(rdev); - } - -} - -static void radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode) { struct drm_device *dev = encoder->dev; @@ -1448,8 +1094,6 @@ radeon_atom_encoder_mode_set(struct drm_encoder *encoder, radeon_encoder->pixel_clock = adjusted_mode->clock; radeon_atombios_encoder_crtc_scratch_regs(encoder, radeon_crtc->crtc_id); - atombios_overscan_setup(encoder, mode, adjusted_mode); - atombios_scaler_setup(encoder); atombios_set_encoder_crtc_source(encoder); if (ASIC_IS_AVIVO(rdev)) { @@ -1667,6 +1311,7 @@ radeon_add_atom_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t su radeon_encoder->encoder_id = encoder_id; radeon_encoder->devices = supported_device; + radeon_encoder->rmx_type = RMX_OFF; switch (radeon_encoder->encoder_id) { case ENCODER_OBJECT_ID_INTERNAL_LVDS: diff --git a/drivers/gpu/drm/radeon/radeon_fb.c b/drivers/gpu/drm/radeon/radeon_fb.c index 9e8f191..3206c0a 100644 --- a/drivers/gpu/drm/radeon/radeon_fb.c +++ b/drivers/gpu/drm/radeon/radeon_fb.c @@ -101,9 +101,10 @@ static int radeonfb_setcolreg(unsigned regno, break; case 24: case 32: - fb->pseudo_palette[regno] = ((red & 0xff00) << 8) | - (green & 0xff00) | - ((blue & 0xff00) >> 8); + fb->pseudo_palette[regno] = + (((red >> 8) & 0xff) << info->var.red.offset) | + (((green >> 8) & 0xff) << info->var.green.offset) | + (((blue >> 8) & 0xff) << info->var.blue.offset); break; } } @@ -154,6 +155,7 @@ static int radeonfb_check_var(struct fb_var_screeninfo *var, var->transp.length = 0; var->transp.offset = 0; break; +#ifdef __LITTLE_ENDIAN case 15: var->red.offset = 10; var->green.offset = 5; @@ -194,6 +196,28 @@ static int radeonfb_check_var(struct fb_var_screeninfo *var, var->transp.length = 8; var->transp.offset = 24; break; +#else + case 24: + var->red.offset = 8; + var->green.offset = 16; + var->blue.offset = 24; + var->red.length = 8; + var->green.length = 8; + var->blue.length = 8; + var->transp.length = 0; + var->transp.offset = 0; + break; + case 32: + var->red.offset = 8; + var->green.offset = 16; + var->blue.offset = 24; + var->red.length = 8; + var->green.length = 8; + var->blue.length = 8; + var->transp.length = 8; + var->transp.offset = 0; + break; +#endif default: return -EINVAL; } @@ -447,10 +471,10 @@ static struct notifier_block paniced = { .notifier_call = radeonfb_panic, }; -static int radeon_align_pitch(struct radeon_device *rdev, int width, int bpp) +static int radeon_align_pitch(struct radeon_device *rdev, int width, int bpp, bool tiled) { int aligned = width; - int align_large = (ASIC_IS_AVIVO(rdev)); + int align_large = (ASIC_IS_AVIVO(rdev)) || tiled; int pitch_mask = 0; switch (bpp / 8) { @@ -488,12 +512,13 @@ int radeonfb_create(struct radeon_device *rdev, u64 fb_gpuaddr; void *fbptr = NULL; unsigned long tmp; + bool fb_tiled = false; /* useful for testing */ mode_cmd.width = surface_width; mode_cmd.height = surface_height; mode_cmd.bpp = 32; /* need to align pitch with crtc limits */ - mode_cmd.pitch = radeon_align_pitch(rdev, mode_cmd.width, mode_cmd.bpp) * ((mode_cmd.bpp + 1) / 8); + mode_cmd.pitch = radeon_align_pitch(rdev, mode_cmd.width, mode_cmd.bpp, fb_tiled) * ((mode_cmd.bpp + 1) / 8); mode_cmd.depth = 24; size = mode_cmd.pitch * mode_cmd.height; @@ -511,6 +536,8 @@ int radeonfb_create(struct radeon_device *rdev, } robj = gobj->driver_private; + if (fb_tiled) + radeon_object_set_tiling_flags(robj, RADEON_TILING_MACRO|RADEON_TILING_SURFACE, mode_cmd.pitch); mutex_lock(&rdev->ddev->struct_mutex); fb = radeon_framebuffer_create(rdev->ddev, &mode_cmd, gobj); if (fb == NULL) { @@ -539,6 +566,9 @@ int radeonfb_create(struct radeon_device *rdev, } rfbdev = info->par; + if (fb_tiled) + radeon_object_check_tiling(robj, 0, 0); + ret = radeon_object_kmap(robj, &fbptr); if (ret) { goto out_unref; @@ -572,6 +602,11 @@ int radeonfb_create(struct radeon_device *rdev, info->var.width = -1; info->var.xres = fb_width; info->var.yres = fb_height; + + /* setup aperture base/size for vesafb takeover */ + info->aperture_base = rdev->ddev->mode_config.fb_base; + info->aperture_size = rdev->mc.real_vram_size; + info->fix.mmio_start = 0; info->fix.mmio_len = 0; info->pixmap.size = 64*1024; @@ -600,6 +635,7 @@ int radeonfb_create(struct radeon_device *rdev, info->var.transp.offset = 0; info->var.transp.length = 0; break; +#ifdef __LITTLE_ENDIAN case 15: info->var.red.offset = 10; info->var.green.offset = 5; @@ -639,7 +675,29 @@ int radeonfb_create(struct radeon_device *rdev, info->var.transp.offset = 24; info->var.transp.length = 8; break; +#else + case 24: + info->var.red.offset = 8; + info->var.green.offset = 16; + info->var.blue.offset = 24; + info->var.red.length = 8; + info->var.green.length = 8; + info->var.blue.length = 8; + info->var.transp.offset = 0; + info->var.transp.length = 0; + break; + case 32: + info->var.red.offset = 8; + info->var.green.offset = 16; + info->var.blue.offset = 24; + info->var.red.length = 8; + info->var.green.length = 8; + info->var.blue.length = 8; + info->var.transp.offset = 0; + info->var.transp.length = 8; + break; default: +#endif break; } diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 96afbf5..b4e48dd 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -195,7 +195,7 @@ retry: r = wait_event_interruptible_timeout(rdev->fence_drv.queue, radeon_fence_signaled(fence), timeout); if (unlikely(r == -ERESTARTSYS)) { - return -ERESTART; + return -EBUSY; } } else { r = wait_event_timeout(rdev->fence_drv.queue, diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c index d343a15..2977539 100644 --- a/drivers/gpu/drm/radeon/radeon_gart.c +++ b/drivers/gpu/drm/radeon/radeon_gart.c @@ -177,7 +177,7 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset, return -ENOMEM; } rdev->gart.pages[p] = pagelist[i]; - page_base = (uint32_t)rdev->gart.pages_addr[p]; + page_base = rdev->gart.pages_addr[p]; for (j = 0; j < (PAGE_SIZE / 4096); j++, t++) { radeon_gart_set_page(rdev, t, page_base); page_base += 4096; diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c index eb51603..cded518 100644 --- a/drivers/gpu/drm/radeon/radeon_gem.c +++ b/drivers/gpu/drm/radeon/radeon_gem.c @@ -157,9 +157,9 @@ int radeon_gem_info_ioctl(struct drm_device *dev, void *data, struct radeon_device *rdev = dev->dev_private; struct drm_radeon_gem_info *args = data; - args->vram_size = rdev->mc.vram_size; + args->vram_size = rdev->mc.real_vram_size; /* FIXME: report somethings that makes sense */ - args->vram_visible = rdev->mc.vram_size - (4 * 1024 * 1024); + args->vram_visible = rdev->mc.real_vram_size - (4 * 1024 * 1024); args->gart_size = rdev->mc.gtt_size; return 0; } @@ -285,3 +285,44 @@ int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data, mutex_unlock(&dev->struct_mutex); return r; } + +int radeon_gem_set_tiling_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct drm_radeon_gem_set_tiling *args = data; + struct drm_gem_object *gobj; + struct radeon_object *robj; + int r = 0; + + DRM_DEBUG("%d \n", args->handle); + gobj = drm_gem_object_lookup(dev, filp, args->handle); + if (gobj == NULL) + return -EINVAL; + robj = gobj->driver_private; + radeon_object_set_tiling_flags(robj, args->tiling_flags, args->pitch); + mutex_lock(&dev->struct_mutex); + drm_gem_object_unreference(gobj); + mutex_unlock(&dev->struct_mutex); + return r; +} + +int radeon_gem_get_tiling_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct drm_radeon_gem_get_tiling *args = data; + struct drm_gem_object *gobj; + struct radeon_object *robj; + int r = 0; + + DRM_DEBUG("\n"); + gobj = drm_gem_object_lookup(dev, filp, args->handle); + if (gobj == NULL) + return -EINVAL; + robj = gobj->driver_private; + radeon_object_get_tiling_flags(robj, &args->tiling_flags, + &args->pitch); + mutex_lock(&dev->struct_mutex); + drm_gem_object_unreference(gobj); + mutex_unlock(&dev->struct_mutex); + return r; +} diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 4612a7c..937a2f1 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -291,5 +291,7 @@ struct drm_ioctl_desc radeon_ioctls_kms[] = { DRM_IOCTL_DEF(DRM_RADEON_GEM_WAIT_IDLE, radeon_gem_wait_idle_ioctl, DRM_AUTH), DRM_IOCTL_DEF(DRM_RADEON_CS, radeon_cs_ioctl, DRM_AUTH), DRM_IOCTL_DEF(DRM_RADEON_INFO, radeon_info_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_TILING, radeon_gem_set_tiling_ioctl, DRM_AUTH), + DRM_IOCTL_DEF(DRM_RADEON_GEM_GET_TILING, radeon_gem_get_tiling_ioctl, DRM_AUTH), }; int radeon_max_kms_ioctl = DRM_ARRAY_SIZE(radeon_ioctls_kms); diff --git a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c index 8086ecf..7d06dc9 100644 --- a/drivers/gpu/drm/radeon/radeon_legacy_crtc.c +++ b/drivers/gpu/drm/radeon/radeon_legacy_crtc.c @@ -29,6 +29,171 @@ #include "radeon_fixed.h" #include "radeon.h" +static void radeon_legacy_rmx_mode_set(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode) +{ + struct drm_device *dev = crtc->dev; + struct radeon_device *rdev = dev->dev_private; + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + int xres = mode->hdisplay; + int yres = mode->vdisplay; + bool hscale = true, vscale = true; + int hsync_wid; + int vsync_wid; + int hsync_start; + int blank_width; + u32 scale, inc, crtc_more_cntl; + u32 fp_horz_stretch, fp_vert_stretch, fp_horz_vert_active; + u32 fp_h_sync_strt_wid, fp_crtc_h_total_disp; + u32 fp_v_sync_strt_wid, fp_crtc_v_total_disp; + struct radeon_native_mode *native_mode = &radeon_crtc->native_mode; + + fp_vert_stretch = RREG32(RADEON_FP_VERT_STRETCH) & + (RADEON_VERT_STRETCH_RESERVED | + RADEON_VERT_AUTO_RATIO_INC); + fp_horz_stretch = RREG32(RADEON_FP_HORZ_STRETCH) & + (RADEON_HORZ_FP_LOOP_STRETCH | + RADEON_HORZ_AUTO_RATIO_INC); + + crtc_more_cntl = 0; + if ((rdev->family == CHIP_RS100) || + (rdev->family == CHIP_RS200)) { + /* This is to workaround the asic bug for RMX, some versions + of BIOS dosen't have this register initialized correctly. */ + crtc_more_cntl |= RADEON_CRTC_H_CUTOFF_ACTIVE_EN; + } + + + fp_crtc_h_total_disp = ((((mode->crtc_htotal / 8) - 1) & 0x3ff) + | ((((mode->crtc_hdisplay / 8) - 1) & 0x1ff) << 16)); + + hsync_wid = (mode->crtc_hsync_end - mode->crtc_hsync_start) / 8; + if (!hsync_wid) + hsync_wid = 1; + hsync_start = mode->crtc_hsync_start - 8; + + fp_h_sync_strt_wid = ((hsync_start & 0x1fff) + | ((hsync_wid & 0x3f) << 16) + | ((mode->flags & DRM_MODE_FLAG_NHSYNC) + ? RADEON_CRTC_H_SYNC_POL + : 0)); + + fp_crtc_v_total_disp = (((mode->crtc_vtotal - 1) & 0xffff) + | ((mode->crtc_vdisplay - 1) << 16)); + + vsync_wid = mode->crtc_vsync_end - mode->crtc_vsync_start; + if (!vsync_wid) + vsync_wid = 1; + + fp_v_sync_strt_wid = (((mode->crtc_vsync_start - 1) & 0xfff) + | ((vsync_wid & 0x1f) << 16) + | ((mode->flags & DRM_MODE_FLAG_NVSYNC) + ? RADEON_CRTC_V_SYNC_POL + : 0)); + + fp_horz_vert_active = 0; + + if (native_mode->panel_xres == 0 || + native_mode->panel_yres == 0) { + hscale = false; + vscale = false; + } else { + if (xres > native_mode->panel_xres) + xres = native_mode->panel_xres; + if (yres > native_mode->panel_yres) + yres = native_mode->panel_yres; + + if (xres == native_mode->panel_xres) + hscale = false; + if (yres == native_mode->panel_yres) + vscale = false; + } + + switch (radeon_crtc->rmx_type) { + case RMX_FULL: + case RMX_ASPECT: + if (!hscale) + fp_horz_stretch |= ((xres/8-1) << 16); + else { + inc = (fp_horz_stretch & RADEON_HORZ_AUTO_RATIO_INC) ? 1 : 0; + scale = ((xres + inc) * RADEON_HORZ_STRETCH_RATIO_MAX) + / native_mode->panel_xres + 1; + fp_horz_stretch |= (((scale) & RADEON_HORZ_STRETCH_RATIO_MASK) | + RADEON_HORZ_STRETCH_BLEND | + RADEON_HORZ_STRETCH_ENABLE | + ((native_mode->panel_xres/8-1) << 16)); + } + + if (!vscale) + fp_vert_stretch |= ((yres-1) << 12); + else { + inc = (fp_vert_stretch & RADEON_VERT_AUTO_RATIO_INC) ? 1 : 0; + scale = ((yres + inc) * RADEON_VERT_STRETCH_RATIO_MAX) + / native_mode->panel_yres + 1; + fp_vert_stretch |= (((scale) & RADEON_VERT_STRETCH_RATIO_MASK) | + RADEON_VERT_STRETCH_ENABLE | + RADEON_VERT_STRETCH_BLEND | + ((native_mode->panel_yres-1) << 12)); + } + break; + case RMX_CENTER: + fp_horz_stretch |= ((xres/8-1) << 16); + fp_vert_stretch |= ((yres-1) << 12); + + crtc_more_cntl |= (RADEON_CRTC_AUTO_HORZ_CENTER_EN | + RADEON_CRTC_AUTO_VERT_CENTER_EN); + + blank_width = (mode->crtc_hblank_end - mode->crtc_hblank_start) / 8; + if (blank_width > 110) + blank_width = 110; + + fp_crtc_h_total_disp = (((blank_width) & 0x3ff) + | ((((mode->crtc_hdisplay / 8) - 1) & 0x1ff) << 16)); + + hsync_wid = (mode->crtc_hsync_end - mode->crtc_hsync_start) / 8; + if (!hsync_wid) + hsync_wid = 1; + + fp_h_sync_strt_wid = ((((mode->crtc_hsync_start - mode->crtc_hblank_start) / 8) & 0x1fff) + | ((hsync_wid & 0x3f) << 16) + | ((mode->flags & DRM_MODE_FLAG_NHSYNC) + ? RADEON_CRTC_H_SYNC_POL + : 0)); + + fp_crtc_v_total_disp = (((mode->crtc_vblank_end - mode->crtc_vblank_start) & 0xffff) + | ((mode->crtc_vdisplay - 1) << 16)); + + vsync_wid = mode->crtc_vsync_end - mode->crtc_vsync_start; + if (!vsync_wid) + vsync_wid = 1; + + fp_v_sync_strt_wid = ((((mode->crtc_vsync_start - mode->crtc_vblank_start) & 0xfff) + | ((vsync_wid & 0x1f) << 16) + | ((mode->flags & DRM_MODE_FLAG_NVSYNC) + ? RADEON_CRTC_V_SYNC_POL + : 0))); + + fp_horz_vert_active = (((native_mode->panel_yres) & 0xfff) | + (((native_mode->panel_xres / 8) & 0x1ff) << 16)); + break; + case RMX_OFF: + default: + fp_horz_stretch |= ((xres/8-1) << 16); + fp_vert_stretch |= ((yres-1) << 12); + break; + } + + WREG32(RADEON_FP_HORZ_STRETCH, fp_horz_stretch); + WREG32(RADEON_FP_VERT_STRETCH, fp_vert_stretch); + WREG32(RADEON_CRTC_MORE_CNTL, crtc_more_cntl); + WREG32(RADEON_FP_HORZ_VERT_ACTIVE, fp_horz_vert_active); + WREG32(RADEON_FP_H_SYNC_STRT_WID, fp_h_sync_strt_wid); + WREG32(RADEON_FP_V_SYNC_STRT_WID, fp_v_sync_strt_wid); + WREG32(RADEON_FP_CRTC_H_TOTAL_DISP, fp_crtc_h_total_disp); + WREG32(RADEON_FP_CRTC_V_TOTAL_DISP, fp_crtc_v_total_disp); +} + void radeon_restore_common_regs(struct drm_device *dev) { /* don't need this yet */ @@ -235,6 +400,7 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y, uint64_t base; uint32_t crtc_offset, crtc_offset_cntl, crtc_tile_x0_y0 = 0; uint32_t crtc_pitch, pitch_pixels; + uint32_t tiling_flags; DRM_DEBUG("\n"); @@ -244,7 +410,12 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y, if (radeon_gem_object_pin(obj, RADEON_GEM_DOMAIN_VRAM, &base)) { return -EINVAL; } - crtc_offset = (u32)base; + /* if scanout was in GTT this really wouldn't work */ + /* crtc offset is from display base addr not FB location */ + radeon_crtc->legacy_display_base_addr = rdev->mc.vram_location; + + base -= radeon_crtc->legacy_display_base_addr; + crtc_offset_cntl = 0; pitch_pixels = crtc->fb->pitch / (crtc->fb->bits_per_pixel / 8); @@ -253,8 +424,12 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y, (crtc->fb->bits_per_pixel * 8)); crtc_pitch |= crtc_pitch << 16; - /* TODO tiling */ - if (0) { + radeon_object_get_tiling_flags(obj->driver_private, + &tiling_flags, NULL); + if (tiling_flags & RADEON_TILING_MICRO) + DRM_ERROR("trying to scanout microtiled buffer\n"); + + if (tiling_flags & RADEON_TILING_MACRO) { if (ASIC_IS_R300(rdev)) crtc_offset_cntl |= (R300_CRTC_X_Y_MODE_EN | R300_CRTC_MICRO_TILE_BUFFER_DIS | @@ -270,15 +445,13 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y, crtc_offset_cntl &= ~RADEON_CRTC_TILE_EN; } - - /* TODO more tiling */ - if (0) { + if (tiling_flags & RADEON_TILING_MACRO) { if (ASIC_IS_R300(rdev)) { crtc_tile_x0_y0 = x | (y << 16); base &= ~0x7ff; } else { int byteshift = crtc->fb->bits_per_pixel >> 4; - int tile_addr = (((y >> 3) * crtc->fb->width + x) >> (8 - byteshift)) << 11; + int tile_addr = (((y >> 3) * pitch_pixels + x) >> (8 - byteshift)) << 11; base += tile_addr + ((x << byteshift) % 256) + ((y % 8) << 8); crtc_offset_cntl |= (y % 16); } @@ -303,11 +476,9 @@ int radeon_crtc_set_base(struct drm_crtc *crtc, int x, int y, base &= ~7; - /* update sarea TODO */ - crtc_offset = (u32)base; - WREG32(RADEON_DISPLAY_BASE_ADDR + radeon_crtc->crtc_offset, rdev->mc.vram_location); + WREG32(RADEON_DISPLAY_BASE_ADDR + radeon_crtc->crtc_offset, radeon_crtc->legacy_display_base_addr); if (ASIC_IS_R300(rdev)) { if (radeon_crtc->crtc_id) @@ -751,6 +922,8 @@ static bool radeon_crtc_mode_fixup(struct drm_crtc *crtc, struct drm_display_mode *mode, struct drm_display_mode *adjusted_mode) { + if (!radeon_crtc_scaling_mode_fixup(crtc, mode, adjusted_mode)) + return false; return true; } @@ -759,16 +932,25 @@ static int radeon_crtc_mode_set(struct drm_crtc *crtc, struct drm_display_mode *adjusted_mode, int x, int y, struct drm_framebuffer *old_fb) { - - DRM_DEBUG("\n"); + struct radeon_crtc *radeon_crtc = to_radeon_crtc(crtc); + struct drm_device *dev = crtc->dev; + struct radeon_device *rdev = dev->dev_private; /* TODO TV */ - radeon_crtc_set_base(crtc, x, y, old_fb); radeon_set_crtc_timing(crtc, adjusted_mode); radeon_set_pll(crtc, adjusted_mode); - radeon_init_disp_bandwidth(crtc->dev); - + radeon_bandwidth_update(rdev); + if (radeon_crtc->crtc_id == 0) { + radeon_legacy_rmx_mode_set(crtc, mode, adjusted_mode); + } else { + if (radeon_crtc->rmx_type != RMX_OFF) { + /* FIXME: only first crtc has rmx what should we + * do ? + */ + DRM_ERROR("Mode need scaling but only first crtc can do that.\n"); + } + } return 0; } @@ -799,478 +981,3 @@ void radeon_legacy_init_crtc(struct drm_device *dev, radeon_crtc->crtc_offset = RADEON_CRTC2_H_TOTAL_DISP - RADEON_CRTC_H_TOTAL_DISP; drm_crtc_helper_add(&radeon_crtc->base, &legacy_helper_funcs); } - -void radeon_init_disp_bw_legacy(struct drm_device *dev, - struct drm_display_mode *mode1, - uint32_t pixel_bytes1, - struct drm_display_mode *mode2, - uint32_t pixel_bytes2) -{ - struct radeon_device *rdev = dev->dev_private; - fixed20_12 trcd_ff, trp_ff, tras_ff, trbs_ff, tcas_ff; - fixed20_12 sclk_ff, mclk_ff, sclk_eff_ff, sclk_delay_ff; - fixed20_12 peak_disp_bw, mem_bw, pix_clk, pix_clk2, temp_ff, crit_point_ff; - uint32_t temp, data, mem_trcd, mem_trp, mem_tras; - fixed20_12 memtcas_ff[8] = { - fixed_init(1), - fixed_init(2), - fixed_init(3), - fixed_init(0), - fixed_init_half(1), - fixed_init_half(2), - fixed_init(0), - }; - fixed20_12 memtcas_rs480_ff[8] = { - fixed_init(0), - fixed_init(1), - fixed_init(2), - fixed_init(3), - fixed_init(0), - fixed_init_half(1), - fixed_init_half(2), - fixed_init_half(3), - }; - fixed20_12 memtcas2_ff[8] = { - fixed_init(0), - fixed_init(1), - fixed_init(2), - fixed_init(3), - fixed_init(4), - fixed_init(5), - fixed_init(6), - fixed_init(7), - }; - fixed20_12 memtrbs[8] = { - fixed_init(1), - fixed_init_half(1), - fixed_init(2), - fixed_init_half(2), - fixed_init(3), - fixed_init_half(3), - fixed_init(4), - fixed_init_half(4) - }; - fixed20_12 memtrbs_r4xx[8] = { - fixed_init(4), - fixed_init(5), - fixed_init(6), - fixed_init(7), - fixed_init(8), - fixed_init(9), - fixed_init(10), - fixed_init(11) - }; - fixed20_12 min_mem_eff; - fixed20_12 mc_latency_sclk, mc_latency_mclk, k1; - fixed20_12 cur_latency_mclk, cur_latency_sclk; - fixed20_12 disp_latency, disp_latency_overhead, disp_drain_rate, - disp_drain_rate2, read_return_rate; - fixed20_12 time_disp1_drop_priority; - int c; - int cur_size = 16; /* in octawords */ - int critical_point = 0, critical_point2; -/* uint32_t read_return_rate, time_disp1_drop_priority; */ - int stop_req, max_stop_req; - - min_mem_eff.full = rfixed_const_8(0); - /* get modes */ - if ((rdev->disp_priority == 2) && ASIC_IS_R300(rdev)) { - uint32_t mc_init_misc_lat_timer = RREG32(R300_MC_INIT_MISC_LAT_TIMER); - mc_init_misc_lat_timer &= ~(R300_MC_DISP1R_INIT_LAT_MASK << R300_MC_DISP1R_INIT_LAT_SHIFT); - mc_init_misc_lat_timer &= ~(R300_MC_DISP0R_INIT_LAT_MASK << R300_MC_DISP0R_INIT_LAT_SHIFT); - /* check crtc enables */ - if (mode2) - mc_init_misc_lat_timer |= (1 << R300_MC_DISP1R_INIT_LAT_SHIFT); - if (mode1) - mc_init_misc_lat_timer |= (1 << R300_MC_DISP0R_INIT_LAT_SHIFT); - WREG32(R300_MC_INIT_MISC_LAT_TIMER, mc_init_misc_lat_timer); - } - - /* - * determine is there is enough bw for current mode - */ - mclk_ff.full = rfixed_const(rdev->clock.default_mclk); - temp_ff.full = rfixed_const(100); - mclk_ff.full = rfixed_div(mclk_ff, temp_ff); - sclk_ff.full = rfixed_const(rdev->clock.default_sclk); - sclk_ff.full = rfixed_div(sclk_ff, temp_ff); - - temp = (rdev->mc.vram_width / 8) * (rdev->mc.vram_is_ddr ? 2 : 1); - temp_ff.full = rfixed_const(temp); - mem_bw.full = rfixed_mul(mclk_ff, temp_ff); - - pix_clk.full = 0; - pix_clk2.full = 0; - peak_disp_bw.full = 0; - if (mode1) { - temp_ff.full = rfixed_const(1000); - pix_clk.full = rfixed_const(mode1->clock); /* convert to fixed point */ - pix_clk.full = rfixed_div(pix_clk, temp_ff); - temp_ff.full = rfixed_const(pixel_bytes1); - peak_disp_bw.full += rfixed_mul(pix_clk, temp_ff); - } - if (mode2) { - temp_ff.full = rfixed_const(1000); - pix_clk2.full = rfixed_const(mode2->clock); /* convert to fixed point */ - pix_clk2.full = rfixed_div(pix_clk2, temp_ff); - temp_ff.full = rfixed_const(pixel_bytes2); - peak_disp_bw.full += rfixed_mul(pix_clk2, temp_ff); - } - - mem_bw.full = rfixed_mul(mem_bw, min_mem_eff); - if (peak_disp_bw.full >= mem_bw.full) { - DRM_ERROR("You may not have enough display bandwidth for current mode\n" - "If you have flickering problem, try to lower resolution, refresh rate, or color depth\n"); - } - - /* Get values from the EXT_MEM_CNTL register...converting its contents. */ - temp = RREG32(RADEON_MEM_TIMING_CNTL); - if ((rdev->family == CHIP_RV100) || (rdev->flags & RADEON_IS_IGP)) { /* RV100, M6, IGPs */ - mem_trcd = ((temp >> 2) & 0x3) + 1; - mem_trp = ((temp & 0x3)) + 1; - mem_tras = ((temp & 0x70) >> 4) + 1; - } else if (rdev->family == CHIP_R300 || - rdev->family == CHIP_R350) { /* r300, r350 */ - mem_trcd = (temp & 0x7) + 1; - mem_trp = ((temp >> 8) & 0x7) + 1; - mem_tras = ((temp >> 11) & 0xf) + 4; - } else if (rdev->family == CHIP_RV350 || - rdev->family <= CHIP_RV380) { - /* rv3x0 */ - mem_trcd = (temp & 0x7) + 3; - mem_trp = ((temp >> 8) & 0x7) + 3; - mem_tras = ((temp >> 11) & 0xf) + 6; - } else if (rdev->family == CHIP_R420 || - rdev->family == CHIP_R423 || - rdev->family == CHIP_RV410) { - /* r4xx */ - mem_trcd = (temp & 0xf) + 3; - if (mem_trcd > 15) - mem_trcd = 15; - mem_trp = ((temp >> 8) & 0xf) + 3; - if (mem_trp > 15) - mem_trp = 15; - mem_tras = ((temp >> 12) & 0x1f) + 6; - if (mem_tras > 31) - mem_tras = 31; - } else { /* RV200, R200 */ - mem_trcd = (temp & 0x7) + 1; - mem_trp = ((temp >> 8) & 0x7) + 1; - mem_tras = ((temp >> 12) & 0xf) + 4; - } - /* convert to FF */ - trcd_ff.full = rfixed_const(mem_trcd); - trp_ff.full = rfixed_const(mem_trp); - tras_ff.full = rfixed_const(mem_tras); - - /* Get values from the MEM_SDRAM_MODE_REG register...converting its */ - temp = RREG32(RADEON_MEM_SDRAM_MODE_REG); - data = (temp & (7 << 20)) >> 20; - if ((rdev->family == CHIP_RV100) || rdev->flags & RADEON_IS_IGP) { - if (rdev->family == CHIP_RS480) /* don't think rs400 */ - tcas_ff = memtcas_rs480_ff[data]; - else - tcas_ff = memtcas_ff[data]; - } else - tcas_ff = memtcas2_ff[data]; - - if (rdev->family == CHIP_RS400 || - rdev->family == CHIP_RS480) { - /* extra cas latency stored in bits 23-25 0-4 clocks */ - data = (temp >> 23) & 0x7; - if (data < 5) - tcas_ff.full += rfixed_const(data); - } - - if (ASIC_IS_R300(rdev) && !(rdev->flags & RADEON_IS_IGP)) { - /* on the R300, Tcas is included in Trbs. - */ - temp = RREG32(RADEON_MEM_CNTL); - data = (R300_MEM_NUM_CHANNELS_MASK & temp); - if (data == 1) { - if (R300_MEM_USE_CD_CH_ONLY & temp) { - temp = RREG32(R300_MC_IND_INDEX); - temp &= ~R300_MC_IND_ADDR_MASK; - temp |= R300_MC_READ_CNTL_CD_mcind; - WREG32(R300_MC_IND_INDEX, temp); - temp = RREG32(R300_MC_IND_DATA); - data = (R300_MEM_RBS_POSITION_C_MASK & temp); - } else { - temp = RREG32(R300_MC_READ_CNTL_AB); - data = (R300_MEM_RBS_POSITION_A_MASK & temp); - } - } else { - temp = RREG32(R300_MC_READ_CNTL_AB); - data = (R300_MEM_RBS_POSITION_A_MASK & temp); - } - if (rdev->family == CHIP_RV410 || - rdev->family == CHIP_R420 || - rdev->family == CHIP_R423) - trbs_ff = memtrbs_r4xx[data]; - else - trbs_ff = memtrbs[data]; - tcas_ff.full += trbs_ff.full; - } - - sclk_eff_ff.full = sclk_ff.full; - - if (rdev->flags & RADEON_IS_AGP) { - fixed20_12 agpmode_ff; - agpmode_ff.full = rfixed_const(radeon_agpmode); - temp_ff.full = rfixed_const_666(16); - sclk_eff_ff.full -= rfixed_mul(agpmode_ff, temp_ff); - } - /* TODO PCIE lanes may affect this - agpmode == 16?? */ - - if (ASIC_IS_R300(rdev)) { - sclk_delay_ff.full = rfixed_const(250); - } else { - if ((rdev->family == CHIP_RV100) || - rdev->flags & RADEON_IS_IGP) { - if (rdev->mc.vram_is_ddr) - sclk_delay_ff.full = rfixed_const(41); - else - sclk_delay_ff.full = rfixed_const(33); - } else { - if (rdev->mc.vram_width == 128) - sclk_delay_ff.full = rfixed_const(57); - else - sclk_delay_ff.full = rfixed_const(41); - } - } - - mc_latency_sclk.full = rfixed_div(sclk_delay_ff, sclk_eff_ff); - - if (rdev->mc.vram_is_ddr) { - if (rdev->mc.vram_width == 32) { - k1.full = rfixed_const(40); - c = 3; - } else { - k1.full = rfixed_const(20); - c = 1; - } - } else { - k1.full = rfixed_const(40); - c = 3; - } - - temp_ff.full = rfixed_const(2); - mc_latency_mclk.full = rfixed_mul(trcd_ff, temp_ff); - temp_ff.full = rfixed_const(c); - mc_latency_mclk.full += rfixed_mul(tcas_ff, temp_ff); - temp_ff.full = rfixed_const(4); - mc_latency_mclk.full += rfixed_mul(tras_ff, temp_ff); - mc_latency_mclk.full += rfixed_mul(trp_ff, temp_ff); - mc_latency_mclk.full += k1.full; - - mc_latency_mclk.full = rfixed_div(mc_latency_mclk, mclk_ff); - mc_latency_mclk.full += rfixed_div(temp_ff, sclk_eff_ff); - - /* - HW cursor time assuming worst case of full size colour cursor. - */ - temp_ff.full = rfixed_const((2 * (cur_size - (rdev->mc.vram_is_ddr + 1)))); - temp_ff.full += trcd_ff.full; - if (temp_ff.full < tras_ff.full) - temp_ff.full = tras_ff.full; - cur_latency_mclk.full = rfixed_div(temp_ff, mclk_ff); - - temp_ff.full = rfixed_const(cur_size); - cur_latency_sclk.full = rfixed_div(temp_ff, sclk_eff_ff); - /* - Find the total latency for the display data. - */ - disp_latency_overhead.full = rfixed_const(80); - disp_latency_overhead.full = rfixed_div(disp_latency_overhead, sclk_ff); - mc_latency_mclk.full += disp_latency_overhead.full + cur_latency_mclk.full; - mc_latency_sclk.full += disp_latency_overhead.full + cur_latency_sclk.full; - - if (mc_latency_mclk.full > mc_latency_sclk.full) - disp_latency.full = mc_latency_mclk.full; - else - disp_latency.full = mc_latency_sclk.full; - - /* setup Max GRPH_STOP_REQ default value */ - if (ASIC_IS_RV100(rdev)) - max_stop_req = 0x5c; - else - max_stop_req = 0x7c; - - if (mode1) { - /* CRTC1 - Set GRPH_BUFFER_CNTL register using h/w defined optimal values. - GRPH_STOP_REQ <= MIN[ 0x7C, (CRTC_H_DISP + 1) * (bit depth) / 0x10 ] - */ - stop_req = mode1->hdisplay * pixel_bytes1 / 16; - - if (stop_req > max_stop_req) - stop_req = max_stop_req; - - /* - Find the drain rate of the display buffer. - */ - temp_ff.full = rfixed_const((16/pixel_bytes1)); - disp_drain_rate.full = rfixed_div(pix_clk, temp_ff); - - /* - Find the critical point of the display buffer. - */ - crit_point_ff.full = rfixed_mul(disp_drain_rate, disp_latency); - crit_point_ff.full += rfixed_const_half(0); - - critical_point = rfixed_trunc(crit_point_ff); - - if (rdev->disp_priority == 2) { - critical_point = 0; - } - - /* - The critical point should never be above max_stop_req-4. Setting - GRPH_CRITICAL_CNTL = 0 will thus force high priority all the time. - */ - if (max_stop_req - critical_point < 4) - critical_point = 0; - - if (critical_point == 0 && mode2 && rdev->family == CHIP_R300) { - /* some R300 cards have problem with this set to 0, when CRTC2 is enabled.*/ - critical_point = 0x10; - } - - temp = RREG32(RADEON_GRPH_BUFFER_CNTL); - temp &= ~(RADEON_GRPH_STOP_REQ_MASK); - temp |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); - temp &= ~(RADEON_GRPH_START_REQ_MASK); - if ((rdev->family == CHIP_R350) && - (stop_req > 0x15)) { - stop_req -= 0x10; - } - temp |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); - temp |= RADEON_GRPH_BUFFER_SIZE; - temp &= ~(RADEON_GRPH_CRITICAL_CNTL | - RADEON_GRPH_CRITICAL_AT_SOF | - RADEON_GRPH_STOP_CNTL); - /* - Write the result into the register. - */ - WREG32(RADEON_GRPH_BUFFER_CNTL, ((temp & ~RADEON_GRPH_CRITICAL_POINT_MASK) | - (critical_point << RADEON_GRPH_CRITICAL_POINT_SHIFT))); - -#if 0 - if ((rdev->family == CHIP_RS400) || - (rdev->family == CHIP_RS480)) { - /* attempt to program RS400 disp regs correctly ??? */ - temp = RREG32(RS400_DISP1_REG_CNTL); - temp &= ~(RS400_DISP1_START_REQ_LEVEL_MASK | - RS400_DISP1_STOP_REQ_LEVEL_MASK); - WREG32(RS400_DISP1_REQ_CNTL1, (temp | - (critical_point << RS400_DISP1_START_REQ_LEVEL_SHIFT) | - (critical_point << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); - temp = RREG32(RS400_DMIF_MEM_CNTL1); - temp &= ~(RS400_DISP1_CRITICAL_POINT_START_MASK | - RS400_DISP1_CRITICAL_POINT_STOP_MASK); - WREG32(RS400_DMIF_MEM_CNTL1, (temp | - (critical_point << RS400_DISP1_CRITICAL_POINT_START_SHIFT) | - (critical_point << RS400_DISP1_CRITICAL_POINT_STOP_SHIFT))); - } -#endif - - DRM_DEBUG("GRPH_BUFFER_CNTL from to %x\n", - /* (unsigned int)info->SavedReg->grph_buffer_cntl, */ - (unsigned int)RREG32(RADEON_GRPH_BUFFER_CNTL)); - } - - if (mode2) { - u32 grph2_cntl; - stop_req = mode2->hdisplay * pixel_bytes2 / 16; - - if (stop_req > max_stop_req) - stop_req = max_stop_req; - - /* - Find the drain rate of the display buffer. - */ - temp_ff.full = rfixed_const((16/pixel_bytes2)); - disp_drain_rate2.full = rfixed_div(pix_clk2, temp_ff); - - grph2_cntl = RREG32(RADEON_GRPH2_BUFFER_CNTL); - grph2_cntl &= ~(RADEON_GRPH_STOP_REQ_MASK); - grph2_cntl |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); - grph2_cntl &= ~(RADEON_GRPH_START_REQ_MASK); - if ((rdev->family == CHIP_R350) && - (stop_req > 0x15)) { - stop_req -= 0x10; - } - grph2_cntl |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); - grph2_cntl |= RADEON_GRPH_BUFFER_SIZE; - grph2_cntl &= ~(RADEON_GRPH_CRITICAL_CNTL | - RADEON_GRPH_CRITICAL_AT_SOF | - RADEON_GRPH_STOP_CNTL); - - if ((rdev->family == CHIP_RS100) || - (rdev->family == CHIP_RS200)) - critical_point2 = 0; - else { - temp = (rdev->mc.vram_width * rdev->mc.vram_is_ddr + 1)/128; - temp_ff.full = rfixed_const(temp); - temp_ff.full = rfixed_mul(mclk_ff, temp_ff); - if (sclk_ff.full < temp_ff.full) - temp_ff.full = sclk_ff.full; - - read_return_rate.full = temp_ff.full; - - if (mode1) { - temp_ff.full = read_return_rate.full - disp_drain_rate.full; - time_disp1_drop_priority.full = rfixed_div(crit_point_ff, temp_ff); - } else { - time_disp1_drop_priority.full = 0; - } - crit_point_ff.full = disp_latency.full + time_disp1_drop_priority.full + disp_latency.full; - crit_point_ff.full = rfixed_mul(crit_point_ff, disp_drain_rate2); - crit_point_ff.full += rfixed_const_half(0); - - critical_point2 = rfixed_trunc(crit_point_ff); - - if (rdev->disp_priority == 2) { - critical_point2 = 0; - } - - if (max_stop_req - critical_point2 < 4) - critical_point2 = 0; - - } - - if (critical_point2 == 0 && rdev->family == CHIP_R300) { - /* some R300 cards have problem with this set to 0 */ - critical_point2 = 0x10; - } - - WREG32(RADEON_GRPH2_BUFFER_CNTL, ((grph2_cntl & ~RADEON_GRPH_CRITICAL_POINT_MASK) | - (critical_point2 << RADEON_GRPH_CRITICAL_POINT_SHIFT))); - - if ((rdev->family == CHIP_RS400) || - (rdev->family == CHIP_RS480)) { -#if 0 - /* attempt to program RS400 disp2 regs correctly ??? */ - temp = RREG32(RS400_DISP2_REQ_CNTL1); - temp &= ~(RS400_DISP2_START_REQ_LEVEL_MASK | - RS400_DISP2_STOP_REQ_LEVEL_MASK); - WREG32(RS400_DISP2_REQ_CNTL1, (temp | - (critical_point2 << RS400_DISP1_START_REQ_LEVEL_SHIFT) | - (critical_point2 << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); - temp = RREG32(RS400_DISP2_REQ_CNTL2); - temp &= ~(RS400_DISP2_CRITICAL_POINT_START_MASK | - RS400_DISP2_CRITICAL_POINT_STOP_MASK); - WREG32(RS400_DISP2_REQ_CNTL2, (temp | - (critical_point2 << RS400_DISP2_CRITICAL_POINT_START_SHIFT) | - (critical_point2 << RS400_DISP2_CRITICAL_POINT_STOP_SHIFT))); -#endif - WREG32(RS400_DISP2_REQ_CNTL1, 0x105DC1CC); - WREG32(RS400_DISP2_REQ_CNTL2, 0x2749D000); - WREG32(RS400_DMIF_MEM_CNTL1, 0x29CA71DC); - WREG32(RS400_DISP1_REQ_CNTL1, 0x28FBC3AC); - } - - DRM_DEBUG("GRPH2_BUFFER_CNTL from to %x\n", - (unsigned int)RREG32(RADEON_GRPH2_BUFFER_CNTL)); - } -} diff --git a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c index 2c2f42d..34d0f58 100644 --- a/drivers/gpu/drm/radeon/radeon_legacy_encoders.c +++ b/drivers/gpu/drm/radeon/radeon_legacy_encoders.c @@ -30,170 +30,6 @@ #include "atom.h" -static void radeon_legacy_rmx_mode_set(struct drm_encoder *encoder, - struct drm_display_mode *mode, - struct drm_display_mode *adjusted_mode) -{ - struct drm_device *dev = encoder->dev; - struct radeon_device *rdev = dev->dev_private; - struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); - int xres = mode->hdisplay; - int yres = mode->vdisplay; - bool hscale = true, vscale = true; - int hsync_wid; - int vsync_wid; - int hsync_start; - uint32_t scale, inc; - uint32_t fp_horz_stretch, fp_vert_stretch, crtc_more_cntl, fp_horz_vert_active; - uint32_t fp_h_sync_strt_wid, fp_v_sync_strt_wid, fp_crtc_h_total_disp, fp_crtc_v_total_disp; - struct radeon_native_mode *native_mode = &radeon_encoder->native_mode; - - DRM_DEBUG("\n"); - - fp_vert_stretch = RREG32(RADEON_FP_VERT_STRETCH) & - (RADEON_VERT_STRETCH_RESERVED | - RADEON_VERT_AUTO_RATIO_INC); - fp_horz_stretch = RREG32(RADEON_FP_HORZ_STRETCH) & - (RADEON_HORZ_FP_LOOP_STRETCH | - RADEON_HORZ_AUTO_RATIO_INC); - - crtc_more_cntl = 0; - if ((rdev->family == CHIP_RS100) || - (rdev->family == CHIP_RS200)) { - /* This is to workaround the asic bug for RMX, some versions - of BIOS dosen't have this register initialized correctly. */ - crtc_more_cntl |= RADEON_CRTC_H_CUTOFF_ACTIVE_EN; - } - - - fp_crtc_h_total_disp = ((((mode->crtc_htotal / 8) - 1) & 0x3ff) - | ((((mode->crtc_hdisplay / 8) - 1) & 0x1ff) << 16)); - - hsync_wid = (mode->crtc_hsync_end - mode->crtc_hsync_start) / 8; - if (!hsync_wid) - hsync_wid = 1; - hsync_start = mode->crtc_hsync_start - 8; - - fp_h_sync_strt_wid = ((hsync_start & 0x1fff) - | ((hsync_wid & 0x3f) << 16) - | ((mode->flags & DRM_MODE_FLAG_NHSYNC) - ? RADEON_CRTC_H_SYNC_POL - : 0)); - - fp_crtc_v_total_disp = (((mode->crtc_vtotal - 1) & 0xffff) - | ((mode->crtc_vdisplay - 1) << 16)); - - vsync_wid = mode->crtc_vsync_end - mode->crtc_vsync_start; - if (!vsync_wid) - vsync_wid = 1; - - fp_v_sync_strt_wid = (((mode->crtc_vsync_start - 1) & 0xfff) - | ((vsync_wid & 0x1f) << 16) - | ((mode->flags & DRM_MODE_FLAG_NVSYNC) - ? RADEON_CRTC_V_SYNC_POL - : 0)); - - fp_horz_vert_active = 0; - - if (native_mode->panel_xres == 0 || - native_mode->panel_yres == 0) { - hscale = false; - vscale = false; - } else { - if (xres > native_mode->panel_xres) - xres = native_mode->panel_xres; - if (yres > native_mode->panel_yres) - yres = native_mode->panel_yres; - - if (xres == native_mode->panel_xres) - hscale = false; - if (yres == native_mode->panel_yres) - vscale = false; - } - - if (radeon_encoder->flags & RADEON_USE_RMX) { - if (radeon_encoder->rmx_type != RMX_CENTER) { - if (!hscale) - fp_horz_stretch |= ((xres/8-1) << 16); - else { - inc = (fp_horz_stretch & RADEON_HORZ_AUTO_RATIO_INC) ? 1 : 0; - scale = ((xres + inc) * RADEON_HORZ_STRETCH_RATIO_MAX) - / native_mode->panel_xres + 1; - fp_horz_stretch |= (((scale) & RADEON_HORZ_STRETCH_RATIO_MASK) | - RADEON_HORZ_STRETCH_BLEND | - RADEON_HORZ_STRETCH_ENABLE | - ((native_mode->panel_xres/8-1) << 16)); - } - - if (!vscale) - fp_vert_stretch |= ((yres-1) << 12); - else { - inc = (fp_vert_stretch & RADEON_VERT_AUTO_RATIO_INC) ? 1 : 0; - scale = ((yres + inc) * RADEON_VERT_STRETCH_RATIO_MAX) - / native_mode->panel_yres + 1; - fp_vert_stretch |= (((scale) & RADEON_VERT_STRETCH_RATIO_MASK) | - RADEON_VERT_STRETCH_ENABLE | - RADEON_VERT_STRETCH_BLEND | - ((native_mode->panel_yres-1) << 12)); - } - } else if (radeon_encoder->rmx_type == RMX_CENTER) { - int blank_width; - - fp_horz_stretch |= ((xres/8-1) << 16); - fp_vert_stretch |= ((yres-1) << 12); - - crtc_more_cntl |= (RADEON_CRTC_AUTO_HORZ_CENTER_EN | - RADEON_CRTC_AUTO_VERT_CENTER_EN); - - blank_width = (mode->crtc_hblank_end - mode->crtc_hblank_start) / 8; - if (blank_width > 110) - blank_width = 110; - - fp_crtc_h_total_disp = (((blank_width) & 0x3ff) - | ((((mode->crtc_hdisplay / 8) - 1) & 0x1ff) << 16)); - - hsync_wid = (mode->crtc_hsync_end - mode->crtc_hsync_start) / 8; - if (!hsync_wid) - hsync_wid = 1; - - fp_h_sync_strt_wid = ((((mode->crtc_hsync_start - mode->crtc_hblank_start) / 8) & 0x1fff) - | ((hsync_wid & 0x3f) << 16) - | ((mode->flags & DRM_MODE_FLAG_NHSYNC) - ? RADEON_CRTC_H_SYNC_POL - : 0)); - - fp_crtc_v_total_disp = (((mode->crtc_vblank_end - mode->crtc_vblank_start) & 0xffff) - | ((mode->crtc_vdisplay - 1) << 16)); - - vsync_wid = mode->crtc_vsync_end - mode->crtc_vsync_start; - if (!vsync_wid) - vsync_wid = 1; - - fp_v_sync_strt_wid = ((((mode->crtc_vsync_start - mode->crtc_vblank_start) & 0xfff) - | ((vsync_wid & 0x1f) << 16) - | ((mode->flags & DRM_MODE_FLAG_NVSYNC) - ? RADEON_CRTC_V_SYNC_POL - : 0))); - - fp_horz_vert_active = (((native_mode->panel_yres) & 0xfff) | - (((native_mode->panel_xres / 8) & 0x1ff) << 16)); - } - } else { - fp_horz_stretch |= ((xres/8-1) << 16); - fp_vert_stretch |= ((yres-1) << 12); - } - - WREG32(RADEON_FP_HORZ_STRETCH, fp_horz_stretch); - WREG32(RADEON_FP_VERT_STRETCH, fp_vert_stretch); - WREG32(RADEON_CRTC_MORE_CNTL, crtc_more_cntl); - WREG32(RADEON_FP_HORZ_VERT_ACTIVE, fp_horz_vert_active); - WREG32(RADEON_FP_H_SYNC_STRT_WID, fp_h_sync_strt_wid); - WREG32(RADEON_FP_V_SYNC_STRT_WID, fp_v_sync_strt_wid); - WREG32(RADEON_FP_CRTC_H_TOTAL_DISP, fp_crtc_h_total_disp); - WREG32(RADEON_FP_CRTC_V_TOTAL_DISP, fp_crtc_v_total_disp); - -} - static void radeon_legacy_lvds_dpms(struct drm_encoder *encoder, int mode) { struct drm_device *dev = encoder->dev; @@ -287,9 +123,6 @@ static void radeon_legacy_lvds_mode_set(struct drm_encoder *encoder, DRM_DEBUG("\n"); - if (radeon_crtc->crtc_id == 0) - radeon_legacy_rmx_mode_set(encoder, mode, adjusted_mode); - lvds_pll_cntl = RREG32(RADEON_LVDS_PLL_CNTL); lvds_pll_cntl &= ~RADEON_LVDS_PLL_EN; @@ -318,7 +151,7 @@ static void radeon_legacy_lvds_mode_set(struct drm_encoder *encoder, if (radeon_crtc->crtc_id == 0) { if (ASIC_IS_R300(rdev)) { - if (radeon_encoder->flags & RADEON_USE_RMX) + if (radeon_encoder->rmx_type != RMX_OFF) lvds_pll_cntl |= R300_LVDS_SRC_SEL_RMX; } else lvds_gen_cntl &= ~RADEON_LVDS_SEL_CRTC2; @@ -350,8 +183,6 @@ static bool radeon_legacy_lvds_mode_fixup(struct drm_encoder *encoder, drm_mode_set_crtcinfo(adjusted_mode, 0); - radeon_encoder->flags &= ~RADEON_USE_RMX; - if (radeon_encoder->rmx_type != RMX_OFF) radeon_rmx_mode_fixup(encoder, mode, adjusted_mode); @@ -455,9 +286,6 @@ static void radeon_legacy_primary_dac_mode_set(struct drm_encoder *encoder, DRM_DEBUG("\n"); - if (radeon_crtc->crtc_id == 0) - radeon_legacy_rmx_mode_set(encoder, mode, adjusted_mode); - if (radeon_crtc->crtc_id == 0) { if (rdev->family == CHIP_R200 || ASIC_IS_R300(rdev)) { disp_output_cntl = RREG32(RADEON_DISP_OUTPUT_CNTL) & @@ -653,9 +481,6 @@ static void radeon_legacy_tmds_int_mode_set(struct drm_encoder *encoder, DRM_DEBUG("\n"); - if (radeon_crtc->crtc_id == 0) - radeon_legacy_rmx_mode_set(encoder, mode, adjusted_mode); - tmp = tmds_pll_cntl = RREG32(RADEON_TMDS_PLL_CNTL); tmp &= 0xfffff; if (rdev->family == CHIP_RV280) { @@ -711,7 +536,7 @@ static void radeon_legacy_tmds_int_mode_set(struct drm_encoder *encoder, if (radeon_crtc->crtc_id == 0) { if (ASIC_IS_R300(rdev) || rdev->family == CHIP_R200) { fp_gen_cntl &= ~R200_FP_SOURCE_SEL_MASK; - if (radeon_encoder->flags & RADEON_USE_RMX) + if (radeon_encoder->rmx_type != RMX_OFF) fp_gen_cntl |= R200_FP_SOURCE_SEL_RMX; else fp_gen_cntl |= R200_FP_SOURCE_SEL_CRTC1; @@ -820,9 +645,6 @@ static void radeon_legacy_tmds_ext_mode_set(struct drm_encoder *encoder, DRM_DEBUG("\n"); - if (radeon_crtc->crtc_id == 0) - radeon_legacy_rmx_mode_set(encoder, mode, adjusted_mode); - if (rdev->is_atom_bios) { radeon_encoder->pixel_clock = adjusted_mode->clock; atombios_external_tmds_setup(encoder, ATOM_ENABLE); @@ -856,7 +678,7 @@ static void radeon_legacy_tmds_ext_mode_set(struct drm_encoder *encoder, if (radeon_crtc->crtc_id == 0) { if ((rdev->family == CHIP_R200) || ASIC_IS_R300(rdev)) { fp2_gen_cntl &= ~R200_FP2_SOURCE_SEL_MASK; - if (radeon_encoder->flags & RADEON_USE_RMX) + if (radeon_encoder->rmx_type != RMX_OFF) fp2_gen_cntl |= R200_FP2_SOURCE_SEL_RMX; else fp2_gen_cntl |= R200_FP2_SOURCE_SEL_CRTC1; @@ -1014,9 +836,6 @@ static void radeon_legacy_tv_dac_mode_set(struct drm_encoder *encoder, DRM_DEBUG("\n"); - if (radeon_crtc->crtc_id == 0) - radeon_legacy_rmx_mode_set(encoder, mode, adjusted_mode); - if (rdev->family != CHIP_R200) { tv_dac_cntl = RREG32(RADEON_TV_DAC_CNTL); if (rdev->family == CHIP_R420 || @@ -1243,6 +1062,7 @@ radeon_add_legacy_encoder(struct drm_device *dev, uint32_t encoder_id, uint32_t radeon_encoder->encoder_id = encoder_id; radeon_encoder->devices = supported_device; + radeon_encoder->rmx_type = RMX_OFF; switch (radeon_encoder->encoder_id) { case ENCODER_OBJECT_ID_INTERNAL_LVDS: diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 9173b68..3b09a1f 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -36,6 +36,9 @@ #include <linux/i2c.h> #include <linux/i2c-id.h> #include <linux/i2c-algo-bit.h> +#include "radeon_fixed.h" + +struct radeon_device; #define to_radeon_crtc(x) container_of(x, struct radeon_crtc, base) #define to_radeon_connector(x) container_of(x, struct radeon_connector, base) @@ -124,6 +127,7 @@ struct radeon_tmds_pll { #define RADEON_PLL_PREFER_LOW_POST_DIV (1 << 8) #define RADEON_PLL_PREFER_HIGH_POST_DIV (1 << 9) #define RADEON_PLL_USE_FRAC_FB_DIV (1 << 10) +#define RADEON_PLL_PREFER_CLOSEST_LOWER (1 << 11) struct radeon_pll { uint16_t reference_freq; @@ -170,6 +174,18 @@ struct radeon_mode_info { struct atom_context *atom_context; enum radeon_connector_table connector_table; bool mode_config_initialized; + struct radeon_crtc *crtcs[2]; +}; + +struct radeon_native_mode { + /* preferred mode */ + uint32_t panel_xres, panel_yres; + uint32_t hoverplus, hsync_width; + uint32_t hblank; + uint32_t voverplus, vsync_width; + uint32_t vblank; + uint32_t dotclock; + uint32_t flags; }; struct radeon_crtc { @@ -185,19 +201,13 @@ struct radeon_crtc { uint64_t cursor_addr; int cursor_width; int cursor_height; -}; - -#define RADEON_USE_RMX 1 - -struct radeon_native_mode { - /* preferred mode */ - uint32_t panel_xres, panel_yres; - uint32_t hoverplus, hsync_width; - uint32_t hblank; - uint32_t voverplus, vsync_width; - uint32_t vblank; - uint32_t dotclock; - uint32_t flags; + uint32_t legacy_display_base_addr; + uint32_t legacy_cursor_offset; + enum radeon_rmx_type rmx_type; + uint32_t devices; + fixed20_12 vsc; + fixed20_12 hsc; + struct radeon_native_mode native_mode; }; struct radeon_encoder_primary_dac { @@ -383,16 +393,9 @@ void radeon_enc_destroy(struct drm_encoder *encoder); void radeon_copy_fb(struct drm_device *dev, struct drm_gem_object *dst_obj); void radeon_combios_asic_init(struct drm_device *dev); extern int radeon_static_clocks_init(struct drm_device *dev); -void radeon_init_disp_bw_legacy(struct drm_device *dev, - struct drm_display_mode *mode1, - uint32_t pixel_bytes1, - struct drm_display_mode *mode2, - uint32_t pixel_bytes2); -void radeon_init_disp_bw_avivo(struct drm_device *dev, - struct drm_display_mode *mode1, - uint32_t pixel_bytes1, - struct drm_display_mode *mode2, - uint32_t pixel_bytes2); -void radeon_init_disp_bandwidth(struct drm_device *dev); +bool radeon_crtc_scaling_mode_fixup(struct drm_crtc *crtc, + struct drm_display_mode *mode, + struct drm_display_mode *adjusted_mode); +void atom_rv515_force_tv_scaler(struct radeon_device *rdev); #endif diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index bac0d06..dd9ac2f 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -44,6 +44,9 @@ struct radeon_object { uint64_t gpu_addr; void *kptr; bool is_iomem; + uint32_t tiling_flags; + uint32_t pitch; + int surface_reg; }; int radeon_ttm_init(struct radeon_device *rdev); @@ -70,6 +73,7 @@ static void radeon_ttm_object_object_destroy(struct ttm_buffer_object *tobj) robj = container_of(tobj, struct radeon_object, tobj); list_del_init(&robj->list); + radeon_object_clear_surface_reg(robj); kfree(robj); } @@ -99,16 +103,16 @@ static inline uint32_t radeon_object_flags_from_domain(uint32_t domain) { uint32_t flags = 0; if (domain & RADEON_GEM_DOMAIN_VRAM) { - flags |= TTM_PL_FLAG_VRAM; + flags |= TTM_PL_FLAG_VRAM | TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED; } if (domain & RADEON_GEM_DOMAIN_GTT) { - flags |= TTM_PL_FLAG_TT; + flags |= TTM_PL_FLAG_TT | TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED; } if (domain & RADEON_GEM_DOMAIN_CPU) { - flags |= TTM_PL_FLAG_SYSTEM; + flags |= TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING; } if (!flags) { - flags |= TTM_PL_FLAG_SYSTEM; + flags |= TTM_PL_FLAG_SYSTEM | TTM_PL_MASK_CACHING; } return flags; } @@ -141,6 +145,7 @@ int radeon_object_create(struct radeon_device *rdev, } robj->rdev = rdev; robj->gobj = gobj; + robj->surface_reg = -1; INIT_LIST_HEAD(&robj->list); flags = radeon_object_flags_from_domain(domain); @@ -304,7 +309,7 @@ int radeon_object_wait(struct radeon_object *robj) } spin_lock(&robj->tobj.lock); if (robj->tobj.sync_obj) { - r = ttm_bo_wait(&robj->tobj, true, false, false); + r = ttm_bo_wait(&robj->tobj, true, true, false); } spin_unlock(&robj->tobj.lock); radeon_object_unreserve(robj); @@ -403,7 +408,6 @@ int radeon_object_list_validate(struct list_head *head, void *fence) struct radeon_object *robj; struct radeon_fence *old_fence = NULL; struct list_head *i; - uint32_t flags; int r; r = radeon_object_list_reserve(head); @@ -414,27 +418,25 @@ int radeon_object_list_validate(struct list_head *head, void *fence) list_for_each(i, head) { lobj = list_entry(i, struct radeon_object_list, list); robj = lobj->robj; - if (lobj->wdomain) { - flags = radeon_object_flags_from_domain(lobj->wdomain); - flags |= TTM_PL_FLAG_TT; - } else { - flags = radeon_object_flags_from_domain(lobj->rdomain); - flags |= TTM_PL_FLAG_TT; - flags |= TTM_PL_FLAG_VRAM; - } if (!robj->pin_count) { - robj->tobj.proposed_placement = flags | TTM_PL_MASK_CACHING; + if (lobj->wdomain) { + robj->tobj.proposed_placement = + radeon_object_flags_from_domain(lobj->wdomain); + } else { + robj->tobj.proposed_placement = + radeon_object_flags_from_domain(lobj->rdomain); + } r = ttm_buffer_object_validate(&robj->tobj, robj->tobj.proposed_placement, true, false); if (unlikely(r)) { - radeon_object_list_unreserve(head); DRM_ERROR("radeon: failed to validate.\n"); return r; } radeon_object_gpu_addr(robj); } lobj->gpu_offset = robj->gpu_addr; + lobj->tiling_flags = robj->tiling_flags; if (fence) { old_fence = (struct radeon_fence *)robj->tobj.sync_obj; robj->tobj.sync_obj = radeon_fence_ref(fence); @@ -479,3 +481,127 @@ unsigned long radeon_object_size(struct radeon_object *robj) { return robj->tobj.num_pages << PAGE_SHIFT; } + +int radeon_object_get_surface_reg(struct radeon_object *robj) +{ + struct radeon_device *rdev = robj->rdev; + struct radeon_surface_reg *reg; + struct radeon_object *old_object; + int steal; + int i; + + if (!robj->tiling_flags) + return 0; + + if (robj->surface_reg >= 0) { + reg = &rdev->surface_regs[robj->surface_reg]; + i = robj->surface_reg; + goto out; + } + + steal = -1; + for (i = 0; i < RADEON_GEM_MAX_SURFACES; i++) { + + reg = &rdev->surface_regs[i]; + if (!reg->robj) + break; + + old_object = reg->robj; + if (old_object->pin_count == 0) + steal = i; + } + + /* if we are all out */ + if (i == RADEON_GEM_MAX_SURFACES) { + if (steal == -1) + return -ENOMEM; + /* find someone with a surface reg and nuke their BO */ + reg = &rdev->surface_regs[steal]; + old_object = reg->robj; + /* blow away the mapping */ + DRM_DEBUG("stealing surface reg %d from %p\n", steal, old_object); + ttm_bo_unmap_virtual(&old_object->tobj); + old_object->surface_reg = -1; + i = steal; + } + + robj->surface_reg = i; + reg->robj = robj; + +out: + radeon_set_surface_reg(rdev, i, robj->tiling_flags, robj->pitch, + robj->tobj.mem.mm_node->start << PAGE_SHIFT, + robj->tobj.num_pages << PAGE_SHIFT); + return 0; +} + +void radeon_object_clear_surface_reg(struct radeon_object *robj) +{ + struct radeon_device *rdev = robj->rdev; + struct radeon_surface_reg *reg; + + if (robj->surface_reg == -1) + return; + + reg = &rdev->surface_regs[robj->surface_reg]; + radeon_clear_surface_reg(rdev, robj->surface_reg); + + reg->robj = NULL; + robj->surface_reg = -1; +} + +void radeon_object_set_tiling_flags(struct radeon_object *robj, + uint32_t tiling_flags, uint32_t pitch) +{ + robj->tiling_flags = tiling_flags; + robj->pitch = pitch; +} + +void radeon_object_get_tiling_flags(struct radeon_object *robj, + uint32_t *tiling_flags, + uint32_t *pitch) +{ + if (tiling_flags) + *tiling_flags = robj->tiling_flags; + if (pitch) + *pitch = robj->pitch; +} + +int radeon_object_check_tiling(struct radeon_object *robj, bool has_moved, + bool force_drop) +{ + if (!(robj->tiling_flags & RADEON_TILING_SURFACE)) + return 0; + + if (force_drop) { + radeon_object_clear_surface_reg(robj); + return 0; + } + + if (robj->tobj.mem.mem_type != TTM_PL_VRAM) { + if (!has_moved) + return 0; + + if (robj->surface_reg >= 0) + radeon_object_clear_surface_reg(robj); + return 0; + } + + if ((robj->surface_reg >= 0) && !has_moved) + return 0; + + return radeon_object_get_surface_reg(robj); +} + +void radeon_bo_move_notify(struct ttm_buffer_object *bo, + struct ttm_mem_reg *mem) +{ + struct radeon_object *robj = container_of(bo, struct radeon_object, tobj); + radeon_object_check_tiling(robj, 0, 1); +} + +void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo) +{ + struct radeon_object *robj = container_of(bo, struct radeon_object, tobj); + radeon_object_check_tiling(robj, 0, 0); +} diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index a853261..60d1593 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -126,32 +126,19 @@ static void radeon_ib_align(struct radeon_device *rdev, struct radeon_ib *ib) } } -static void radeon_ib_cpu_flush(struct radeon_device *rdev, - struct radeon_ib *ib) -{ - unsigned long tmp; - unsigned i; - - /* To force CPU cache flush ugly but seems reliable */ - for (i = 0; i < ib->length_dw; i += (rdev->cp.align_mask + 1)) { - tmp = readl(&ib->ptr[i]); - } -} - int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib) { int r = 0; mutex_lock(&rdev->ib_pool.mutex); radeon_ib_align(rdev, ib); - radeon_ib_cpu_flush(rdev, ib); if (!ib->length_dw || !rdev->cp.ready) { /* TODO: Nothings in the ib we should report. */ mutex_unlock(&rdev->ib_pool.mutex); DRM_ERROR("radeon: couldn't schedule IB(%lu).\n", ib->idx); return -EINVAL; } - /* 64 dwords should be enought for fence too */ + /* 64 dwords should be enough for fence too */ r = radeon_ring_lock(rdev, 64); if (r) { DRM_ERROR("radeon: scheduling IB failled (%d).\n", r); diff --git a/drivers/gpu/drm/radeon/radeon_share.h b/drivers/gpu/drm/radeon/radeon_share.h new file mode 100644 index 0000000..63a7735 --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_share.h @@ -0,0 +1,39 @@ +/* + * Copyright 2008 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 + */ +#ifndef __RADEON_SHARE_H__ +#define __RADEON_SHARE_H__ + +void r100_vram_init_sizes(struct radeon_device *rdev); + +void rs690_line_buffer_adjust(struct radeon_device *rdev, + struct drm_display_mode *mode1, + struct drm_display_mode *mode2); + +void rv515_bandwidth_avivo_update(struct radeon_device *rdev); + +#endif diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c new file mode 100644 index 0000000..03c33cf --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_test.c @@ -0,0 +1,209 @@ +/* + * Copyright 2009 VMware, Inc. + * + * 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: Michel Dänzer + */ +#include <drm/drmP.h> +#include <drm/radeon_drm.h> +#include "radeon_reg.h" +#include "radeon.h" + + +/* Test BO GTT->VRAM and VRAM->GTT GPU copies across the whole GTT aperture */ +void radeon_test_moves(struct radeon_device *rdev) +{ + struct radeon_object *vram_obj = NULL; + struct radeon_object **gtt_obj = NULL; + struct radeon_fence *fence = NULL; + uint64_t gtt_addr, vram_addr; + unsigned i, n, size; + int r; + + size = 1024 * 1024; + + /* Number of tests = + * (Total GTT - IB pool - writeback page - ring buffer) / test size + */ + n = (rdev->mc.gtt_size - RADEON_IB_POOL_SIZE*64*1024 - 4096 - + rdev->cp.ring_size) / size; + + gtt_obj = kzalloc(n * sizeof(*gtt_obj), GFP_KERNEL); + if (!gtt_obj) { + DRM_ERROR("Failed to allocate %d pointers\n", n); + r = 1; + goto out_cleanup; + } + + r = radeon_object_create(rdev, NULL, size, true, RADEON_GEM_DOMAIN_VRAM, + false, &vram_obj); + if (r) { + DRM_ERROR("Failed to create VRAM object\n"); + goto out_cleanup; + } + + r = radeon_object_pin(vram_obj, RADEON_GEM_DOMAIN_VRAM, &vram_addr); + if (r) { + DRM_ERROR("Failed to pin VRAM object\n"); + goto out_cleanup; + } + + for (i = 0; i < n; i++) { + void *gtt_map, *vram_map; + void **gtt_start, **gtt_end; + void **vram_start, **vram_end; + + r = radeon_object_create(rdev, NULL, size, true, + RADEON_GEM_DOMAIN_GTT, false, gtt_obj + i); + if (r) { + DRM_ERROR("Failed to create GTT object %d\n", i); + goto out_cleanup; + } + + r = radeon_object_pin(gtt_obj[i], RADEON_GEM_DOMAIN_GTT, >t_addr); + if (r) { + DRM_ERROR("Failed to pin GTT object %d\n", i); + goto out_cleanup; + } + + r = radeon_object_kmap(gtt_obj[i], >t_map); + if (r) { + DRM_ERROR("Failed to map GTT object %d\n", i); + goto out_cleanup; + } + + for (gtt_start = gtt_map, gtt_end = gtt_map + size; + gtt_start < gtt_end; + gtt_start++) + *gtt_start = gtt_start; + + radeon_object_kunmap(gtt_obj[i]); + + r = radeon_fence_create(rdev, &fence); + if (r) { + DRM_ERROR("Failed to create GTT->VRAM fence %d\n", i); + goto out_cleanup; + } + + r = radeon_copy(rdev, gtt_addr, vram_addr, size / 4096, fence); + if (r) { + DRM_ERROR("Failed GTT->VRAM copy %d\n", i); + goto out_cleanup; + } + + r = radeon_fence_wait(fence, false); + if (r) { + DRM_ERROR("Failed to wait for GTT->VRAM fence %d\n", i); + goto out_cleanup; + } + + radeon_fence_unref(&fence); + + r = radeon_object_kmap(vram_obj, &vram_map); + if (r) { + DRM_ERROR("Failed to map VRAM object after copy %d\n", i); + goto out_cleanup; + } + + for (gtt_start = gtt_map, gtt_end = gtt_map + size, + vram_start = vram_map, vram_end = vram_map + size; + vram_start < vram_end; + gtt_start++, vram_start++) { + if (*vram_start != gtt_start) { + DRM_ERROR("Incorrect GTT->VRAM copy %d: Got 0x%p, " + "expected 0x%p (GTT map 0x%p-0x%p)\n", + i, *vram_start, gtt_start, gtt_map, + gtt_end); + radeon_object_kunmap(vram_obj); + goto out_cleanup; + } + *vram_start = vram_start; + } + + radeon_object_kunmap(vram_obj); + + r = radeon_fence_create(rdev, &fence); + if (r) { + DRM_ERROR("Failed to create VRAM->GTT fence %d\n", i); + goto out_cleanup; + } + + r = radeon_copy(rdev, vram_addr, gtt_addr, size / 4096, fence); + if (r) { + DRM_ERROR("Failed VRAM->GTT copy %d\n", i); + goto out_cleanup; + } + + r = radeon_fence_wait(fence, false); + if (r) { + DRM_ERROR("Failed to wait for VRAM->GTT fence %d\n", i); + goto out_cleanup; + } + + radeon_fence_unref(&fence); + + r = radeon_object_kmap(gtt_obj[i], >t_map); + if (r) { + DRM_ERROR("Failed to map GTT object after copy %d\n", i); + goto out_cleanup; + } + + for (gtt_start = gtt_map, gtt_end = gtt_map + size, + vram_start = vram_map, vram_end = vram_map + size; + gtt_start < gtt_end; + gtt_start++, vram_start++) { + if (*gtt_start != vram_start) { + DRM_ERROR("Incorrect VRAM->GTT copy %d: Got 0x%p, " + "expected 0x%p (VRAM map 0x%p-0x%p)\n", + i, *gtt_start, vram_start, vram_map, + vram_end); + radeon_object_kunmap(gtt_obj[i]); + goto out_cleanup; + } + } + + radeon_object_kunmap(gtt_obj[i]); + + DRM_INFO("Tested GTT->VRAM and VRAM->GTT copy for GTT offset 0x%llx\n", + gtt_addr - rdev->mc.gtt_location); + } + +out_cleanup: + if (vram_obj) { + radeon_object_unpin(vram_obj); + radeon_object_unref(&vram_obj); + } + if (gtt_obj) { + for (i = 0; i < n; i++) { + if (gtt_obj[i]) { + radeon_object_unpin(gtt_obj[i]); + radeon_object_unref(>t_obj[i]); + } + } + kfree(gtt_obj); + } + if (fence) { + radeon_fence_unref(&fence); + } + if (r) { + printk(KERN_WARNING "Error while testing BO move.\n"); + } +} + diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index 1227a97..15c3531 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -355,23 +355,26 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, if (!rdev->cp.ready) { /* use memcpy */ DRM_ERROR("CP is not ready use memcpy.\n"); - return ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); + goto memcpy; } if (old_mem->mem_type == TTM_PL_VRAM && new_mem->mem_type == TTM_PL_SYSTEM) { - return radeon_move_vram_ram(bo, evict, interruptible, + r = radeon_move_vram_ram(bo, evict, interruptible, no_wait, new_mem); } else if (old_mem->mem_type == TTM_PL_SYSTEM && new_mem->mem_type == TTM_PL_VRAM) { - return radeon_move_ram_vram(bo, evict, interruptible, + r = radeon_move_ram_vram(bo, evict, interruptible, no_wait, new_mem); } else { r = radeon_move_blit(bo, evict, no_wait, new_mem, old_mem); - if (unlikely(r)) { - return r; - } } + + if (r) { +memcpy: + r = ttm_bo_move_memcpy(bo, evict, no_wait, new_mem); + } + return r; } @@ -429,6 +432,8 @@ static struct ttm_bo_driver radeon_bo_driver = { .sync_obj_flush = &radeon_sync_obj_flush, .sync_obj_unref = &radeon_sync_obj_unref, .sync_obj_ref = &radeon_sync_obj_ref, + .move_notify = &radeon_bo_move_notify, + .fault_reserve_notify = &radeon_bo_fault_reserve_notify, }; int radeon_ttm_init(struct radeon_device *rdev) @@ -442,13 +447,14 @@ int radeon_ttm_init(struct radeon_device *rdev) /* No others user of address space so set it to 0 */ r = ttm_bo_device_init(&rdev->mman.bdev, rdev->mman.mem_global_ref.object, - &radeon_bo_driver, DRM_FILE_PAGE_OFFSET); + &radeon_bo_driver, DRM_FILE_PAGE_OFFSET, + rdev->need_dma32); if (r) { DRM_ERROR("failed initializing buffer object driver(%d).\n", r); return r; } r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM, 0, - ((rdev->mc.aper_size) >> PAGE_SHIFT)); + ((rdev->mc.real_vram_size) >> PAGE_SHIFT)); if (r) { DRM_ERROR("Failed initializing VRAM heap.\n"); return r; @@ -465,7 +471,7 @@ int radeon_ttm_init(struct radeon_device *rdev) return r; } DRM_INFO("radeon: %uM of VRAM memory ready\n", - rdev->mc.vram_size / (1024 * 1024)); + rdev->mc.real_vram_size / (1024 * 1024)); r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_TT, 0, ((rdev->mc.gtt_size) >> PAGE_SHIFT)); if (r) { diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c index cc074b5..b29affd 100644 --- a/drivers/gpu/drm/radeon/rs400.c +++ b/drivers/gpu/drm/radeon/rs400.c @@ -29,6 +29,7 @@ #include <drm/drmP.h> #include "radeon_reg.h" #include "radeon.h" +#include "radeon_share.h" /* rs400,rs480 depends on : */ void r100_hdp_reset(struct radeon_device *rdev); @@ -164,7 +165,9 @@ int rs400_gart_enable(struct radeon_device *rdev) WREG32(RADEON_BUS_CNTL, tmp); } /* Table should be in 32bits address space so ignore bits above. */ - tmp = rdev->gart.table_addr & 0xfffff000; + tmp = (u32)rdev->gart.table_addr & 0xfffff000; + tmp |= (upper_32_bits(rdev->gart.table_addr) & 0xff) << 4; + WREG32_MC(RS480_GART_BASE, tmp); /* TODO: more tweaking here */ WREG32_MC(RS480_GART_FEATURE_ID, @@ -201,10 +204,17 @@ void rs400_gart_disable(struct radeon_device *rdev) int rs400_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) { + uint32_t entry; + if (i < 0 || i > rdev->gart.num_gpu_pages) { return -EINVAL; } - rdev->gart.table.ram.ptr[i] = cpu_to_le32(((uint32_t)addr) | 0xC); + + entry = (lower_32_bits(addr) & PAGE_MASK) | + ((upper_32_bits(addr) & 0xff) << 4) | + 0xc; + entry = cpu_to_le32(entry); + rdev->gart.table.ram.ptr[i] = entry; return 0; } @@ -223,10 +233,9 @@ int rs400_mc_init(struct radeon_device *rdev) rs400_gpu_init(rdev); rs400_gart_disable(rdev); - rdev->mc.gtt_location = rdev->mc.vram_size; + rdev->mc.gtt_location = rdev->mc.mc_vram_size; rdev->mc.gtt_location += (rdev->mc.gtt_size - 1); rdev->mc.gtt_location &= ~(rdev->mc.gtt_size - 1); - rdev->mc.vram_location = 0xFFFFFFFFUL; r = radeon_mc_setup(rdev); if (r) { return r; @@ -238,7 +247,7 @@ int rs400_mc_init(struct radeon_device *rdev) "programming pipes. Bad things might happen.\n"); } - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(RADEON_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(RADEON_MC_FB_START, rdev->mc.vram_location >> 16); WREG32(RADEON_MC_FB_LOCATION, tmp); @@ -284,21 +293,12 @@ void rs400_gpu_init(struct radeon_device *rdev) */ void rs400_vram_info(struct radeon_device *rdev) { - uint32_t tom; - rs400_gart_adjust_size(rdev); /* DDR for all card after R300 & IGP */ rdev->mc.vram_is_ddr = true; rdev->mc.vram_width = 128; - /* read NB_TOM to get the amount of ram stolen for the GPU */ - tom = RREG32(RADEON_NB_TOM); - rdev->mc.vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); - - /* Could aper size report 0 ? */ - rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); - rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + r100_vram_init_sizes(rdev); } diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c index ab0c967..bbea6dee 100644 --- a/drivers/gpu/drm/radeon/rs600.c +++ b/drivers/gpu/drm/radeon/rs600.c @@ -223,7 +223,7 @@ int rs600_mc_init(struct radeon_device *rdev) printk(KERN_WARNING "Failed to wait MC idle while " "programming pipes. Bad things might happen.\n"); } - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(RS600_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(RS600_MC_FB_START, rdev->mc.vram_location >> 16); WREG32_MC(RS600_MC_FB_LOCATION, tmp); @@ -301,6 +301,11 @@ void rs600_vram_info(struct radeon_device *rdev) rdev->mc.vram_width = 128; } +void rs600_bandwidth_update(struct radeon_device *rdev) +{ + /* FIXME: implement, should this be like rs690 ? */ +} + /* * Indirect registers accessor diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c index 79ba850..839595b 100644 --- a/drivers/gpu/drm/radeon/rs690.c +++ b/drivers/gpu/drm/radeon/rs690.c @@ -28,6 +28,9 @@ #include "drmP.h" #include "radeon_reg.h" #include "radeon.h" +#include "rs690r.h" +#include "atom.h" +#include "atom-bits.h" /* rs690,rs740 depends on : */ void r100_hdp_reset(struct radeon_device *rdev); @@ -64,7 +67,7 @@ int rs690_mc_init(struct radeon_device *rdev) rs400_gart_disable(rdev); /* Setup GPU memory space */ - rdev->mc.gtt_location = rdev->mc.vram_size; + rdev->mc.gtt_location = rdev->mc.mc_vram_size; rdev->mc.gtt_location += (rdev->mc.gtt_size - 1); rdev->mc.gtt_location &= ~(rdev->mc.gtt_size - 1); rdev->mc.vram_location = 0xFFFFFFFFUL; @@ -79,7 +82,7 @@ int rs690_mc_init(struct radeon_device *rdev) printk(KERN_WARNING "Failed to wait MC idle while " "programming pipes. Bad things might happen.\n"); } - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(RS690_MC_FB_TOP, tmp >> 16); tmp |= REG_SET(RS690_MC_FB_START, rdev->mc.vram_location >> 16); WREG32_MC(RS690_MCCFG_FB_LOCATION, tmp); @@ -138,9 +141,82 @@ void rs690_gpu_init(struct radeon_device *rdev) /* * VRAM info. */ +void rs690_pm_info(struct radeon_device *rdev) +{ + int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo); + struct _ATOM_INTEGRATED_SYSTEM_INFO *info; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 *info_v2; + void *ptr; + uint16_t data_offset; + uint8_t frev, crev; + fixed20_12 tmp; + + atom_parse_data_header(rdev->mode_info.atom_context, index, NULL, + &frev, &crev, &data_offset); + ptr = rdev->mode_info.atom_context->bios + data_offset; + info = (struct _ATOM_INTEGRATED_SYSTEM_INFO *)ptr; + info_v2 = (struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 *)ptr; + /* Get various system informations from bios */ + switch (crev) { + case 1: + tmp.full = rfixed_const(100); + rdev->pm.igp_sideport_mclk.full = rfixed_const(info->ulBootUpMemoryClock); + rdev->pm.igp_sideport_mclk.full = rfixed_div(rdev->pm.igp_sideport_mclk, tmp); + rdev->pm.igp_system_mclk.full = rfixed_const(le16_to_cpu(info->usK8MemoryClock)); + rdev->pm.igp_ht_link_clk.full = rfixed_const(le16_to_cpu(info->usFSBClock)); + rdev->pm.igp_ht_link_width.full = rfixed_const(info->ucHTLinkWidth); + break; + case 2: + tmp.full = rfixed_const(100); + rdev->pm.igp_sideport_mclk.full = rfixed_const(info_v2->ulBootUpSidePortClock); + rdev->pm.igp_sideport_mclk.full = rfixed_div(rdev->pm.igp_sideport_mclk, tmp); + rdev->pm.igp_system_mclk.full = rfixed_const(info_v2->ulBootUpUMAClock); + rdev->pm.igp_system_mclk.full = rfixed_div(rdev->pm.igp_system_mclk, tmp); + rdev->pm.igp_ht_link_clk.full = rfixed_const(info_v2->ulHTLinkFreq); + rdev->pm.igp_ht_link_clk.full = rfixed_div(rdev->pm.igp_ht_link_clk, tmp); + rdev->pm.igp_ht_link_width.full = rfixed_const(le16_to_cpu(info_v2->usMinHTLinkWidth)); + break; + default: + tmp.full = rfixed_const(100); + /* We assume the slower possible clock ie worst case */ + /* DDR 333Mhz */ + rdev->pm.igp_sideport_mclk.full = rfixed_const(333); + /* FIXME: system clock ? */ + rdev->pm.igp_system_mclk.full = rfixed_const(100); + rdev->pm.igp_system_mclk.full = rfixed_div(rdev->pm.igp_system_mclk, tmp); + rdev->pm.igp_ht_link_clk.full = rfixed_const(200); + rdev->pm.igp_ht_link_width.full = rfixed_const(8); + DRM_ERROR("No integrated system info for your GPU, using safe default\n"); + break; + } + /* Compute various bandwidth */ + /* k8_bandwidth = (memory_clk / 2) * 2 * 8 * 0.5 = memory_clk * 4 */ + tmp.full = rfixed_const(4); + rdev->pm.k8_bandwidth.full = rfixed_mul(rdev->pm.igp_system_mclk, tmp); + /* ht_bandwidth = ht_clk * 2 * ht_width / 8 * 0.8 + * = ht_clk * ht_width / 5 + */ + tmp.full = rfixed_const(5); + rdev->pm.ht_bandwidth.full = rfixed_mul(rdev->pm.igp_ht_link_clk, + rdev->pm.igp_ht_link_width); + rdev->pm.ht_bandwidth.full = rfixed_div(rdev->pm.ht_bandwidth, tmp); + if (tmp.full < rdev->pm.max_bandwidth.full) { + /* HT link is a limiting factor */ + rdev->pm.max_bandwidth.full = tmp.full; + } + /* sideport_bandwidth = (sideport_clk / 2) * 2 * 2 * 0.7 + * = (sideport_clk * 14) / 10 + */ + tmp.full = rfixed_const(14); + rdev->pm.sideport_bandwidth.full = rfixed_mul(rdev->pm.igp_sideport_mclk, tmp); + tmp.full = rfixed_const(10); + rdev->pm.sideport_bandwidth.full = rfixed_div(rdev->pm.sideport_bandwidth, tmp); +} + void rs690_vram_info(struct radeon_device *rdev) { uint32_t tmp; + fixed20_12 a; rs400_gart_adjust_size(rdev); /* DDR for all card after R300 & IGP */ @@ -152,12 +228,409 @@ void rs690_vram_info(struct radeon_device *rdev) } else { rdev->mc.vram_width = 64; } - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); + rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE); + rdev->mc.mc_vram_size = rdev->mc.real_vram_size; rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + rs690_pm_info(rdev); + /* FIXME: we should enforce default clock in case GPU is not in + * default setup + */ + a.full = rfixed_const(100); + rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk); + rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a); + a.full = rfixed_const(16); + /* core_bandwidth = sclk(Mhz) * 16 */ + rdev->pm.core_bandwidth.full = rfixed_div(rdev->pm.sclk, a); +} + +void rs690_line_buffer_adjust(struct radeon_device *rdev, + struct drm_display_mode *mode1, + struct drm_display_mode *mode2) +{ + u32 tmp; + + /* + * Line Buffer Setup + * There is a single line buffer shared by both display controllers. + * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between + * the display controllers. The paritioning can either be done + * manually or via one of four preset allocations specified in bits 1:0: + * 0 - line buffer is divided in half and shared between crtc + * 1 - D1 gets 3/4 of the line buffer, D2 gets 1/4 + * 2 - D1 gets the whole buffer + * 3 - D1 gets 1/4 of the line buffer, D2 gets 3/4 + * Setting bit 2 of DC_LB_MEMORY_SPLIT controls switches to manual + * allocation mode. In manual allocation mode, D1 always starts at 0, + * D1 end/2 is specified in bits 14:4; D2 allocation follows D1. + */ + tmp = RREG32(DC_LB_MEMORY_SPLIT) & ~DC_LB_MEMORY_SPLIT_MASK; + tmp &= ~DC_LB_MEMORY_SPLIT_SHIFT_MODE; + /* auto */ + if (mode1 && mode2) { + if (mode1->hdisplay > mode2->hdisplay) { + if (mode1->hdisplay > 2560) + tmp |= DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q; + else + tmp |= DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; + } else if (mode2->hdisplay > mode1->hdisplay) { + if (mode2->hdisplay > 2560) + tmp |= DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q; + else + tmp |= DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; + } else + tmp |= AVIVO_DC_LB_MEMORY_SPLIT_D1HALF_D2HALF; + } else if (mode1) { + tmp |= DC_LB_MEMORY_SPLIT_D1_ONLY; + } else if (mode2) { + tmp |= DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q; + } + WREG32(DC_LB_MEMORY_SPLIT, tmp); } +struct rs690_watermark { + u32 lb_request_fifo_depth; + fixed20_12 num_line_pair; + fixed20_12 estimated_width; + fixed20_12 worst_case_latency; + fixed20_12 consumption_rate; + fixed20_12 active_time; + fixed20_12 dbpp; + fixed20_12 priority_mark_max; + fixed20_12 priority_mark; + fixed20_12 sclk; +}; + +void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, + struct radeon_crtc *crtc, + struct rs690_watermark *wm) +{ + struct drm_display_mode *mode = &crtc->base.mode; + fixed20_12 a, b, c; + fixed20_12 pclk, request_fifo_depth, tolerable_latency, estimated_width; + fixed20_12 consumption_time, line_time, chunk_time, read_delay_latency; + /* FIXME: detect IGP with sideport memory, i don't think there is any + * such product available + */ + bool sideport = false; + + if (!crtc->base.enabled) { + /* FIXME: wouldn't it better to set priority mark to maximum */ + wm->lb_request_fifo_depth = 4; + return; + } + + if (crtc->vsc.full > rfixed_const(2)) + wm->num_line_pair.full = rfixed_const(2); + else + wm->num_line_pair.full = rfixed_const(1); + + b.full = rfixed_const(mode->crtc_hdisplay); + c.full = rfixed_const(256); + a.full = rfixed_mul(wm->num_line_pair, b); + request_fifo_depth.full = rfixed_div(a, c); + if (a.full < rfixed_const(4)) { + wm->lb_request_fifo_depth = 4; + } else { + wm->lb_request_fifo_depth = rfixed_trunc(request_fifo_depth); + } + + /* Determine consumption rate + * pclk = pixel clock period(ns) = 1000 / (mode.clock / 1000) + * vtaps = number of vertical taps, + * vsc = vertical scaling ratio, defined as source/destination + * hsc = horizontal scaling ration, defined as source/destination + */ + a.full = rfixed_const(mode->clock); + b.full = rfixed_const(1000); + a.full = rfixed_div(a, b); + pclk.full = rfixed_div(b, a); + if (crtc->rmx_type != RMX_OFF) { + b.full = rfixed_const(2); + if (crtc->vsc.full > b.full) + b.full = crtc->vsc.full; + b.full = rfixed_mul(b, crtc->hsc); + c.full = rfixed_const(2); + b.full = rfixed_div(b, c); + consumption_time.full = rfixed_div(pclk, b); + } else { + consumption_time.full = pclk.full; + } + a.full = rfixed_const(1); + wm->consumption_rate.full = rfixed_div(a, consumption_time); + + + /* Determine line time + * LineTime = total time for one line of displayhtotal + * LineTime = total number of horizontal pixels + * pclk = pixel clock period(ns) + */ + a.full = rfixed_const(crtc->base.mode.crtc_htotal); + line_time.full = rfixed_mul(a, pclk); + + /* Determine active time + * ActiveTime = time of active region of display within one line, + * hactive = total number of horizontal active pixels + * htotal = total number of horizontal pixels + */ + a.full = rfixed_const(crtc->base.mode.crtc_htotal); + b.full = rfixed_const(crtc->base.mode.crtc_hdisplay); + wm->active_time.full = rfixed_mul(line_time, b); + wm->active_time.full = rfixed_div(wm->active_time, a); + + /* Maximun bandwidth is the minimun bandwidth of all component */ + rdev->pm.max_bandwidth = rdev->pm.core_bandwidth; + if (sideport) { + if (rdev->pm.max_bandwidth.full > rdev->pm.sideport_bandwidth.full && + rdev->pm.sideport_bandwidth.full) + rdev->pm.max_bandwidth = rdev->pm.sideport_bandwidth; + read_delay_latency.full = rfixed_const(370 * 800 * 1000); + read_delay_latency.full = rfixed_div(read_delay_latency, + rdev->pm.igp_sideport_mclk); + } else { + if (rdev->pm.max_bandwidth.full > rdev->pm.k8_bandwidth.full && + rdev->pm.k8_bandwidth.full) + rdev->pm.max_bandwidth = rdev->pm.k8_bandwidth; + if (rdev->pm.max_bandwidth.full > rdev->pm.ht_bandwidth.full && + rdev->pm.ht_bandwidth.full) + rdev->pm.max_bandwidth = rdev->pm.ht_bandwidth; + read_delay_latency.full = rfixed_const(5000); + } + + /* sclk = system clocks(ns) = 1000 / max_bandwidth / 16 */ + a.full = rfixed_const(16); + rdev->pm.sclk.full = rfixed_mul(rdev->pm.max_bandwidth, a); + a.full = rfixed_const(1000); + rdev->pm.sclk.full = rfixed_div(a, rdev->pm.sclk); + /* Determine chunk time + * ChunkTime = the time it takes the DCP to send one chunk of data + * to the LB which consists of pipeline delay and inter chunk gap + * sclk = system clock(ns) + */ + a.full = rfixed_const(256 * 13); + chunk_time.full = rfixed_mul(rdev->pm.sclk, a); + a.full = rfixed_const(10); + chunk_time.full = rfixed_div(chunk_time, a); + + /* Determine the worst case latency + * NumLinePair = Number of line pairs to request(1=2 lines, 2=4 lines) + * WorstCaseLatency = worst case time from urgent to when the MC starts + * to return data + * READ_DELAY_IDLE_MAX = constant of 1us + * ChunkTime = time it takes the DCP to send one chunk of data to the LB + * which consists of pipeline delay and inter chunk gap + */ + if (rfixed_trunc(wm->num_line_pair) > 1) { + a.full = rfixed_const(3); + wm->worst_case_latency.full = rfixed_mul(a, chunk_time); + wm->worst_case_latency.full += read_delay_latency.full; + } else { + a.full = rfixed_const(2); + wm->worst_case_latency.full = rfixed_mul(a, chunk_time); + wm->worst_case_latency.full += read_delay_latency.full; + } + + /* Determine the tolerable latency + * TolerableLatency = Any given request has only 1 line time + * for the data to be returned + * LBRequestFifoDepth = Number of chunk requests the LB can + * put into the request FIFO for a display + * LineTime = total time for one line of display + * ChunkTime = the time it takes the DCP to send one chunk + * of data to the LB which consists of + * pipeline delay and inter chunk gap + */ + if ((2+wm->lb_request_fifo_depth) >= rfixed_trunc(request_fifo_depth)) { + tolerable_latency.full = line_time.full; + } else { + tolerable_latency.full = rfixed_const(wm->lb_request_fifo_depth - 2); + tolerable_latency.full = request_fifo_depth.full - tolerable_latency.full; + tolerable_latency.full = rfixed_mul(tolerable_latency, chunk_time); + tolerable_latency.full = line_time.full - tolerable_latency.full; + } + /* We assume worst case 32bits (4 bytes) */ + wm->dbpp.full = rfixed_const(4 * 8); + + /* Determine the maximum priority mark + * width = viewport width in pixels + */ + a.full = rfixed_const(16); + wm->priority_mark_max.full = rfixed_const(crtc->base.mode.crtc_hdisplay); + wm->priority_mark_max.full = rfixed_div(wm->priority_mark_max, a); + + /* Determine estimated width */ + estimated_width.full = tolerable_latency.full - wm->worst_case_latency.full; + estimated_width.full = rfixed_div(estimated_width, consumption_time); + if (rfixed_trunc(estimated_width) > crtc->base.mode.crtc_hdisplay) { + wm->priority_mark.full = rfixed_const(10); + } else { + a.full = rfixed_const(16); + wm->priority_mark.full = rfixed_div(estimated_width, a); + wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full; + } +} + +void rs690_bandwidth_update(struct radeon_device *rdev) +{ + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + struct rs690_watermark wm0; + struct rs690_watermark wm1; + u32 tmp; + fixed20_12 priority_mark02, priority_mark12, fill_rate; + fixed20_12 a, b; + + if (rdev->mode_info.crtcs[0]->base.enabled) + mode0 = &rdev->mode_info.crtcs[0]->base.mode; + if (rdev->mode_info.crtcs[1]->base.enabled) + mode1 = &rdev->mode_info.crtcs[1]->base.mode; + /* + * Set display0/1 priority up in the memory controller for + * modes if the user specifies HIGH for displaypriority + * option. + */ + if (rdev->disp_priority == 2) { + tmp = RREG32_MC(MC_INIT_MISC_LAT_TIMER); + tmp &= ~MC_DISP1R_INIT_LAT_MASK; + tmp &= ~MC_DISP0R_INIT_LAT_MASK; + if (mode1) + tmp |= (1 << MC_DISP1R_INIT_LAT_SHIFT); + if (mode0) + tmp |= (1 << MC_DISP0R_INIT_LAT_SHIFT); + WREG32_MC(MC_INIT_MISC_LAT_TIMER, tmp); + } + rs690_line_buffer_adjust(rdev, mode0, mode1); + + if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740)) + WREG32(DCP_CONTROL, 0); + if ((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880)) + WREG32(DCP_CONTROL, 2); + + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0); + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1); + + tmp = (wm0.lb_request_fifo_depth - 1); + tmp |= (wm1.lb_request_fifo_depth - 1) << 16; + WREG32(LB_MAX_REQ_OUTSTANDING, tmp); + + if (mode0 && mode1) { + if (rfixed_trunc(wm0.dbpp) > 64) + a.full = rfixed_mul(wm0.dbpp, wm0.num_line_pair); + else + a.full = wm0.num_line_pair.full; + if (rfixed_trunc(wm1.dbpp) > 64) + b.full = rfixed_mul(wm1.dbpp, wm1.num_line_pair); + else + b.full = wm1.num_line_pair.full; + a.full += b.full; + fill_rate.full = rfixed_div(wm0.sclk, a); + if (wm0.consumption_rate.full > fill_rate.full) { + b.full = wm0.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm0.active_time); + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + a.full = a.full + b.full; + b.full = rfixed_const(16 * 1000); + priority_mark02.full = rfixed_div(a, b); + } else { + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark02.full = rfixed_div(a, b); + } + if (wm1.consumption_rate.full > fill_rate.full) { + b.full = wm1.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm1.active_time); + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + a.full = a.full + b.full; + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } else { + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } + if (wm0.priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark.full; + if (rfixed_trunc(priority_mark02) < 0) + priority_mark02.full = 0; + if (wm0.priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark_max.full; + if (wm1.priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark.full; + if (rfixed_trunc(priority_mark12) < 0) + priority_mark12.full = 0; + if (wm1.priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02)); + WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02)); + WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12)); + WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12)); + } else if (mode0) { + if (rfixed_trunc(wm0.dbpp) > 64) + a.full = rfixed_mul(wm0.dbpp, wm0.num_line_pair); + else + a.full = wm0.num_line_pair.full; + fill_rate.full = rfixed_div(wm0.sclk, a); + if (wm0.consumption_rate.full > fill_rate.full) { + b.full = wm0.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm0.active_time); + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + a.full = a.full + b.full; + b.full = rfixed_const(16 * 1000); + priority_mark02.full = rfixed_div(a, b); + } else { + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark02.full = rfixed_div(a, b); + } + if (wm0.priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark.full; + if (rfixed_trunc(priority_mark02) < 0) + priority_mark02.full = 0; + if (wm0.priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02)); + WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02)); + WREG32(D2MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF); + WREG32(D2MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF); + } else { + if (rfixed_trunc(wm1.dbpp) > 64) + a.full = rfixed_mul(wm1.dbpp, wm1.num_line_pair); + else + a.full = wm1.num_line_pair.full; + fill_rate.full = rfixed_div(wm1.sclk, a); + if (wm1.consumption_rate.full > fill_rate.full) { + b.full = wm1.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm1.active_time); + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + a.full = a.full + b.full; + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } else { + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } + if (wm1.priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark.full; + if (rfixed_trunc(priority_mark12) < 0) + priority_mark12.full = 0; + if (wm1.priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF); + WREG32(D1MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF); + WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12)); + WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12)); + } +} /* * Indirect registers accessor diff --git a/drivers/gpu/drm/radeon/rs690r.h b/drivers/gpu/drm/radeon/rs690r.h new file mode 100644 index 0000000..c0d9faa --- /dev/null +++ b/drivers/gpu/drm/radeon/rs690r.h @@ -0,0 +1,99 @@ +/* + * Copyright 2008 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 + */ +#ifndef RS690R_H +#define RS690R_H + +/* RS690/RS740 registers */ +#define MC_INDEX 0x0078 +# define MC_INDEX_MASK 0x1FF +# define MC_INDEX_WR_EN (1 << 9) +# define MC_INDEX_WR_ACK 0x7F +#define MC_DATA 0x007C +#define HDP_FB_LOCATION 0x0134 +#define DC_LB_MEMORY_SPLIT 0x6520 +#define DC_LB_MEMORY_SPLIT_MASK 0x00000003 +#define DC_LB_MEMORY_SPLIT_SHIFT 0 +#define DC_LB_MEMORY_SPLIT_D1HALF_D2HALF 0 +#define DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q 1 +#define DC_LB_MEMORY_SPLIT_D1_ONLY 2 +#define DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q 3 +#define DC_LB_MEMORY_SPLIT_SHIFT_MODE (1 << 2) +#define DC_LB_DISP1_END_ADR_SHIFT 4 +#define DC_LB_DISP1_END_ADR_MASK 0x00007FF0 +#define D1MODE_PRIORITY_A_CNT 0x6548 +#define MODE_PRIORITY_MARK_MASK 0x00007FFF +#define MODE_PRIORITY_OFF (1 << 16) +#define MODE_PRIORITY_ALWAYS_ON (1 << 20) +#define MODE_PRIORITY_FORCE_MASK (1 << 24) +#define D1MODE_PRIORITY_B_CNT 0x654C +#define LB_MAX_REQ_OUTSTANDING 0x6D58 +#define LB_D1_MAX_REQ_OUTSTANDING_MASK 0x0000000F +#define LB_D1_MAX_REQ_OUTSTANDING_SHIFT 0 +#define LB_D2_MAX_REQ_OUTSTANDING_MASK 0x000F0000 +#define LB_D2_MAX_REQ_OUTSTANDING_SHIFT 16 +#define DCP_CONTROL 0x6C9C +#define D2MODE_PRIORITY_A_CNT 0x6D48 +#define D2MODE_PRIORITY_B_CNT 0x6D4C + +/* MC indirect registers */ +#define MC_STATUS_IDLE (1 << 0) +#define MC_MISC_CNTL 0x18 +#define DISABLE_GTW (1 << 1) +#define GART_INDEX_REG_EN (1 << 12) +#define BLOCK_GFX_D3_EN (1 << 14) +#define GART_FEATURE_ID 0x2B +#define HANG_EN (1 << 11) +#define TLB_ENABLE (1 << 18) +#define P2P_ENABLE (1 << 19) +#define GTW_LAC_EN (1 << 25) +#define LEVEL2_GART (0 << 30) +#define LEVEL1_GART (1 << 30) +#define PDC_EN (1 << 31) +#define GART_BASE 0x2C +#define GART_CACHE_CNTRL 0x2E +# define GART_CACHE_INVALIDATE (1 << 0) +#define MC_STATUS 0x90 +#define MCCFG_FB_LOCATION 0x100 +#define MC_FB_START_MASK 0x0000FFFF +#define MC_FB_START_SHIFT 0 +#define MC_FB_TOP_MASK 0xFFFF0000 +#define MC_FB_TOP_SHIFT 16 +#define MCCFG_AGP_LOCATION 0x101 +#define MC_AGP_START_MASK 0x0000FFFF +#define MC_AGP_START_SHIFT 0 +#define MC_AGP_TOP_MASK 0xFFFF0000 +#define MC_AGP_TOP_SHIFT 16 +#define MCCFG_AGP_BASE 0x102 +#define MCCFG_AGP_BASE_2 0x103 +#define MC_INIT_MISC_LAT_TIMER 0x104 +#define MC_DISP0R_INIT_LAT_SHIFT 8 +#define MC_DISP0R_INIT_LAT_MASK 0x00000F00 +#define MC_DISP1R_INIT_LAT_SHIFT 12 +#define MC_DISP1R_INIT_LAT_MASK 0x0000F000 + +#endif diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c index ffea37b..551e608 100644 --- a/drivers/gpu/drm/radeon/rv515.c +++ b/drivers/gpu/drm/radeon/rv515.c @@ -27,8 +27,9 @@ */ #include <linux/seq_file.h> #include "drmP.h" -#include "radeon_reg.h" +#include "rv515r.h" #include "radeon.h" +#include "radeon_share.h" /* rv515 depends on : */ void r100_hdp_reset(struct radeon_device *rdev); @@ -99,26 +100,26 @@ int rv515_mc_init(struct radeon_device *rdev) "programming pipes. Bad things might happen.\n"); } /* Write VRAM size in case we are limiting it */ - WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.vram_size); - tmp = REG_SET(RV515_MC_FB_START, rdev->mc.vram_location >> 16); + WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); + tmp = REG_SET(MC_FB_START, rdev->mc.vram_location >> 16); WREG32(0x134, tmp); - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; - tmp = REG_SET(RV515_MC_FB_TOP, tmp >> 16); - tmp |= REG_SET(RV515_MC_FB_START, rdev->mc.vram_location >> 16); - WREG32_MC(RV515_MC_FB_LOCATION, tmp); - WREG32(RS690_HDP_FB_LOCATION, rdev->mc.vram_location >> 16); + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; + tmp = REG_SET(MC_FB_TOP, tmp >> 16); + tmp |= REG_SET(MC_FB_START, rdev->mc.vram_location >> 16); + WREG32_MC(MC_FB_LOCATION, tmp); + WREG32(HDP_FB_LOCATION, rdev->mc.vram_location >> 16); WREG32(0x310, rdev->mc.vram_location); if (rdev->flags & RADEON_IS_AGP) { tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - 1; - tmp = REG_SET(RV515_MC_AGP_TOP, tmp >> 16); - tmp |= REG_SET(RV515_MC_AGP_START, rdev->mc.gtt_location >> 16); - WREG32_MC(RV515_MC_AGP_LOCATION, tmp); - WREG32_MC(RV515_MC_AGP_BASE, rdev->mc.agp_base); - WREG32_MC(RV515_MC_AGP_BASE_2, 0); + tmp = REG_SET(MC_AGP_TOP, tmp >> 16); + tmp |= REG_SET(MC_AGP_START, rdev->mc.gtt_location >> 16); + WREG32_MC(MC_AGP_LOCATION, tmp); + WREG32_MC(MC_AGP_BASE, rdev->mc.agp_base); + WREG32_MC(MC_AGP_BASE_2, 0); } else { - WREG32_MC(RV515_MC_AGP_LOCATION, 0x0FFFFFFF); - WREG32_MC(RV515_MC_AGP_BASE, 0); - WREG32_MC(RV515_MC_AGP_BASE_2, 0); + WREG32_MC(MC_AGP_LOCATION, 0x0FFFFFFF); + WREG32_MC(MC_AGP_BASE, 0); + WREG32_MC(MC_AGP_BASE_2, 0); } return 0; } @@ -136,95 +137,67 @@ void rv515_mc_fini(struct radeon_device *rdev) */ void rv515_ring_start(struct radeon_device *rdev) { - unsigned gb_tile_config; int r; - /* Sub pixel 1/12 so we can have 4K rendering according to doc */ - gb_tile_config = R300_ENABLE_TILING | R300_TILE_SIZE_16; - switch (rdev->num_gb_pipes) { - case 2: - gb_tile_config |= R300_PIPE_COUNT_R300; - break; - case 3: - gb_tile_config |= R300_PIPE_COUNT_R420_3P; - break; - case 4: - gb_tile_config |= R300_PIPE_COUNT_R420; - break; - case 1: - default: - gb_tile_config |= R300_PIPE_COUNT_RV350; - break; - } - r = radeon_ring_lock(rdev, 64); if (r) { return; } - radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0)); - radeon_ring_write(rdev, - RADEON_ISYNC_ANY2D_IDLE3D | - RADEON_ISYNC_ANY3D_IDLE2D | - RADEON_ISYNC_WAIT_IDLEGUI | - RADEON_ISYNC_CPSCRATCH_IDLEGUI); - radeon_ring_write(rdev, PACKET0(R300_GB_TILE_CONFIG, 0)); - radeon_ring_write(rdev, gb_tile_config); - radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0)); + radeon_ring_write(rdev, PACKET0(ISYNC_CNTL, 0)); radeon_ring_write(rdev, - RADEON_WAIT_2D_IDLECLEAN | - RADEON_WAIT_3D_IDLECLEAN); + ISYNC_ANY2D_IDLE3D | + ISYNC_ANY3D_IDLE2D | + ISYNC_WAIT_IDLEGUI | + ISYNC_CPSCRATCH_IDLEGUI); + radeon_ring_write(rdev, PACKET0(WAIT_UNTIL, 0)); + radeon_ring_write(rdev, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN); radeon_ring_write(rdev, PACKET0(0x170C, 0)); radeon_ring_write(rdev, 1 << 31); - radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0)); + radeon_ring_write(rdev, PACKET0(GB_SELECT, 0)); radeon_ring_write(rdev, 0); - radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0)); + radeon_ring_write(rdev, PACKET0(GB_ENABLE, 0)); radeon_ring_write(rdev, 0); radeon_ring_write(rdev, PACKET0(0x42C8, 0)); radeon_ring_write(rdev, (1 << rdev->num_gb_pipes) - 1); - radeon_ring_write(rdev, PACKET0(R500_VAP_INDEX_OFFSET, 0)); + radeon_ring_write(rdev, PACKET0(VAP_INDEX_OFFSET, 0)); radeon_ring_write(rdev, 0); - radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); - radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE); - radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0)); - radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE); - radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0)); - radeon_ring_write(rdev, - RADEON_WAIT_2D_IDLECLEAN | - RADEON_WAIT_3D_IDLECLEAN); - radeon_ring_write(rdev, PACKET0(R300_GB_AA_CONFIG, 0)); + radeon_ring_write(rdev, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0)); + radeon_ring_write(rdev, RB3D_DC_FLUSH | RB3D_DC_FREE); + radeon_ring_write(rdev, PACKET0(ZB_ZCACHE_CTLSTAT, 0)); + radeon_ring_write(rdev, ZC_FLUSH | ZC_FREE); + radeon_ring_write(rdev, PACKET0(WAIT_UNTIL, 0)); + radeon_ring_write(rdev, WAIT_2D_IDLECLEAN | WAIT_3D_IDLECLEAN); + radeon_ring_write(rdev, PACKET0(GB_AA_CONFIG, 0)); radeon_ring_write(rdev, 0); - radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0)); - radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE); - radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0)); - radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE); - radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS0, 0)); - radeon_ring_write(rdev, - ((6 << R300_MS_X0_SHIFT) | - (6 << R300_MS_Y0_SHIFT) | - (6 << R300_MS_X1_SHIFT) | - (6 << R300_MS_Y1_SHIFT) | - (6 << R300_MS_X2_SHIFT) | - (6 << R300_MS_Y2_SHIFT) | - (6 << R300_MSBD0_Y_SHIFT) | - (6 << R300_MSBD0_X_SHIFT))); - radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS1, 0)); + radeon_ring_write(rdev, PACKET0(RB3D_DSTCACHE_CTLSTAT, 0)); + radeon_ring_write(rdev, RB3D_DC_FLUSH | RB3D_DC_FREE); + radeon_ring_write(rdev, PACKET0(ZB_ZCACHE_CTLSTAT, 0)); + radeon_ring_write(rdev, ZC_FLUSH | ZC_FREE); + radeon_ring_write(rdev, PACKET0(GB_MSPOS0, 0)); radeon_ring_write(rdev, - ((6 << R300_MS_X3_SHIFT) | - (6 << R300_MS_Y3_SHIFT) | - (6 << R300_MS_X4_SHIFT) | - (6 << R300_MS_Y4_SHIFT) | - (6 << R300_MS_X5_SHIFT) | - (6 << R300_MS_Y5_SHIFT) | - (6 << R300_MSBD1_SHIFT))); - radeon_ring_write(rdev, PACKET0(R300_GA_ENHANCE, 0)); - radeon_ring_write(rdev, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL); - radeon_ring_write(rdev, PACKET0(R300_GA_POLY_MODE, 0)); + ((6 << MS_X0_SHIFT) | + (6 << MS_Y0_SHIFT) | + (6 << MS_X1_SHIFT) | + (6 << MS_Y1_SHIFT) | + (6 << MS_X2_SHIFT) | + (6 << MS_Y2_SHIFT) | + (6 << MSBD0_Y_SHIFT) | + (6 << MSBD0_X_SHIFT))); + radeon_ring_write(rdev, PACKET0(GB_MSPOS1, 0)); radeon_ring_write(rdev, - R300_FRONT_PTYPE_TRIANGE | R300_BACK_PTYPE_TRIANGE); - radeon_ring_write(rdev, PACKET0(R300_GA_ROUND_MODE, 0)); - radeon_ring_write(rdev, - R300_GEOMETRY_ROUND_NEAREST | - R300_COLOR_ROUND_NEAREST); + ((6 << MS_X3_SHIFT) | + (6 << MS_Y3_SHIFT) | + (6 << MS_X4_SHIFT) | + (6 << MS_Y4_SHIFT) | + (6 << MS_X5_SHIFT) | + (6 << MS_Y5_SHIFT) | + (6 << MSBD1_SHIFT))); + radeon_ring_write(rdev, PACKET0(GA_ENHANCE, 0)); + radeon_ring_write(rdev, GA_DEADLOCK_CNTL | GA_FASTSYNC_CNTL); + radeon_ring_write(rdev, PACKET0(GA_POLY_MODE, 0)); + radeon_ring_write(rdev, FRONT_PTYPE_TRIANGE | BACK_PTYPE_TRIANGE); + radeon_ring_write(rdev, PACKET0(GA_ROUND_MODE, 0)); + radeon_ring_write(rdev, GEOMETRY_ROUND_NEAREST | COLOR_ROUND_NEAREST); radeon_ring_write(rdev, PACKET0(0x20C8, 0)); radeon_ring_write(rdev, 0); radeon_ring_unlock_commit(rdev); @@ -242,8 +215,8 @@ int rv515_mc_wait_for_idle(struct radeon_device *rdev) for (i = 0; i < rdev->usec_timeout; i++) { /* read MC_STATUS */ - tmp = RREG32_MC(RV515_MC_STATUS); - if (tmp & RV515_MC_STATUS_IDLE) { + tmp = RREG32_MC(MC_STATUS); + if (tmp & MC_STATUS_IDLE) { return 0; } DRM_UDELAY(1); @@ -291,33 +264,33 @@ int rv515_ga_reset(struct radeon_device *rdev) reinit_cp = rdev->cp.ready; rdev->cp.ready = false; for (i = 0; i < rdev->usec_timeout; i++) { - WREG32(RADEON_CP_CSQ_MODE, 0); - WREG32(RADEON_CP_CSQ_CNTL, 0); - WREG32(RADEON_RBBM_SOFT_RESET, 0x32005); - (void)RREG32(RADEON_RBBM_SOFT_RESET); + WREG32(CP_CSQ_MODE, 0); + WREG32(CP_CSQ_CNTL, 0); + WREG32(RBBM_SOFT_RESET, 0x32005); + (void)RREG32(RBBM_SOFT_RESET); udelay(200); - WREG32(RADEON_RBBM_SOFT_RESET, 0); + WREG32(RBBM_SOFT_RESET, 0); /* Wait to prevent race in RBBM_STATUS */ mdelay(1); - tmp = RREG32(RADEON_RBBM_STATUS); + tmp = RREG32(RBBM_STATUS); if (tmp & ((1 << 20) | (1 << 26))) { DRM_ERROR("VAP & CP still busy (RBBM_STATUS=0x%08X)\n", tmp); /* GA still busy soft reset it */ WREG32(0x429C, 0x200); - WREG32(R300_VAP_PVS_STATE_FLUSH_REG, 0); + WREG32(VAP_PVS_STATE_FLUSH_REG, 0); WREG32(0x43E0, 0); WREG32(0x43E4, 0); WREG32(0x24AC, 0); } /* Wait to prevent race in RBBM_STATUS */ mdelay(1); - tmp = RREG32(RADEON_RBBM_STATUS); + tmp = RREG32(RBBM_STATUS); if (!(tmp & ((1 << 20) | (1 << 26)))) { break; } } for (i = 0; i < rdev->usec_timeout; i++) { - tmp = RREG32(RADEON_RBBM_STATUS); + tmp = RREG32(RBBM_STATUS); if (!(tmp & ((1 << 20) | (1 << 26)))) { DRM_INFO("GA reset succeed (RBBM_STATUS=0x%08X)\n", tmp); @@ -331,7 +304,7 @@ int rv515_ga_reset(struct radeon_device *rdev) } DRM_UDELAY(1); } - tmp = RREG32(RADEON_RBBM_STATUS); + tmp = RREG32(RBBM_STATUS); DRM_ERROR("Failed to reset GA ! (RBBM_STATUS=0x%08X)\n", tmp); return -1; } @@ -341,7 +314,7 @@ int rv515_gpu_reset(struct radeon_device *rdev) uint32_t status; /* reset order likely matter */ - status = RREG32(RADEON_RBBM_STATUS); + status = RREG32(RBBM_STATUS); /* reset HDP */ r100_hdp_reset(rdev); /* reset rb2d */ @@ -353,12 +326,12 @@ int rv515_gpu_reset(struct radeon_device *rdev) rv515_ga_reset(rdev); } /* reset CP */ - status = RREG32(RADEON_RBBM_STATUS); + status = RREG32(RBBM_STATUS); if (status & (1 << 16)) { r100_cp_reset(rdev); } /* Check if GPU is idle */ - status = RREG32(RADEON_RBBM_STATUS); + status = RREG32(RBBM_STATUS); if (status & (1 << 31)) { DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status); return -1; @@ -377,8 +350,7 @@ static void rv515_vram_get_type(struct radeon_device *rdev) rdev->mc.vram_width = 128; rdev->mc.vram_is_ddr = true; - tmp = RREG32_MC(RV515_MC_CNTL); - tmp &= RV515_MEM_NUM_CHANNELS_MASK; + tmp = RREG32_MC(RV515_MC_CNTL) & MEM_NUM_CHANNELS_MASK; switch (tmp) { case 0: rdev->mc.vram_width = 64; @@ -394,11 +366,16 @@ static void rv515_vram_get_type(struct radeon_device *rdev) void rv515_vram_info(struct radeon_device *rdev) { + fixed20_12 a; + rv515_vram_get_type(rdev); - rdev->mc.vram_size = RREG32(RADEON_CONFIG_MEMSIZE); - rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); - rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + /* FIXME: we should enforce default clock in case GPU is not in + * default setup + */ + a.full = rfixed_const(100); + rdev->pm.sclk.full = rfixed_const(rdev->clock.default_sclk); + rdev->pm.sclk.full = rfixed_div(rdev->pm.sclk, a); } @@ -409,35 +386,35 @@ uint32_t rv515_mc_rreg(struct radeon_device *rdev, uint32_t reg) { uint32_t r; - WREG32(R520_MC_IND_INDEX, 0x7f0000 | (reg & 0xffff)); - r = RREG32(R520_MC_IND_DATA); - WREG32(R520_MC_IND_INDEX, 0); + WREG32(MC_IND_INDEX, 0x7f0000 | (reg & 0xffff)); + r = RREG32(MC_IND_DATA); + WREG32(MC_IND_INDEX, 0); return r; } void rv515_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) { - WREG32(R520_MC_IND_INDEX, 0xff0000 | ((reg) & 0xffff)); - WREG32(R520_MC_IND_DATA, (v)); - WREG32(R520_MC_IND_INDEX, 0); + WREG32(MC_IND_INDEX, 0xff0000 | ((reg) & 0xffff)); + WREG32(MC_IND_DATA, (v)); + WREG32(MC_IND_INDEX, 0); } uint32_t rv515_pcie_rreg(struct radeon_device *rdev, uint32_t reg) { uint32_t r; - WREG32(RADEON_PCIE_INDEX, ((reg) & 0x7ff)); - (void)RREG32(RADEON_PCIE_INDEX); - r = RREG32(RADEON_PCIE_DATA); + WREG32(PCIE_INDEX, ((reg) & 0x7ff)); + (void)RREG32(PCIE_INDEX); + r = RREG32(PCIE_DATA); return r; } void rv515_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) { - WREG32(RADEON_PCIE_INDEX, ((reg) & 0x7ff)); - (void)RREG32(RADEON_PCIE_INDEX); - WREG32(RADEON_PCIE_DATA, (v)); - (void)RREG32(RADEON_PCIE_DATA); + WREG32(PCIE_INDEX, ((reg) & 0x7ff)); + (void)RREG32(PCIE_INDEX); + WREG32(PCIE_DATA, (v)); + (void)RREG32(PCIE_DATA); } @@ -452,13 +429,13 @@ static int rv515_debugfs_pipes_info(struct seq_file *m, void *data) struct radeon_device *rdev = dev->dev_private; uint32_t tmp; - tmp = RREG32(R400_GB_PIPE_SELECT); + tmp = RREG32(GB_PIPE_SELECT); seq_printf(m, "GB_PIPE_SELECT 0x%08x\n", tmp); - tmp = RREG32(R500_SU_REG_DEST); + tmp = RREG32(SU_REG_DEST); seq_printf(m, "SU_REG_DEST 0x%08x\n", tmp); - tmp = RREG32(R300_GB_TILE_CONFIG); + tmp = RREG32(GB_TILE_CONFIG); seq_printf(m, "GB_TILE_CONFIG 0x%08x\n", tmp); - tmp = RREG32(R300_DST_PIPE_CONFIG); + tmp = RREG32(DST_PIPE_CONFIG); seq_printf(m, "DST_PIPE_CONFIG 0x%08x\n", tmp); return 0; } @@ -509,9 +486,9 @@ int rv515_debugfs_ga_info_init(struct radeon_device *rdev) /* * Asic initialization */ -static const unsigned r500_reg_safe_bm[159] = { +static const unsigned r500_reg_safe_bm[219] = { + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, - 0xFFFFFFBF, 0xFFFFFFFF, 0xFFFFFFBF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, @@ -549,14 +526,575 @@ static const unsigned r500_reg_safe_bm[159] = { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFF80FFFF, 0x00000000, 0x00000000, 0x00000000, 0x00000000, - 0x0003FC01, 0x3FFFFCF8, 0xFE800B19, + 0x0003FC01, 0x3FFFFCF8, 0xFE800B19, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, }; - - int rv515_init(struct radeon_device *rdev) { rdev->config.r300.reg_safe_bm = r500_reg_safe_bm; rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(r500_reg_safe_bm); return 0; } + +void atom_rv515_force_tv_scaler(struct radeon_device *rdev) +{ + + WREG32(0x659C, 0x0); + WREG32(0x6594, 0x705); + WREG32(0x65A4, 0x10001); + WREG32(0x65D8, 0x0); + WREG32(0x65B0, 0x0); + WREG32(0x65C0, 0x0); + WREG32(0x65D4, 0x0); + WREG32(0x6578, 0x0); + WREG32(0x657C, 0x841880A8); + WREG32(0x6578, 0x1); + WREG32(0x657C, 0x84208680); + WREG32(0x6578, 0x2); + WREG32(0x657C, 0xBFF880B0); + WREG32(0x6578, 0x100); + WREG32(0x657C, 0x83D88088); + WREG32(0x6578, 0x101); + WREG32(0x657C, 0x84608680); + WREG32(0x6578, 0x102); + WREG32(0x657C, 0xBFF080D0); + WREG32(0x6578, 0x200); + WREG32(0x657C, 0x83988068); + WREG32(0x6578, 0x201); + WREG32(0x657C, 0x84A08680); + WREG32(0x6578, 0x202); + WREG32(0x657C, 0xBFF080F8); + WREG32(0x6578, 0x300); + WREG32(0x657C, 0x83588058); + WREG32(0x6578, 0x301); + WREG32(0x657C, 0x84E08660); + WREG32(0x6578, 0x302); + WREG32(0x657C, 0xBFF88120); + WREG32(0x6578, 0x400); + WREG32(0x657C, 0x83188040); + WREG32(0x6578, 0x401); + WREG32(0x657C, 0x85008660); + WREG32(0x6578, 0x402); + WREG32(0x657C, 0xBFF88150); + WREG32(0x6578, 0x500); + WREG32(0x657C, 0x82D88030); + WREG32(0x6578, 0x501); + WREG32(0x657C, 0x85408640); + WREG32(0x6578, 0x502); + WREG32(0x657C, 0xBFF88180); + WREG32(0x6578, 0x600); + WREG32(0x657C, 0x82A08018); + WREG32(0x6578, 0x601); + WREG32(0x657C, 0x85808620); + WREG32(0x6578, 0x602); + WREG32(0x657C, 0xBFF081B8); + WREG32(0x6578, 0x700); + WREG32(0x657C, 0x82608010); + WREG32(0x6578, 0x701); + WREG32(0x657C, 0x85A08600); + WREG32(0x6578, 0x702); + WREG32(0x657C, 0x800081F0); + WREG32(0x6578, 0x800); + WREG32(0x657C, 0x8228BFF8); + WREG32(0x6578, 0x801); + WREG32(0x657C, 0x85E085E0); + WREG32(0x6578, 0x802); + WREG32(0x657C, 0xBFF88228); + WREG32(0x6578, 0x10000); + WREG32(0x657C, 0x82A8BF00); + WREG32(0x6578, 0x10001); + WREG32(0x657C, 0x82A08CC0); + WREG32(0x6578, 0x10002); + WREG32(0x657C, 0x8008BEF8); + WREG32(0x6578, 0x10100); + WREG32(0x657C, 0x81F0BF28); + WREG32(0x6578, 0x10101); + WREG32(0x657C, 0x83608CA0); + WREG32(0x6578, 0x10102); + WREG32(0x657C, 0x8018BED0); + WREG32(0x6578, 0x10200); + WREG32(0x657C, 0x8148BF38); + WREG32(0x6578, 0x10201); + WREG32(0x657C, 0x84408C80); + WREG32(0x6578, 0x10202); + WREG32(0x657C, 0x8008BEB8); + WREG32(0x6578, 0x10300); + WREG32(0x657C, 0x80B0BF78); + WREG32(0x6578, 0x10301); + WREG32(0x657C, 0x85008C20); + WREG32(0x6578, 0x10302); + WREG32(0x657C, 0x8020BEA0); + WREG32(0x6578, 0x10400); + WREG32(0x657C, 0x8028BF90); + WREG32(0x6578, 0x10401); + WREG32(0x657C, 0x85E08BC0); + WREG32(0x6578, 0x10402); + WREG32(0x657C, 0x8018BE90); + WREG32(0x6578, 0x10500); + WREG32(0x657C, 0xBFB8BFB0); + WREG32(0x6578, 0x10501); + WREG32(0x657C, 0x86C08B40); + WREG32(0x6578, 0x10502); + WREG32(0x657C, 0x8010BE90); + WREG32(0x6578, 0x10600); + WREG32(0x657C, 0xBF58BFC8); + WREG32(0x6578, 0x10601); + WREG32(0x657C, 0x87A08AA0); + WREG32(0x6578, 0x10602); + WREG32(0x657C, 0x8010BE98); + WREG32(0x6578, 0x10700); + WREG32(0x657C, 0xBF10BFF0); + WREG32(0x6578, 0x10701); + WREG32(0x657C, 0x886089E0); + WREG32(0x6578, 0x10702); + WREG32(0x657C, 0x8018BEB0); + WREG32(0x6578, 0x10800); + WREG32(0x657C, 0xBED8BFE8); + WREG32(0x6578, 0x10801); + WREG32(0x657C, 0x89408940); + WREG32(0x6578, 0x10802); + WREG32(0x657C, 0xBFE8BED8); + WREG32(0x6578, 0x20000); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20001); + WREG32(0x657C, 0x90008000); + WREG32(0x6578, 0x20002); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20003); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20100); + WREG32(0x657C, 0x80108000); + WREG32(0x6578, 0x20101); + WREG32(0x657C, 0x8FE0BF70); + WREG32(0x6578, 0x20102); + WREG32(0x657C, 0xBFE880C0); + WREG32(0x6578, 0x20103); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20200); + WREG32(0x657C, 0x8018BFF8); + WREG32(0x6578, 0x20201); + WREG32(0x657C, 0x8F80BF08); + WREG32(0x6578, 0x20202); + WREG32(0x657C, 0xBFD081A0); + WREG32(0x6578, 0x20203); + WREG32(0x657C, 0xBFF88000); + WREG32(0x6578, 0x20300); + WREG32(0x657C, 0x80188000); + WREG32(0x6578, 0x20301); + WREG32(0x657C, 0x8EE0BEC0); + WREG32(0x6578, 0x20302); + WREG32(0x657C, 0xBFB082A0); + WREG32(0x6578, 0x20303); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20400); + WREG32(0x657C, 0x80188000); + WREG32(0x6578, 0x20401); + WREG32(0x657C, 0x8E00BEA0); + WREG32(0x6578, 0x20402); + WREG32(0x657C, 0xBF8883C0); + WREG32(0x6578, 0x20403); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x20500); + WREG32(0x657C, 0x80188000); + WREG32(0x6578, 0x20501); + WREG32(0x657C, 0x8D00BE90); + WREG32(0x6578, 0x20502); + WREG32(0x657C, 0xBF588500); + WREG32(0x6578, 0x20503); + WREG32(0x657C, 0x80008008); + WREG32(0x6578, 0x20600); + WREG32(0x657C, 0x80188000); + WREG32(0x6578, 0x20601); + WREG32(0x657C, 0x8BC0BE98); + WREG32(0x6578, 0x20602); + WREG32(0x657C, 0xBF308660); + WREG32(0x6578, 0x20603); + WREG32(0x657C, 0x80008008); + WREG32(0x6578, 0x20700); + WREG32(0x657C, 0x80108000); + WREG32(0x6578, 0x20701); + WREG32(0x657C, 0x8A80BEB0); + WREG32(0x6578, 0x20702); + WREG32(0x657C, 0xBF0087C0); + WREG32(0x6578, 0x20703); + WREG32(0x657C, 0x80008008); + WREG32(0x6578, 0x20800); + WREG32(0x657C, 0x80108000); + WREG32(0x6578, 0x20801); + WREG32(0x657C, 0x8920BED0); + WREG32(0x6578, 0x20802); + WREG32(0x657C, 0xBED08920); + WREG32(0x6578, 0x20803); + WREG32(0x657C, 0x80008010); + WREG32(0x6578, 0x30000); + WREG32(0x657C, 0x90008000); + WREG32(0x6578, 0x30001); + WREG32(0x657C, 0x80008000); + WREG32(0x6578, 0x30100); + WREG32(0x657C, 0x8FE0BF90); + WREG32(0x6578, 0x30101); + WREG32(0x657C, 0xBFF880A0); + WREG32(0x6578, 0x30200); + WREG32(0x657C, 0x8F60BF40); + WREG32(0x6578, 0x30201); + WREG32(0x657C, 0xBFE88180); + WREG32(0x6578, 0x30300); + WREG32(0x657C, 0x8EC0BF00); + WREG32(0x6578, 0x30301); + WREG32(0x657C, 0xBFC88280); + WREG32(0x6578, 0x30400); + WREG32(0x657C, 0x8DE0BEE0); + WREG32(0x6578, 0x30401); + WREG32(0x657C, 0xBFA083A0); + WREG32(0x6578, 0x30500); + WREG32(0x657C, 0x8CE0BED0); + WREG32(0x6578, 0x30501); + WREG32(0x657C, 0xBF7884E0); + WREG32(0x6578, 0x30600); + WREG32(0x657C, 0x8BA0BED8); + WREG32(0x6578, 0x30601); + WREG32(0x657C, 0xBF508640); + WREG32(0x6578, 0x30700); + WREG32(0x657C, 0x8A60BEE8); + WREG32(0x6578, 0x30701); + WREG32(0x657C, 0xBF2087A0); + WREG32(0x6578, 0x30800); + WREG32(0x657C, 0x8900BF00); + WREG32(0x6578, 0x30801); + WREG32(0x657C, 0xBF008900); +} + +struct rv515_watermark { + u32 lb_request_fifo_depth; + fixed20_12 num_line_pair; + fixed20_12 estimated_width; + fixed20_12 worst_case_latency; + fixed20_12 consumption_rate; + fixed20_12 active_time; + fixed20_12 dbpp; + fixed20_12 priority_mark_max; + fixed20_12 priority_mark; + fixed20_12 sclk; +}; + +void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, + struct radeon_crtc *crtc, + struct rv515_watermark *wm) +{ + struct drm_display_mode *mode = &crtc->base.mode; + fixed20_12 a, b, c; + fixed20_12 pclk, request_fifo_depth, tolerable_latency, estimated_width; + fixed20_12 consumption_time, line_time, chunk_time, read_delay_latency; + + if (!crtc->base.enabled) { + /* FIXME: wouldn't it better to set priority mark to maximum */ + wm->lb_request_fifo_depth = 4; + return; + } + + if (crtc->vsc.full > rfixed_const(2)) + wm->num_line_pair.full = rfixed_const(2); + else + wm->num_line_pair.full = rfixed_const(1); + + b.full = rfixed_const(mode->crtc_hdisplay); + c.full = rfixed_const(256); + a.full = rfixed_mul(wm->num_line_pair, b); + request_fifo_depth.full = rfixed_div(a, c); + if (a.full < rfixed_const(4)) { + wm->lb_request_fifo_depth = 4; + } else { + wm->lb_request_fifo_depth = rfixed_trunc(request_fifo_depth); + } + + /* Determine consumption rate + * pclk = pixel clock period(ns) = 1000 / (mode.clock / 1000) + * vtaps = number of vertical taps, + * vsc = vertical scaling ratio, defined as source/destination + * hsc = horizontal scaling ration, defined as source/destination + */ + a.full = rfixed_const(mode->clock); + b.full = rfixed_const(1000); + a.full = rfixed_div(a, b); + pclk.full = rfixed_div(b, a); + if (crtc->rmx_type != RMX_OFF) { + b.full = rfixed_const(2); + if (crtc->vsc.full > b.full) + b.full = crtc->vsc.full; + b.full = rfixed_mul(b, crtc->hsc); + c.full = rfixed_const(2); + b.full = rfixed_div(b, c); + consumption_time.full = rfixed_div(pclk, b); + } else { + consumption_time.full = pclk.full; + } + a.full = rfixed_const(1); + wm->consumption_rate.full = rfixed_div(a, consumption_time); + + + /* Determine line time + * LineTime = total time for one line of displayhtotal + * LineTime = total number of horizontal pixels + * pclk = pixel clock period(ns) + */ + a.full = rfixed_const(crtc->base.mode.crtc_htotal); + line_time.full = rfixed_mul(a, pclk); + + /* Determine active time + * ActiveTime = time of active region of display within one line, + * hactive = total number of horizontal active pixels + * htotal = total number of horizontal pixels + */ + a.full = rfixed_const(crtc->base.mode.crtc_htotal); + b.full = rfixed_const(crtc->base.mode.crtc_hdisplay); + wm->active_time.full = rfixed_mul(line_time, b); + wm->active_time.full = rfixed_div(wm->active_time, a); + + /* Determine chunk time + * ChunkTime = the time it takes the DCP to send one chunk of data + * to the LB which consists of pipeline delay and inter chunk gap + * sclk = system clock(Mhz) + */ + a.full = rfixed_const(600 * 1000); + chunk_time.full = rfixed_div(a, rdev->pm.sclk); + read_delay_latency.full = rfixed_const(1000); + + /* Determine the worst case latency + * NumLinePair = Number of line pairs to request(1=2 lines, 2=4 lines) + * WorstCaseLatency = worst case time from urgent to when the MC starts + * to return data + * READ_DELAY_IDLE_MAX = constant of 1us + * ChunkTime = time it takes the DCP to send one chunk of data to the LB + * which consists of pipeline delay and inter chunk gap + */ + if (rfixed_trunc(wm->num_line_pair) > 1) { + a.full = rfixed_const(3); + wm->worst_case_latency.full = rfixed_mul(a, chunk_time); + wm->worst_case_latency.full += read_delay_latency.full; + } else { + wm->worst_case_latency.full = chunk_time.full + read_delay_latency.full; + } + + /* Determine the tolerable latency + * TolerableLatency = Any given request has only 1 line time + * for the data to be returned + * LBRequestFifoDepth = Number of chunk requests the LB can + * put into the request FIFO for a display + * LineTime = total time for one line of display + * ChunkTime = the time it takes the DCP to send one chunk + * of data to the LB which consists of + * pipeline delay and inter chunk gap + */ + if ((2+wm->lb_request_fifo_depth) >= rfixed_trunc(request_fifo_depth)) { + tolerable_latency.full = line_time.full; + } else { + tolerable_latency.full = rfixed_const(wm->lb_request_fifo_depth - 2); + tolerable_latency.full = request_fifo_depth.full - tolerable_latency.full; + tolerable_latency.full = rfixed_mul(tolerable_latency, chunk_time); + tolerable_latency.full = line_time.full - tolerable_latency.full; + } + /* We assume worst case 32bits (4 bytes) */ + wm->dbpp.full = rfixed_const(2 * 16); + + /* Determine the maximum priority mark + * width = viewport width in pixels + */ + a.full = rfixed_const(16); + wm->priority_mark_max.full = rfixed_const(crtc->base.mode.crtc_hdisplay); + wm->priority_mark_max.full = rfixed_div(wm->priority_mark_max, a); + + /* Determine estimated width */ + estimated_width.full = tolerable_latency.full - wm->worst_case_latency.full; + estimated_width.full = rfixed_div(estimated_width, consumption_time); + if (rfixed_trunc(estimated_width) > crtc->base.mode.crtc_hdisplay) { + wm->priority_mark.full = rfixed_const(10); + } else { + a.full = rfixed_const(16); + wm->priority_mark.full = rfixed_div(estimated_width, a); + wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full; + } +} + +void rv515_bandwidth_avivo_update(struct radeon_device *rdev) +{ + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + struct rv515_watermark wm0; + struct rv515_watermark wm1; + u32 tmp; + fixed20_12 priority_mark02, priority_mark12, fill_rate; + fixed20_12 a, b; + + if (rdev->mode_info.crtcs[0]->base.enabled) + mode0 = &rdev->mode_info.crtcs[0]->base.mode; + if (rdev->mode_info.crtcs[1]->base.enabled) + mode1 = &rdev->mode_info.crtcs[1]->base.mode; + rs690_line_buffer_adjust(rdev, mode0, mode1); + + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0); + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1); + + tmp = wm0.lb_request_fifo_depth; + tmp |= wm1.lb_request_fifo_depth << 16; + WREG32(LB_MAX_REQ_OUTSTANDING, tmp); + + if (mode0 && mode1) { + if (rfixed_trunc(wm0.dbpp) > 64) + a.full = rfixed_div(wm0.dbpp, wm0.num_line_pair); + else + a.full = wm0.num_line_pair.full; + if (rfixed_trunc(wm1.dbpp) > 64) + b.full = rfixed_div(wm1.dbpp, wm1.num_line_pair); + else + b.full = wm1.num_line_pair.full; + a.full += b.full; + fill_rate.full = rfixed_div(wm0.sclk, a); + if (wm0.consumption_rate.full > fill_rate.full) { + b.full = wm0.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm0.active_time); + a.full = rfixed_const(16); + b.full = rfixed_div(b, a); + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + priority_mark02.full = a.full + b.full; + } else { + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark02.full = rfixed_div(a, b); + } + if (wm1.consumption_rate.full > fill_rate.full) { + b.full = wm1.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm1.active_time); + a.full = rfixed_const(16); + b.full = rfixed_div(b, a); + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + priority_mark12.full = a.full + b.full; + } else { + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } + if (wm0.priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark.full; + if (rfixed_trunc(priority_mark02) < 0) + priority_mark02.full = 0; + if (wm0.priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark_max.full; + if (wm1.priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark.full; + if (rfixed_trunc(priority_mark12) < 0) + priority_mark12.full = 0; + if (wm1.priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02)); + WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02)); + WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12)); + WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12)); + } else if (mode0) { + if (rfixed_trunc(wm0.dbpp) > 64) + a.full = rfixed_div(wm0.dbpp, wm0.num_line_pair); + else + a.full = wm0.num_line_pair.full; + fill_rate.full = rfixed_div(wm0.sclk, a); + if (wm0.consumption_rate.full > fill_rate.full) { + b.full = wm0.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm0.active_time); + a.full = rfixed_const(16); + b.full = rfixed_div(b, a); + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + priority_mark02.full = a.full + b.full; + } else { + a.full = rfixed_mul(wm0.worst_case_latency, + wm0.consumption_rate); + b.full = rfixed_const(16); + priority_mark02.full = rfixed_div(a, b); + } + if (wm0.priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark.full; + if (rfixed_trunc(priority_mark02) < 0) + priority_mark02.full = 0; + if (wm0.priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark02)); + WREG32(D1MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark02)); + WREG32(D2MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF); + WREG32(D2MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF); + } else { + if (rfixed_trunc(wm1.dbpp) > 64) + a.full = rfixed_div(wm1.dbpp, wm1.num_line_pair); + else + a.full = wm1.num_line_pair.full; + fill_rate.full = rfixed_div(wm1.sclk, a); + if (wm1.consumption_rate.full > fill_rate.full) { + b.full = wm1.consumption_rate.full - fill_rate.full; + b.full = rfixed_mul(b, wm1.active_time); + a.full = rfixed_const(16); + b.full = rfixed_div(b, a); + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + priority_mark12.full = a.full + b.full; + } else { + a.full = rfixed_mul(wm1.worst_case_latency, + wm1.consumption_rate); + b.full = rfixed_const(16 * 1000); + priority_mark12.full = rfixed_div(a, b); + } + if (wm1.priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark.full; + if (rfixed_trunc(priority_mark12) < 0) + priority_mark12.full = 0; + if (wm1.priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1.priority_mark_max.full; + WREG32(D1MODE_PRIORITY_A_CNT, MODE_PRIORITY_OFF); + WREG32(D1MODE_PRIORITY_B_CNT, MODE_PRIORITY_OFF); + WREG32(D2MODE_PRIORITY_A_CNT, rfixed_trunc(priority_mark12)); + WREG32(D2MODE_PRIORITY_B_CNT, rfixed_trunc(priority_mark12)); + } +} + +void rv515_bandwidth_update(struct radeon_device *rdev) +{ + uint32_t tmp; + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + + if (rdev->mode_info.crtcs[0]->base.enabled) + mode0 = &rdev->mode_info.crtcs[0]->base.mode; + if (rdev->mode_info.crtcs[1]->base.enabled) + mode1 = &rdev->mode_info.crtcs[1]->base.mode; + /* + * Set display0/1 priority up in the memory controller for + * modes if the user specifies HIGH for displaypriority + * option. + */ + if (rdev->disp_priority == 2) { + tmp = RREG32_MC(MC_MISC_LAT_TIMER); + tmp &= ~MC_DISP1R_INIT_LAT_MASK; + tmp &= ~MC_DISP0R_INIT_LAT_MASK; + if (mode1) + tmp |= (1 << MC_DISP1R_INIT_LAT_SHIFT); + if (mode0) + tmp |= (1 << MC_DISP0R_INIT_LAT_SHIFT); + WREG32_MC(MC_MISC_LAT_TIMER, tmp); + } + rv515_bandwidth_avivo_update(rdev); +} diff --git a/drivers/gpu/drm/radeon/rv515r.h b/drivers/gpu/drm/radeon/rv515r.h new file mode 100644 index 0000000..f3cf840 --- /dev/null +++ b/drivers/gpu/drm/radeon/rv515r.h @@ -0,0 +1,170 @@ +/* + * Copyright 2008 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 + */ +#ifndef RV515R_H +#define RV515R_H + +/* RV515 registers */ +#define PCIE_INDEX 0x0030 +#define PCIE_DATA 0x0034 +#define MC_IND_INDEX 0x0070 +#define MC_IND_WR_EN (1 << 24) +#define MC_IND_DATA 0x0074 +#define RBBM_SOFT_RESET 0x00F0 +#define CONFIG_MEMSIZE 0x00F8 +#define HDP_FB_LOCATION 0x0134 +#define CP_CSQ_CNTL 0x0740 +#define CP_CSQ_MODE 0x0744 +#define CP_CSQ_ADDR 0x07F0 +#define CP_CSQ_DATA 0x07F4 +#define CP_CSQ_STAT 0x07F8 +#define CP_CSQ2_STAT 0x07FC +#define RBBM_STATUS 0x0E40 +#define DST_PIPE_CONFIG 0x170C +#define WAIT_UNTIL 0x1720 +#define WAIT_2D_IDLE (1 << 14) +#define WAIT_3D_IDLE (1 << 15) +#define WAIT_2D_IDLECLEAN (1 << 16) +#define WAIT_3D_IDLECLEAN (1 << 17) +#define ISYNC_CNTL 0x1724 +#define ISYNC_ANY2D_IDLE3D (1 << 0) +#define ISYNC_ANY3D_IDLE2D (1 << 1) +#define ISYNC_TRIG2D_IDLE3D (1 << 2) +#define ISYNC_TRIG3D_IDLE2D (1 << 3) +#define ISYNC_WAIT_IDLEGUI (1 << 4) +#define ISYNC_CPSCRATCH_IDLEGUI (1 << 5) +#define VAP_INDEX_OFFSET 0x208C +#define VAP_PVS_STATE_FLUSH_REG 0x2284 +#define GB_ENABLE 0x4008 +#define GB_MSPOS0 0x4010 +#define MS_X0_SHIFT 0 +#define MS_Y0_SHIFT 4 +#define MS_X1_SHIFT 8 +#define MS_Y1_SHIFT 12 +#define MS_X2_SHIFT 16 +#define MS_Y2_SHIFT 20 +#define MSBD0_Y_SHIFT 24 +#define MSBD0_X_SHIFT 28 +#define GB_MSPOS1 0x4014 +#define MS_X3_SHIFT 0 +#define MS_Y3_SHIFT 4 +#define MS_X4_SHIFT 8 +#define MS_Y4_SHIFT 12 +#define MS_X5_SHIFT 16 +#define MS_Y5_SHIFT 20 +#define MSBD1_SHIFT 24 +#define GB_TILE_CONFIG 0x4018 +#define ENABLE_TILING (1 << 0) +#define PIPE_COUNT_MASK 0x0000000E +#define PIPE_COUNT_SHIFT 1 +#define TILE_SIZE_8 (0 << 4) +#define TILE_SIZE_16 (1 << 4) +#define TILE_SIZE_32 (2 << 4) +#define SUBPIXEL_1_12 (0 << 16) +#define SUBPIXEL_1_16 (1 << 16) +#define GB_SELECT 0x401C +#define GB_AA_CONFIG 0x4020 +#define GB_PIPE_SELECT 0x402C +#define GA_ENHANCE 0x4274 +#define GA_DEADLOCK_CNTL (1 << 0) +#define GA_FASTSYNC_CNTL (1 << 1) +#define GA_POLY_MODE 0x4288 +#define FRONT_PTYPE_POINT (0 << 4) +#define FRONT_PTYPE_LINE (1 << 4) +#define FRONT_PTYPE_TRIANGE (2 << 4) +#define BACK_PTYPE_POINT (0 << 7) +#define BACK_PTYPE_LINE (1 << 7) +#define BACK_PTYPE_TRIANGE (2 << 7) +#define GA_ROUND_MODE 0x428C +#define GEOMETRY_ROUND_TRUNC (0 << 0) +#define GEOMETRY_ROUND_NEAREST (1 << 0) +#define COLOR_ROUND_TRUNC (0 << 2) +#define COLOR_ROUND_NEAREST (1 << 2) +#define SU_REG_DEST 0x42C8 +#define RB3D_DSTCACHE_CTLSTAT 0x4E4C +#define RB3D_DC_FLUSH (2 << 0) +#define RB3D_DC_FREE (2 << 2) +#define RB3D_DC_FINISH (1 << 4) +#define ZB_ZCACHE_CTLSTAT 0x4F18 +#define ZC_FLUSH (1 << 0) +#define ZC_FREE (1 << 1) +#define DC_LB_MEMORY_SPLIT 0x6520 +#define DC_LB_MEMORY_SPLIT_MASK 0x00000003 +#define DC_LB_MEMORY_SPLIT_SHIFT 0 +#define DC_LB_MEMORY_SPLIT_D1HALF_D2HALF 0 +#define DC_LB_MEMORY_SPLIT_D1_3Q_D2_1Q 1 +#define DC_LB_MEMORY_SPLIT_D1_ONLY 2 +#define DC_LB_MEMORY_SPLIT_D1_1Q_D2_3Q 3 +#define DC_LB_MEMORY_SPLIT_SHIFT_MODE (1 << 2) +#define DC_LB_DISP1_END_ADR_SHIFT 4 +#define DC_LB_DISP1_END_ADR_MASK 0x00007FF0 +#define D1MODE_PRIORITY_A_CNT 0x6548 +#define MODE_PRIORITY_MARK_MASK 0x00007FFF +#define MODE_PRIORITY_OFF (1 << 16) +#define MODE_PRIORITY_ALWAYS_ON (1 << 20) +#define MODE_PRIORITY_FORCE_MASK (1 << 24) +#define D1MODE_PRIORITY_B_CNT 0x654C +#define LB_MAX_REQ_OUTSTANDING 0x6D58 +#define LB_D1_MAX_REQ_OUTSTANDING_MASK 0x0000000F +#define LB_D1_MAX_REQ_OUTSTANDING_SHIFT 0 +#define LB_D2_MAX_REQ_OUTSTANDING_MASK 0x000F0000 +#define LB_D2_MAX_REQ_OUTSTANDING_SHIFT 16 +#define D2MODE_PRIORITY_A_CNT 0x6D48 +#define D2MODE_PRIORITY_B_CNT 0x6D4C + +/* ix[MC] registers */ +#define MC_FB_LOCATION 0x01 +#define MC_FB_START_MASK 0x0000FFFF +#define MC_FB_START_SHIFT 0 +#define MC_FB_TOP_MASK 0xFFFF0000 +#define MC_FB_TOP_SHIFT 16 +#define MC_AGP_LOCATION 0x02 +#define MC_AGP_START_MASK 0x0000FFFF +#define MC_AGP_START_SHIFT 0 +#define MC_AGP_TOP_MASK 0xFFFF0000 +#define MC_AGP_TOP_SHIFT 16 +#define MC_AGP_BASE 0x03 +#define MC_AGP_BASE_2 0x04 +#define MC_CNTL 0x5 +#define MEM_NUM_CHANNELS_MASK 0x00000003 +#define MC_STATUS 0x08 +#define MC_STATUS_IDLE (1 << 4) +#define MC_MISC_LAT_TIMER 0x09 +#define MC_CPR_INIT_LAT_MASK 0x0000000F +#define MC_VF_INIT_LAT_MASK 0x000000F0 +#define MC_DISP0R_INIT_LAT_MASK 0x00000F00 +#define MC_DISP0R_INIT_LAT_SHIFT 8 +#define MC_DISP1R_INIT_LAT_MASK 0x0000F000 +#define MC_DISP1R_INIT_LAT_SHIFT 12 +#define MC_FIXED_INIT_LAT_MASK 0x000F0000 +#define MC_E2R_INIT_LAT_MASK 0x00F00000 +#define SAME_PAGE_PRIO_MASK 0x0F000000 +#define MC_GLOBW_INIT_LAT_MASK 0xF0000000 + + +#endif + diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index da50cc51e..21d8ffd 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -67,7 +67,7 @@ int rv770_mc_init(struct radeon_device *rdev) "programming pipes. Bad things might happen.\n"); } - tmp = rdev->mc.vram_location + rdev->mc.vram_size - 1; + tmp = rdev->mc.vram_location + rdev->mc.mc_vram_size - 1; tmp = REG_SET(R700_MC_FB_TOP, tmp >> 24); tmp |= REG_SET(R700_MC_FB_BASE, rdev->mc.vram_location >> 24); WREG32(R700_MC_VM_FB_LOCATION, tmp); diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index c1c407f..6538d42 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -43,7 +43,6 @@ #define TTM_BO_HASH_ORDER 13 static int ttm_bo_setup_vm(struct ttm_buffer_object *bo); -static void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo); static int ttm_bo_swapout(struct ttm_mem_shrink *shrink); static inline uint32_t ttm_bo_type_flags(unsigned type) @@ -224,6 +223,9 @@ static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc) TTM_ASSERT_LOCKED(&bo->mutex); bo->ttm = NULL; + if (bdev->need_dma32) + page_flags |= TTM_PAGE_FLAG_DMA32; + switch (bo->type) { case ttm_bo_type_device: if (zero_alloc) @@ -304,6 +306,9 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, } + if (bdev->driver->move_notify) + bdev->driver->move_notify(bo, mem); + if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) && !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) ret = ttm_bo_move_ttm(bo, evict, no_wait, mem); @@ -655,31 +660,52 @@ retry_pre_get: return 0; } +static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man, + uint32_t cur_placement, + uint32_t proposed_placement) +{ + uint32_t caching = proposed_placement & TTM_PL_MASK_CACHING; + uint32_t result = proposed_placement & ~TTM_PL_MASK_CACHING; + + /** + * Keep current caching if possible. + */ + + if ((cur_placement & caching) != 0) + result |= (cur_placement & caching); + else if ((man->default_caching & caching) != 0) + result |= man->default_caching; + else if ((TTM_PL_FLAG_CACHED & caching) != 0) + result |= TTM_PL_FLAG_CACHED; + else if ((TTM_PL_FLAG_WC & caching) != 0) + result |= TTM_PL_FLAG_WC; + else if ((TTM_PL_FLAG_UNCACHED & caching) != 0) + result |= TTM_PL_FLAG_UNCACHED; + + return result; +} + + static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man, bool disallow_fixed, uint32_t mem_type, - uint32_t mask, uint32_t *res_mask) + uint32_t proposed_placement, + uint32_t *masked_placement) { uint32_t cur_flags = ttm_bo_type_flags(mem_type); if ((man->flags & TTM_MEMTYPE_FLAG_FIXED) && disallow_fixed) return false; - if ((cur_flags & mask & TTM_PL_MASK_MEM) == 0) + if ((cur_flags & proposed_placement & TTM_PL_MASK_MEM) == 0) return false; - if ((mask & man->available_caching) == 0) + if ((proposed_placement & man->available_caching) == 0) return false; - if (mask & man->default_caching) - cur_flags |= man->default_caching; - else if (mask & TTM_PL_FLAG_CACHED) - cur_flags |= TTM_PL_FLAG_CACHED; - else if (mask & TTM_PL_FLAG_WC) - cur_flags |= TTM_PL_FLAG_WC; - else - cur_flags |= TTM_PL_FLAG_UNCACHED; - *res_mask = cur_flags; + cur_flags |= (proposed_placement & man->available_caching); + + *masked_placement = cur_flags; return true; } @@ -723,6 +749,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo, if (!type_ok) continue; + cur_flags = ttm_bo_select_caching(man, bo->mem.placement, + cur_flags); + if (mem_type == TTM_PL_SYSTEM) break; @@ -779,6 +808,9 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo, proposed_placement, &cur_flags)) continue; + cur_flags = ttm_bo_select_caching(man, bo->mem.placement, + cur_flags); + ret = ttm_bo_mem_force_space(bdev, mem, mem_type, interruptible, no_wait); @@ -1305,7 +1337,8 @@ EXPORT_SYMBOL(ttm_bo_device_release); int ttm_bo_device_init(struct ttm_bo_device *bdev, struct ttm_mem_global *mem_glob, - struct ttm_bo_driver *driver, uint64_t file_page_offset) + struct ttm_bo_driver *driver, uint64_t file_page_offset, + bool need_dma32) { int ret = -EINVAL; @@ -1342,6 +1375,7 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev, INIT_LIST_HEAD(&bdev->ddestroy); INIT_LIST_HEAD(&bdev->swap_lru); bdev->dev_mapping = NULL; + bdev->need_dma32 = need_dma32; ttm_mem_init_shrink(&bdev->shrink, ttm_bo_swapout); ret = ttm_mem_register_shrink(mem_glob, &bdev->shrink); if (unlikely(ret != 0)) { @@ -1419,6 +1453,7 @@ void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo) unmap_mapping_range(bdev->dev_mapping, offset, holelen, 1); } +EXPORT_SYMBOL(ttm_bo_unmap_virtual); static void ttm_bo_vm_insert_rb(struct ttm_buffer_object *bo) { diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index bdec583..ce2e6f3 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -136,7 +136,8 @@ static int ttm_copy_io_page(void *dst, void *src, unsigned long page) } static int ttm_copy_io_ttm_page(struct ttm_tt *ttm, void *src, - unsigned long page) + unsigned long page, + pgprot_t prot) { struct page *d = ttm_tt_get_page(ttm, page); void *dst; @@ -145,17 +146,35 @@ static int ttm_copy_io_ttm_page(struct ttm_tt *ttm, void *src, return -ENOMEM; src = (void *)((unsigned long)src + (page << PAGE_SHIFT)); - dst = kmap(d); + +#ifdef CONFIG_X86 + dst = kmap_atomic_prot(d, KM_USER0, prot); +#else + if (prot != PAGE_KERNEL) + dst = vmap(&d, 1, 0, prot); + else + dst = kmap(d); +#endif if (!dst) return -ENOMEM; memcpy_fromio(dst, src, PAGE_SIZE); - kunmap(d); + +#ifdef CONFIG_X86 + kunmap_atomic(dst, KM_USER0); +#else + if (prot != PAGE_KERNEL) + vunmap(dst); + else + kunmap(d); +#endif + return 0; } static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst, - unsigned long page) + unsigned long page, + pgprot_t prot) { struct page *s = ttm_tt_get_page(ttm, page); void *src; @@ -164,12 +183,28 @@ static int ttm_copy_ttm_io_page(struct ttm_tt *ttm, void *dst, return -ENOMEM; dst = (void *)((unsigned long)dst + (page << PAGE_SHIFT)); - src = kmap(s); +#ifdef CONFIG_X86 + src = kmap_atomic_prot(s, KM_USER0, prot); +#else + if (prot != PAGE_KERNEL) + src = vmap(&s, 1, 0, prot); + else + src = kmap(s); +#endif if (!src) return -ENOMEM; memcpy_toio(dst, src, PAGE_SIZE); - kunmap(s); + +#ifdef CONFIG_X86 + kunmap_atomic(src, KM_USER0); +#else + if (prot != PAGE_KERNEL) + vunmap(src); + else + kunmap(s); +#endif + return 0; } @@ -214,11 +249,17 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo, for (i = 0; i < new_mem->num_pages; ++i) { page = i * dir + add; - if (old_iomap == NULL) - ret = ttm_copy_ttm_io_page(ttm, new_iomap, page); - else if (new_iomap == NULL) - ret = ttm_copy_io_ttm_page(ttm, old_iomap, page); - else + if (old_iomap == NULL) { + pgprot_t prot = ttm_io_prot(old_mem->placement, + PAGE_KERNEL); + ret = ttm_copy_ttm_io_page(ttm, new_iomap, page, + prot); + } else if (new_iomap == NULL) { + pgprot_t prot = ttm_io_prot(new_mem->placement, + PAGE_KERNEL); + ret = ttm_copy_io_ttm_page(ttm, old_iomap, page, + prot); + } else ret = ttm_copy_io_page(new_iomap, old_iomap, page); if (ret) goto out1; @@ -509,8 +550,8 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo, if (evict) { ret = ttm_bo_wait(bo, false, false, false); spin_unlock(&bo->lock); - driver->sync_obj_unref(&bo->sync_obj); - + if (tmp_obj) + driver->sync_obj_unref(&tmp_obj); if (ret) return ret; @@ -532,6 +573,8 @@ int ttm_bo_move_accel_cleanup(struct ttm_buffer_object *bo, set_bit(TTM_BO_PRIV_FLAG_MOVING, &bo->priv_flags); spin_unlock(&bo->lock); + if (tmp_obj) + driver->sync_obj_unref(&tmp_obj); ret = ttm_buffer_object_transfer(bo, &ghost_obj); if (ret) diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c index fe949a1..33de763 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c @@ -101,6 +101,9 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) return VM_FAULT_NOPAGE; } + if (bdev->driver->fault_reserve_notify) + bdev->driver->fault_reserve_notify(bo); + /* * Wait for buffer data in transit, due to a pipelined * move. diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c index 75dc8bd..b8b6c4a 100644 --- a/drivers/gpu/drm/ttm/ttm_tt.c +++ b/drivers/gpu/drm/ttm/ttm_tt.c @@ -86,10 +86,16 @@ void ttm_tt_cache_flush(struct page *pages[], unsigned long num_pages) unsigned long i; for (i = 0; i < num_pages; ++i) { - if (pages[i]) { - unsigned long start = (unsigned long)page_address(pages[i]); - flush_dcache_range(start, start + PAGE_SIZE); - } + struct page *page = pages[i]; + void *page_virtual; + + if (unlikely(page == NULL)) + continue; + + page_virtual = kmap_atomic(page, KM_USER0); + flush_dcache_range((unsigned long) page_virtual, + (unsigned long) page_virtual + PAGE_SIZE); + kunmap_atomic(page_virtual, KM_USER0); } #else if (on_each_cpu(ttm_tt_ipi_handler, NULL, 1) != 0) @@ -131,10 +137,17 @@ static void ttm_tt_free_page_directory(struct ttm_tt *ttm) static struct page *ttm_tt_alloc_page(unsigned page_flags) { + gfp_t gfp_flags = GFP_USER; + if (page_flags & TTM_PAGE_FLAG_ZERO_ALLOC) - return alloc_page(GFP_HIGHUSER | __GFP_ZERO); + gfp_flags |= __GFP_ZERO; + + if (page_flags & TTM_PAGE_FLAG_DMA32) + gfp_flags |= __GFP_DMA32; + else + gfp_flags |= __GFP_HIGHMEM; - return alloc_page(GFP_HIGHUSER); + return alloc_page(gfp_flags); } static void ttm_tt_free_user_pages(struct ttm_tt *ttm) diff --git a/include/drm/radeon_drm.h b/include/drm/radeon_drm.h index 41862e9..af4b482 100644 --- a/include/drm/radeon_drm.h +++ b/include/drm/radeon_drm.h @@ -506,6 +506,8 @@ typedef struct { #define DRM_RADEON_GEM_WAIT_IDLE 0x24 #define DRM_RADEON_CS 0x26 #define DRM_RADEON_INFO 0x27 +#define DRM_RADEON_GEM_SET_TILING 0x28 +#define DRM_RADEON_GEM_GET_TILING 0x29 #define DRM_IOCTL_RADEON_CP_INIT DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_INIT, drm_radeon_init_t) #define DRM_IOCTL_RADEON_CP_START DRM_IO( DRM_COMMAND_BASE + DRM_RADEON_CP_START) @@ -544,7 +546,8 @@ typedef struct { #define DRM_IOCTL_RADEON_GEM_WAIT_IDLE DRM_IOW(DRM_COMMAND_BASE + DRM_RADEON_GEM_WAIT_IDLE, struct drm_radeon_gem_wait_idle) #define DRM_IOCTL_RADEON_CS DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_CS, struct drm_radeon_cs) #define DRM_IOCTL_RADEON_INFO DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_INFO, struct drm_radeon_info) - +#define DRM_IOCTL_RADEON_SET_TILING DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_SET_TILING, struct drm_radeon_gem_set_tiling) +#define DRM_IOCTL_RADEON_GET_TILING DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_GET_TILING, struct drm_radeon_gem_get_tiling) typedef struct drm_radeon_init { enum { @@ -796,6 +799,24 @@ struct drm_radeon_gem_create { uint32_t flags; }; +#define RADEON_TILING_MACRO 0x1 +#define RADEON_TILING_MICRO 0x2 +#define RADEON_TILING_SWAP 0x4 +#define RADEON_TILING_SURFACE 0x8 /* this object requires a surface + * when mapped - i.e. front buffer */ + +struct drm_radeon_gem_set_tiling { + uint32_t handle; + uint32_t tiling_flags; + uint32_t pitch; +}; + +struct drm_radeon_gem_get_tiling { + uint32_t handle; + uint32_t tiling_flags; + uint32_t pitch; +}; + struct drm_radeon_gem_mmap { uint32_t handle; uint32_t pad; diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index 62ed733..a68829d 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -121,6 +121,7 @@ struct ttm_backend { #define TTM_PAGE_FLAG_SWAPPED (1 << 4) #define TTM_PAGE_FLAG_PERSISTANT_SWAP (1 << 5) #define TTM_PAGE_FLAG_ZERO_ALLOC (1 << 6) +#define TTM_PAGE_FLAG_DMA32 (1 << 7) enum ttm_caching_state { tt_uncached, @@ -353,6 +354,14 @@ struct ttm_bo_driver { int (*sync_obj_flush) (void *sync_obj, void *sync_arg); void (*sync_obj_unref) (void **sync_obj); void *(*sync_obj_ref) (void *sync_obj); + + /* hook to notify driver about a driver move so it + * can do tiling things */ + void (*move_notify)(struct ttm_buffer_object *bo, + struct ttm_mem_reg *new_mem); + /* notify the driver we are taking a fault on this BO + * and have reserved it */ + void (*fault_reserve_notify)(struct ttm_buffer_object *bo); }; #define TTM_NUM_MEM_TYPES 8 @@ -429,6 +438,8 @@ struct ttm_bo_device { */ struct delayed_work wq; + + bool need_dma32; }; /** @@ -648,7 +659,14 @@ extern int ttm_bo_device_release(struct ttm_bo_device *bdev); extern int ttm_bo_device_init(struct ttm_bo_device *bdev, struct ttm_mem_global *mem_glob, struct ttm_bo_driver *driver, - uint64_t file_page_offset); + uint64_t file_page_offset, bool need_dma32); + +/** + * ttm_bo_unmap_virtual + * + * @bo: tear down the virtual mappings for this BO + */ +extern void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo); /** * ttm_bo_reserve: diff --git a/include/drm/ttm/ttm_module.h b/include/drm/ttm/ttm_module.h index 889a4c79..d1d4338 100644 --- a/include/drm/ttm/ttm_module.h +++ b/include/drm/ttm/ttm_module.h @@ -33,7 +33,7 @@ #include <linux/kernel.h> -#define TTM_PFX "[TTM]" +#define TTM_PFX "[TTM] " enum ttm_global_types { TTM_GLOBAL_TTM_MEM = 0, |
