diff options
32 files changed, 849 insertions, 512 deletions
diff --git a/drivers/gpu/drm/drm_crtc_helper.c b/drivers/gpu/drm/drm_crtc_helper.c index 3963b3c..4231d6d 100644 --- a/drivers/gpu/drm/drm_crtc_helper.c +++ b/drivers/gpu/drm/drm_crtc_helper.c @@ -1020,6 +1020,9 @@ bool drm_helper_initial_config(struct drm_device *dev) { int count = 0; + /* disable all the possible outputs/crtcs before entering KMS mode */ + drm_helper_disable_unused_functions(dev); + drm_fb_helper_parse_command_line(dev); count = drm_helper_probe_connector_modes(dev, diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c index 1f0d717..d7d7eac 100644 --- a/drivers/gpu/drm/drm_mm.c +++ b/drivers/gpu/drm/drm_mm.c @@ -226,6 +226,44 @@ struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node, } EXPORT_SYMBOL(drm_mm_get_block_generic); +struct drm_mm_node *drm_mm_get_block_range_generic(struct drm_mm_node *node, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end, + int atomic) +{ + struct drm_mm_node *align_splitoff = NULL; + unsigned tmp = 0; + unsigned wasted = 0; + + if (node->start < start) + wasted += start - node->start; + if (alignment) + tmp = ((node->start + wasted) % alignment); + + if (tmp) + wasted += alignment - tmp; + if (wasted) { + align_splitoff = drm_mm_split_at_start(node, wasted, atomic); + if (unlikely(align_splitoff == NULL)) + return NULL; + } + + if (node->size == size) { + list_del_init(&node->fl_entry); + node->free = 0; + } else { + node = drm_mm_split_at_start(node, size, atomic); + } + + if (align_splitoff) + drm_mm_put_block(align_splitoff); + + return node; +} +EXPORT_SYMBOL(drm_mm_get_block_range_generic); + /* * Put a block. Merge with the previous and / or next block if they are free. * Otherwise add to the free stack. @@ -331,6 +369,56 @@ struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm, } EXPORT_SYMBOL(drm_mm_search_free); +struct drm_mm_node *drm_mm_search_free_in_range(const struct drm_mm *mm, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end, + int best_match) +{ + struct list_head *list; + const struct list_head *free_stack = &mm->fl_entry; + struct drm_mm_node *entry; + struct drm_mm_node *best; + unsigned long best_size; + unsigned wasted; + + best = NULL; + best_size = ~0UL; + + list_for_each(list, free_stack) { + entry = list_entry(list, struct drm_mm_node, fl_entry); + wasted = 0; + + if (entry->size < size) + continue; + + if (entry->start > end || (entry->start+entry->size) < start) + continue; + + if (entry->start < start) + wasted += start - entry->start; + + if (alignment) { + register unsigned tmp = (entry->start + wasted) % alignment; + if (tmp) + wasted += alignment - tmp; + } + + if (entry->size >= size + wasted) { + if (!best_match) + return entry; + if (size < best_size) { + best = entry; + best_size = entry->size; + } + } + } + + return best; +} +EXPORT_SYMBOL(drm_mm_search_free_in_range); + int drm_mm_clean(struct drm_mm * mm) { struct list_head *head = &mm->ml_entry; @@ -381,6 +469,26 @@ void drm_mm_takedown(struct drm_mm * mm) } EXPORT_SYMBOL(drm_mm_takedown); +void drm_mm_debug_table(struct drm_mm *mm, const char *prefix) +{ + struct drm_mm_node *entry; + int total_used = 0, total_free = 0, total = 0; + + list_for_each_entry(entry, &mm->ml_entry, ml_entry) { + printk(KERN_DEBUG "%s 0x%08lx-0x%08lx: %8ld: %s\n", + prefix, entry->start, entry->start + entry->size, + entry->size, entry->free ? "free" : "used"); + total += entry->size; + if (entry->free) + total_free += entry->size; + else + total_used += entry->size; + } + printk(KERN_DEBUG "%s total: %d, used %d free %d\n", prefix, total, + total_used, total_free); +} +EXPORT_SYMBOL(drm_mm_debug_table); + #if defined(CONFIG_DEBUG_FS) int drm_mm_dump_table(struct seq_file *m, struct drm_mm *mm) { diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c index fba3c96..260fcf5 100644 --- a/drivers/gpu/drm/radeon/atombios_crtc.c +++ b/drivers/gpu/drm/radeon/atombios_crtc.c @@ -499,8 +499,18 @@ void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode *mode) else pll = &rdev->clock.p2pll; - radeon_compute_pll(pll, adjusted_clock, &pll_clock, &fb_div, &frac_fb_div, - &ref_div, &post_div, pll_flags); + if (ASIC_IS_AVIVO(rdev)) { + if (radeon_new_pll) + radeon_compute_pll_avivo(pll, adjusted_clock, &pll_clock, + &fb_div, &frac_fb_div, + &ref_div, &post_div, pll_flags); + else + radeon_compute_pll(pll, adjusted_clock, &pll_clock, + &fb_div, &frac_fb_div, + &ref_div, &post_div, pll_flags); + } else + radeon_compute_pll(pll, adjusted_clock, &pll_clock, &fb_div, &frac_fb_div, + &ref_div, &post_div, pll_flags); index = GetIndexIntoMasterTable(COMMAND, SetPixelClock); atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, @@ -599,8 +609,6 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y, } radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL); radeon_bo_unreserve(rbo); - if (tiling_flags & RADEON_TILING_MACRO) - fb_format |= AVIVO_D1GRPH_MACRO_ADDRESS_MODE; switch (crtc->fb->bits_per_pixel) { case 8: @@ -630,6 +638,9 @@ int atombios_crtc_set_base(struct drm_crtc *crtc, int x, int y, return -EINVAL; } + if (tiling_flags & RADEON_TILING_MACRO) + fb_format |= AVIVO_D1GRPH_MACRO_ADDRESS_MODE; + if (tiling_flags & RADEON_TILING_MICRO) fb_format |= AVIVO_D1GRPH_TILED; diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index b7baf16..824cc64 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -3299,6 +3299,8 @@ int r100_resume(struct radeon_device *rdev) radeon_combios_asic_init(rdev->ddev); /* Resume clock after posting */ r100_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return r100_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c index 86065dc..83378c3 100644 --- a/drivers/gpu/drm/radeon/r300.c +++ b/drivers/gpu/drm/radeon/r300.c @@ -1250,6 +1250,8 @@ int r300_resume(struct radeon_device *rdev) radeon_combios_asic_init(rdev->ddev); /* Resume clock after posting */ r300_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return r300_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/r420.c b/drivers/gpu/drm/radeon/r420.c index 162c390..c05a727 100644 --- a/drivers/gpu/drm/radeon/r420.c +++ b/drivers/gpu/drm/radeon/r420.c @@ -231,7 +231,8 @@ int r420_resume(struct radeon_device *rdev) } /* Resume clock after posting */ r420_clock_resume(rdev); - + /* Initialize surface registers */ + radeon_surface_init(rdev); return r420_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/r520.c b/drivers/gpu/drm/radeon/r520.c index 788eef5..0f3843b 100644 --- a/drivers/gpu/drm/radeon/r520.c +++ b/drivers/gpu/drm/radeon/r520.c @@ -220,6 +220,8 @@ int r520_resume(struct radeon_device *rdev) atom_asic_init(rdev->mode_info.atom_context); /* Resume clock after posting */ rv515_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return r520_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 250ec3f..f5cf874 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -1845,6 +1845,14 @@ int r600_startup(struct radeon_device *rdev) { int r; + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { + r = r600_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + r600_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { r600_agp_enable(rdev); @@ -2026,25 +2034,17 @@ int r600_init(struct radeon_device *rdev) rdev->ih.ring_obj = NULL; r600_ih_ring_init(rdev, 64 * 1024); - if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { - r = r600_init_microcode(rdev); - if (r) { - DRM_ERROR("Failed to load firmware!\n"); - return r; - } - } - r = r600_pcie_gart_init(rdev); if (r) return r; - rdev->accel_working = true; r = r600_blit_init(rdev); if (r) { - DRM_ERROR("radeon: failled blitter (%d).\n", r); + DRM_ERROR("radeon: failed blitter (%d).\n", r); return r; } + rdev->accel_working = true; r = r600_startup(rdev); if (r) { r600_suspend(rdev); @@ -2056,12 +2056,12 @@ int r600_init(struct radeon_device *rdev) if (rdev->accel_working) { r = radeon_ib_pool_init(rdev); if (r) { - DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r); + DRM_ERROR("radeon: failed initializing IB pool (%d).\n", r); rdev->accel_working = false; } r = r600_ib_test(rdev); if (r) { - DRM_ERROR("radeon: failled testing IB (%d).\n", r); + DRM_ERROR("radeon: failed testing IB (%d).\n", r); rdev->accel_working = false; } } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index a15cf9c..c938bb5 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -88,6 +88,7 @@ extern int radeon_benchmarking; extern int radeon_testing; extern int radeon_connector_table; extern int radeon_tv; +extern int radeon_new_pll; /* * Copy from radeon_drv.h so we don't have to include both and have conflicting @@ -208,6 +209,8 @@ struct radeon_bo { /* Protected by gem.mutex */ struct list_head list; /* Protected by tbo.reserved */ + u32 placements[3]; + struct ttm_placement placement; struct ttm_buffer_object tbo; struct ttm_bo_kmap_obj kmap; unsigned pin_count; @@ -1012,6 +1015,7 @@ extern void radeon_surface_init(struct radeon_device *rdev); extern int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data); extern void radeon_legacy_set_clock_gating(struct radeon_device *rdev, int enable); extern void radeon_atom_set_clock_gating(struct radeon_device *rdev, int enable); +extern void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain); /* r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280 */ struct r100_mc_save { diff --git a/drivers/gpu/drm/radeon/radeon_atombios.c b/drivers/gpu/drm/radeon/radeon_atombios.c index d7b0feb..12a0c76 100644 --- a/drivers/gpu/drm/radeon/radeon_atombios.c +++ b/drivers/gpu/drm/radeon/radeon_atombios.c @@ -70,6 +70,7 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_i2c_gpio(struct radeon_dev int index = GetIndexIntoMasterTable(DATA, GPIO_I2C_Info); struct _ATOM_GPIO_I2C_INFO *i2c_info; uint16_t data_offset; + int i; memset(&i2c, 0, sizeof(struct radeon_i2c_bus_rec)); i2c.valid = false; @@ -78,38 +79,43 @@ static inline struct radeon_i2c_bus_rec radeon_lookup_i2c_gpio(struct radeon_dev i2c_info = (struct _ATOM_GPIO_I2C_INFO *)(ctx->bios + data_offset); - gpio = &i2c_info->asGPIO_Info[id]; - - i2c.mask_clk_reg = le16_to_cpu(gpio->usClkMaskRegisterIndex) * 4; - i2c.mask_data_reg = le16_to_cpu(gpio->usDataMaskRegisterIndex) * 4; - i2c.en_clk_reg = le16_to_cpu(gpio->usClkEnRegisterIndex) * 4; - i2c.en_data_reg = le16_to_cpu(gpio->usDataEnRegisterIndex) * 4; - i2c.y_clk_reg = le16_to_cpu(gpio->usClkY_RegisterIndex) * 4; - i2c.y_data_reg = le16_to_cpu(gpio->usDataY_RegisterIndex) * 4; - i2c.a_clk_reg = le16_to_cpu(gpio->usClkA_RegisterIndex) * 4; - i2c.a_data_reg = le16_to_cpu(gpio->usDataA_RegisterIndex) * 4; - i2c.mask_clk_mask = (1 << gpio->ucClkMaskShift); - i2c.mask_data_mask = (1 << gpio->ucDataMaskShift); - i2c.en_clk_mask = (1 << gpio->ucClkEnShift); - i2c.en_data_mask = (1 << gpio->ucDataEnShift); - i2c.y_clk_mask = (1 << gpio->ucClkY_Shift); - i2c.y_data_mask = (1 << gpio->ucDataY_Shift); - i2c.a_clk_mask = (1 << gpio->ucClkA_Shift); - i2c.a_data_mask = (1 << gpio->ucDataA_Shift); - - if (gpio->sucI2cId.sbfAccess.bfHW_Capable) - i2c.hw_capable = true; - else - i2c.hw_capable = false; - if (gpio->sucI2cId.ucAccess == 0xa0) - i2c.mm_i2c = true; - else - i2c.mm_i2c = false; + for (i = 0; i < ATOM_MAX_SUPPORTED_DEVICE; i++) { + gpio = &i2c_info->asGPIO_Info[i]; + + if (gpio->sucI2cId.ucAccess == id) { + i2c.mask_clk_reg = le16_to_cpu(gpio->usClkMaskRegisterIndex) * 4; + i2c.mask_data_reg = le16_to_cpu(gpio->usDataMaskRegisterIndex) * 4; + i2c.en_clk_reg = le16_to_cpu(gpio->usClkEnRegisterIndex) * 4; + i2c.en_data_reg = le16_to_cpu(gpio->usDataEnRegisterIndex) * 4; + i2c.y_clk_reg = le16_to_cpu(gpio->usClkY_RegisterIndex) * 4; + i2c.y_data_reg = le16_to_cpu(gpio->usDataY_RegisterIndex) * 4; + i2c.a_clk_reg = le16_to_cpu(gpio->usClkA_RegisterIndex) * 4; + i2c.a_data_reg = le16_to_cpu(gpio->usDataA_RegisterIndex) * 4; + i2c.mask_clk_mask = (1 << gpio->ucClkMaskShift); + i2c.mask_data_mask = (1 << gpio->ucDataMaskShift); + i2c.en_clk_mask = (1 << gpio->ucClkEnShift); + i2c.en_data_mask = (1 << gpio->ucDataEnShift); + i2c.y_clk_mask = (1 << gpio->ucClkY_Shift); + i2c.y_data_mask = (1 << gpio->ucDataY_Shift); + i2c.a_clk_mask = (1 << gpio->ucClkA_Shift); + i2c.a_data_mask = (1 << gpio->ucDataA_Shift); + + if (gpio->sucI2cId.sbfAccess.bfHW_Capable) + i2c.hw_capable = true; + else + i2c.hw_capable = false; - i2c.i2c_id = gpio->sucI2cId.ucAccess; + if (gpio->sucI2cId.ucAccess == 0xa0) + i2c.mm_i2c = true; + else + i2c.mm_i2c = false; + + i2c.i2c_id = gpio->sucI2cId.ucAccess; - i2c.valid = true; + i2c.valid = true; + } + } return i2c; } @@ -503,6 +509,7 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev) usRecordOffset)); ATOM_I2C_RECORD *i2c_record; ATOM_HPD_INT_RECORD *hpd_record; + ATOM_I2C_ID_CONFIG_ACCESS *i2c_config; hpd.hpd = RADEON_HPD_NONE; while (record->ucRecordType > 0 @@ -514,10 +521,12 @@ bool radeon_get_atom_connector_info_from_object_table(struct drm_device *dev) i2c_record = (ATOM_I2C_RECORD *) record; + i2c_config = + (ATOM_I2C_ID_CONFIG_ACCESS *) + &i2c_record->sucI2cId; ddc_bus = radeon_lookup_i2c_gpio(rdev, - i2c_record-> - sucI2cId. - bfI2C_LineMux); + i2c_config-> + ucAccess); break; case ATOM_HPD_INT_RECORD_TYPE: hpd_record = @@ -670,22 +679,8 @@ bool radeon_get_atom_connector_info_from_supported_devices_table(struct dac = ci.sucConnectorInfo.sbfAccess.bfAssociatedDAC; - if ((rdev->family == CHIP_RS690) || - (rdev->family == CHIP_RS740)) { - if ((i == ATOM_DEVICE_DFP2_INDEX) - && (ci.sucI2cId.sbfAccess.bfI2C_LineMux == 2)) - bios_connectors[i].line_mux = - ci.sucI2cId.sbfAccess.bfI2C_LineMux + 1; - else if ((i == ATOM_DEVICE_DFP3_INDEX) - && (ci.sucI2cId.sbfAccess.bfI2C_LineMux == 1)) - bios_connectors[i].line_mux = - ci.sucI2cId.sbfAccess.bfI2C_LineMux + 1; - else - bios_connectors[i].line_mux = - ci.sucI2cId.sbfAccess.bfI2C_LineMux; - } else - bios_connectors[i].line_mux = - ci.sucI2cId.sbfAccess.bfI2C_LineMux; + bios_connectors[i].line_mux = + ci.sucI2cId.ucAccess; /* give tv unique connector ids */ if (i == ATOM_DEVICE_TV1_INDEX) { @@ -876,7 +871,8 @@ bool radeon_atom_get_clock_info(struct drm_device *dev) * pre-DCE 3.0 r6xx hardware. This might need to be adjusted per * family. */ - p1pll->pll_out_min = 64800; + if (!radeon_new_pll) + p1pll->pll_out_min = 64800; } p1pll->pll_in_min = @@ -1006,6 +1002,7 @@ static struct radeon_atom_ss *radeon_atombios_get_ss_info(struct struct _ATOM_SPREAD_SPECTRUM_INFO *ss_info; uint8_t frev, crev; struct radeon_atom_ss *ss = NULL; + int i; if (id > ATOM_MAX_SS_ENTRY) return NULL; @@ -1023,12 +1020,17 @@ static struct radeon_atom_ss *radeon_atombios_get_ss_info(struct if (!ss) return NULL; - ss->percentage = le16_to_cpu(ss_info->asSS_Info[id].usSpreadSpectrumPercentage); - ss->type = ss_info->asSS_Info[id].ucSpreadSpectrumType; - ss->step = ss_info->asSS_Info[id].ucSS_Step; - ss->delay = ss_info->asSS_Info[id].ucSS_Delay; - ss->range = ss_info->asSS_Info[id].ucSS_Range; - ss->refdiv = ss_info->asSS_Info[id].ucRecommendedRef_Div; + for (i = 0; i < ATOM_MAX_SS_ENTRY; i++) { + if (ss_info->asSS_Info[i].ucSS_Id == id) { + ss->percentage = + le16_to_cpu(ss_info->asSS_Info[i].usSpreadSpectrumPercentage); + ss->type = ss_info->asSS_Info[i].ucSpreadSpectrumType; + ss->step = ss_info->asSS_Info[i].ucSS_Step; + ss->delay = ss_info->asSS_Info[i].ucSS_Delay; + ss->range = ss_info->asSS_Info[i].ucSS_Range; + ss->refdiv = ss_info->asSS_Info[i].ucRecommendedRef_Div; + } + } } return ss; } diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c index cfa2ebb..5eece18 100644 --- a/drivers/gpu/drm/radeon/radeon_connectors.c +++ b/drivers/gpu/drm/radeon/radeon_connectors.c @@ -1103,10 +1103,12 @@ radeon_add_atom_connector(struct drm_device *dev, drm_connector_attach_property(&radeon_connector->base, rdev->mode_info.coherent_mode_property, 1); - radeon_connector->dac_load_detect = true; - drm_connector_attach_property(&radeon_connector->base, - rdev->mode_info.load_detect_property, - 1); + if (connector_type == DRM_MODE_CONNECTOR_DVII) { + radeon_connector->dac_load_detect = true; + drm_connector_attach_property(&radeon_connector->base, + rdev->mode_info.load_detect_property, + 1); + } break; case DRM_MODE_CONNECTOR_HDMIA: case DRM_MODE_CONNECTOR_HDMIB: @@ -1141,14 +1143,19 @@ radeon_add_atom_connector(struct drm_device *dev, ret = drm_connector_helper_add(&radeon_connector->base, &radeon_dp_connector_helper_funcs); if (ret) goto failed; - /* add DP i2c bus */ - radeon_dig_connector->dp_i2c_bus = radeon_i2c_create_dp(dev, i2c_bus, "DP-auxch"); if (i2c_bus->valid) { + /* add DP i2c bus */ + radeon_dig_connector->dp_i2c_bus = radeon_i2c_create_dp(dev, i2c_bus, "DP-auxch"); + if (!radeon_dig_connector->dp_i2c_bus) + goto failed; radeon_connector->ddc_bus = radeon_i2c_create(dev, i2c_bus, "DP"); if (!radeon_connector->ddc_bus) goto failed; } subpixel_order = SubPixelHorizontalRGB; + drm_connector_attach_property(&radeon_connector->base, + rdev->mode_info.coherent_mode_property, + 1); break; case DRM_MODE_CONNECTOR_SVIDEO: case DRM_MODE_CONNECTOR_Composite: @@ -1183,7 +1190,6 @@ radeon_add_atom_connector(struct drm_device *dev, if (!radeon_connector->ddc_bus) goto failed; } - drm_mode_create_scaling_mode_property(dev); drm_connector_attach_property(&radeon_connector->base, dev->mode_config.scaling_mode_property, DRM_MODE_SCALE_FULLSCREEN); diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c index 7e55647..02bcdb1 100644 --- a/drivers/gpu/drm/radeon/radeon_device.c +++ b/drivers/gpu/drm/radeon/radeon_device.c @@ -44,10 +44,11 @@ void radeon_surface_init(struct radeon_device *rdev) if (rdev->family < CHIP_R600) { int i; - for (i = 0; i < 8; i++) { - WREG32(RADEON_SURFACE0_INFO + - i * (RADEON_SURFACE1_INFO - RADEON_SURFACE0_INFO), - 0); + for (i = 0; i < RADEON_GEM_MAX_SURFACES; i++) { + if (rdev->surface_regs[i].bo) + radeon_bo_get_surface_reg(rdev->surface_regs[i].bo); + else + radeon_clear_surface_reg(rdev, i); } /* enable surfaces */ WREG32(RADEON_SURFACE_CNTL, 0); @@ -487,8 +488,10 @@ int radeon_atombios_init(struct radeon_device *rdev) void radeon_atombios_fini(struct radeon_device *rdev) { - kfree(rdev->mode_info.atom_context->scratch); - kfree(rdev->mode_info.atom_context); + if (rdev->mode_info.atom_context) { + kfree(rdev->mode_info.atom_context->scratch); + kfree(rdev->mode_info.atom_context); + } kfree(rdev->mode_info.atom_card_info); } diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c index c115f2e..a133b83 100644 --- a/drivers/gpu/drm/radeon/radeon_display.c +++ b/drivers/gpu/drm/radeon/radeon_display.c @@ -560,6 +560,98 @@ void radeon_compute_pll(struct radeon_pll *pll, *post_div_p = best_post_div; } +void radeon_compute_pll_avivo(struct radeon_pll *pll, + uint64_t freq, + uint32_t *dot_clock_p, + uint32_t *fb_div_p, + uint32_t *frac_fb_div_p, + uint32_t *ref_div_p, + uint32_t *post_div_p, + int flags) +{ + fixed20_12 m, n, frac_n, p, f_vco, f_pclk, best_freq; + fixed20_12 pll_out_max, pll_out_min; + fixed20_12 pll_in_max, pll_in_min; + fixed20_12 reference_freq; + fixed20_12 error, ffreq, a, b; + + pll_out_max.full = rfixed_const(pll->pll_out_max); + pll_out_min.full = rfixed_const(pll->pll_out_min); + pll_in_max.full = rfixed_const(pll->pll_in_max); + pll_in_min.full = rfixed_const(pll->pll_in_min); + reference_freq.full = rfixed_const(pll->reference_freq); + do_div(freq, 10); + ffreq.full = rfixed_const(freq); + error.full = rfixed_const(100 * 100); + + /* max p */ + p.full = rfixed_div(pll_out_max, ffreq); + p.full = rfixed_floor(p); + + /* min m */ + m.full = rfixed_div(reference_freq, pll_in_max); + m.full = rfixed_ceil(m); + + while (1) { + n.full = rfixed_div(ffreq, reference_freq); + n.full = rfixed_mul(n, m); + n.full = rfixed_mul(n, p); + + f_vco.full = rfixed_div(n, m); + f_vco.full = rfixed_mul(f_vco, reference_freq); + + f_pclk.full = rfixed_div(f_vco, p); + + if (f_pclk.full > ffreq.full) + error.full = f_pclk.full - ffreq.full; + else + error.full = ffreq.full - f_pclk.full; + error.full = rfixed_div(error, f_pclk); + a.full = rfixed_const(100 * 100); + error.full = rfixed_mul(error, a); + + a.full = rfixed_mul(m, p); + a.full = rfixed_div(n, a); + best_freq.full = rfixed_mul(reference_freq, a); + + if (rfixed_trunc(error) < 25) + break; + + a.full = rfixed_const(1); + m.full = m.full + a.full; + a.full = rfixed_div(reference_freq, m); + if (a.full >= pll_in_min.full) + continue; + + m.full = rfixed_div(reference_freq, pll_in_max); + m.full = rfixed_ceil(m); + a.full= rfixed_const(1); + p.full = p.full - a.full; + a.full = rfixed_mul(p, ffreq); + if (a.full >= pll_out_min.full) + continue; + else { + DRM_ERROR("Unable to find pll dividers\n"); + break; + } + } + + a.full = rfixed_const(10); + b.full = rfixed_mul(n, a); + + frac_n.full = rfixed_floor(n); + frac_n.full = rfixed_mul(frac_n, a); + frac_n.full = b.full - frac_n.full; + + *dot_clock_p = rfixed_trunc(best_freq); + *fb_div_p = rfixed_trunc(n); + *frac_fb_div_p = rfixed_trunc(frac_n); + *ref_div_p = rfixed_trunc(m); + *post_div_p = rfixed_trunc(p); + + DRM_DEBUG("%u %d.%d, %d, %d\n", *dot_clock_p * 10, *fb_div_p, *frac_fb_div_p, *ref_div_p, *post_div_p); +} + static void radeon_user_framebuffer_destroy(struct drm_framebuffer *fb) { struct radeon_framebuffer *radeon_fb = to_radeon_framebuffer(fb); @@ -660,7 +752,7 @@ int radeon_modeset_create_props(struct radeon_device *rdev) return -ENOMEM; rdev->mode_info.coherent_mode_property->values[0] = 0; - rdev->mode_info.coherent_mode_property->values[0] = 1; + rdev->mode_info.coherent_mode_property->values[1] = 1; } if (!ASIC_IS_AVIVO(rdev)) { @@ -684,7 +776,7 @@ int radeon_modeset_create_props(struct radeon_device *rdev) if (!rdev->mode_info.load_detect_property) return -ENOMEM; rdev->mode_info.load_detect_property->values[0] = 0; - rdev->mode_info.load_detect_property->values[0] = 1; + rdev->mode_info.load_detect_property->values[1] = 1; drm_mode_create_scaling_mode_property(rdev->ddev); diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c index 7f50fb8..2807724 100644 --- a/drivers/gpu/drm/radeon/radeon_drv.c +++ b/drivers/gpu/drm/radeon/radeon_drv.c @@ -86,6 +86,7 @@ int radeon_benchmarking = 0; int radeon_testing = 0; int radeon_connector_table = 0; int radeon_tv = 1; +int radeon_new_pll = 1; MODULE_PARM_DESC(no_wb, "Disable AGP writeback for scratch registers"); module_param_named(no_wb, radeon_no_wb, int, 0444); @@ -120,6 +121,9 @@ module_param_named(connector_table, radeon_connector_table, int, 0444); MODULE_PARM_DESC(tv, "TV enable (0 = disable)"); module_param_named(tv, radeon_tv, int, 0444); +MODULE_PARM_DESC(r4xx_atom, "Select new PLL code for AVIVO chips"); +module_param_named(new_pll, radeon_new_pll, int, 0444); + static int radeon_suspend(struct drm_device *dev, pm_message_t state) { drm_radeon_private_t *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/radeon/radeon_fence.c index 2ac3163..cb4cd97 100644 --- a/drivers/gpu/drm/radeon/radeon_fence.c +++ b/drivers/gpu/drm/radeon/radeon_fence.c @@ -197,9 +197,8 @@ retry: r = wait_event_interruptible_timeout(rdev->fence_drv.queue, radeon_fence_signaled(fence), timeout); radeon_irq_kms_sw_irq_put(rdev); - if (unlikely(r == -ERESTARTSYS)) { - return -EBUSY; - } + if (unlikely(r < 0)) + return r; } else { radeon_irq_kms_sw_irq_get(rdev); r = wait_event_timeout(rdev->fence_drv.queue, diff --git a/drivers/gpu/drm/radeon/radeon_fixed.h b/drivers/gpu/drm/radeon/radeon_fixed.h index 90187d1..3d4d84e 100644 --- a/drivers/gpu/drm/radeon/radeon_fixed.h +++ b/drivers/gpu/drm/radeon/radeon_fixed.h @@ -38,6 +38,23 @@ typedef union rfixed { #define fixed_init_half(A) { .full = rfixed_const_half((A)) } #define rfixed_trunc(A) ((A).full >> 12) +static inline u32 rfixed_floor(fixed20_12 A) +{ + u32 non_frac = rfixed_trunc(A); + + return rfixed_const(non_frac); +} + +static inline u32 rfixed_ceil(fixed20_12 A) +{ + u32 non_frac = rfixed_trunc(A); + + if (A.full > rfixed_const(non_frac)) + return rfixed_const(non_frac + 1); + else + return rfixed_const(non_frac); +} + static inline u32 rfixed_div(fixed20_12 A, fixed20_12 B) { u64 tmp = ((u64)A.full << 13); diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index ba12862..f23b056 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -30,10 +30,19 @@ #include "radeon.h" #include "radeon_drm.h" +int radeon_driver_unload_kms(struct drm_device *dev) +{ + struct radeon_device *rdev = dev->dev_private; + + if (rdev == NULL) + return 0; + radeon_modeset_fini(rdev); + radeon_device_fini(rdev); + kfree(rdev); + dev->dev_private = NULL; + return 0; +} -/* - * Driver load/unload - */ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags) { struct radeon_device *rdev; @@ -62,31 +71,20 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags) */ r = radeon_device_init(rdev, dev, dev->pdev, flags); if (r) { - DRM_ERROR("Fatal error while trying to initialize radeon.\n"); - return r; + dev_err(&dev->pdev->dev, "Fatal error during GPU init\n"); + goto out; } /* Again modeset_init should fail only on fatal error * otherwise it should provide enough functionalities * for shadowfb to run */ r = radeon_modeset_init(rdev); - if (r) { - return r; - } - return 0; -} - -int radeon_driver_unload_kms(struct drm_device *dev) -{ - struct radeon_device *rdev = dev->dev_private; - - if (rdev == NULL) - return 0; - radeon_modeset_fini(rdev); - radeon_device_fini(rdev); - kfree(rdev); - dev->dev_private = NULL; - return 0; + if (r) + dev_err(&dev->pdev->dev, "Fatal error during modeset init\n"); +out: + if (r) + radeon_driver_unload_kms(dev); + return r; } diff --git a/drivers/gpu/drm/radeon/radeon_mode.h b/drivers/gpu/drm/radeon/radeon_mode.h index 15ec7ca..44d4b65 100644 --- a/drivers/gpu/drm/radeon/radeon_mode.h +++ b/drivers/gpu/drm/radeon/radeon_mode.h @@ -437,6 +437,15 @@ extern void radeon_compute_pll(struct radeon_pll *pll, uint32_t *post_div_p, int flags); +extern void radeon_compute_pll_avivo(struct radeon_pll *pll, + uint64_t freq, + uint32_t *dot_clock_p, + uint32_t *fb_div_p, + uint32_t *frac_fb_div_p, + uint32_t *ref_div_p, + uint32_t *post_div_p, + int flags); + extern void radeon_setup_encoder_clones(struct drm_device *dev); struct drm_encoder *radeon_encoder_legacy_lvds_add(struct drm_device *dev, int bios_index); diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index bec4943..2040937 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -75,6 +75,25 @@ static inline u32 radeon_ttm_flags_from_domain(u32 domain) return flags; } +void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain) +{ + u32 c = 0; + + rbo->placement.fpfn = 0; + rbo->placement.lpfn = 0; + rbo->placement.placement = rbo->placements; + rbo->placement.busy_placement = rbo->placements; + if (domain & RADEON_GEM_DOMAIN_VRAM) + rbo->placements[c++] = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | + TTM_PL_FLAG_VRAM; + if (domain & RADEON_GEM_DOMAIN_GTT) + rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; + if (domain & RADEON_GEM_DOMAIN_CPU) + rbo->placements[c++] = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; + rbo->placement.num_placement = c; + rbo->placement.num_busy_placement = c; +} + int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj, unsigned long size, bool kernel, u32 domain, struct radeon_bo **bo_ptr) @@ -102,16 +121,15 @@ int radeon_bo_create(struct radeon_device *rdev, struct drm_gem_object *gobj, INIT_LIST_HEAD(&bo->list); flags = radeon_ttm_flags_from_domain(domain); -retry: + /* Kernel allocation are uninterruptible */ r = ttm_buffer_object_init(&rdev->mman.bdev, &bo->tbo, size, type, - flags, 0, 0, true, NULL, size, + flags, 0, 0, !kernel, NULL, size, &radeon_ttm_bo_destroy); if (unlikely(r != 0)) { - if (r == -ERESTART) - goto retry; - /* ttm call radeon_ttm_object_object_destroy if error happen */ - dev_err(rdev->dev, "object_init failed for (%ld, 0x%08X)\n", - size, flags); + if (r != -ERESTARTSYS) + dev_err(rdev->dev, + "object_init failed for (%ld, 0x%08X)\n", + size, flags); return r; } *bo_ptr = bo; @@ -169,40 +187,32 @@ void radeon_bo_unref(struct radeon_bo **bo) int radeon_bo_pin(struct radeon_bo *bo, u32 domain, u64 *gpu_addr) { - u32 flags; - u32 tmp; - int r; + int r, i; - flags = radeon_ttm_flags_from_domain(domain); + radeon_ttm_placement_from_domain(bo, domain); if (bo->pin_count) { bo->pin_count++; if (gpu_addr) *gpu_addr = radeon_bo_gpu_offset(bo); return 0; } - tmp = bo->tbo.mem.placement; - ttm_flag_masked(&tmp, flags, TTM_PL_MASK_MEM); - bo->tbo.proposed_placement = tmp | TTM_PL_FLAG_NO_EVICT | - TTM_PL_MASK_CACHING; -retry: - r = ttm_buffer_object_validate(&bo->tbo, bo->tbo.proposed_placement, - true, false); + radeon_ttm_placement_from_domain(bo, domain); + for (i = 0; i < bo->placement.num_placement; i++) + bo->placements[i] |= TTM_PL_FLAG_NO_EVICT; + r = ttm_buffer_object_validate(&bo->tbo, &bo->placement, false, false); if (likely(r == 0)) { bo->pin_count = 1; if (gpu_addr != NULL) *gpu_addr = radeon_bo_gpu_offset(bo); } - if (unlikely(r != 0)) { - if (r == -ERESTART) - goto retry; + if (unlikely(r != 0)) dev_err(bo->rdev->dev, "%p pin failed\n", bo); - } return r; } int radeon_bo_unpin(struct radeon_bo *bo) { - int r; + int r, i; if (!bo->pin_count) { dev_warn(bo->rdev->dev, "%p unpin not necessary\n", bo); @@ -211,18 +221,12 @@ int radeon_bo_unpin(struct radeon_bo *bo) bo->pin_count--; if (bo->pin_count) return 0; - bo->tbo.proposed_placement = bo->tbo.mem.placement & - ~TTM_PL_FLAG_NO_EVICT; -retry: - r = ttm_buffer_object_validate(&bo->tbo, bo->tbo.proposed_placement, - true, false); - if (unlikely(r != 0)) { - if (r == -ERESTART) - goto retry; + for (i = 0; i < bo->placement.num_placement; i++) + bo->placements[i] &= ~TTM_PL_FLAG_NO_EVICT; + r = ttm_buffer_object_validate(&bo->tbo, &bo->placement, false, false); + if (unlikely(r != 0)) dev_err(bo->rdev->dev, "%p validate failed for unpin\n", bo); - return r; - } - return 0; + return r; } int radeon_bo_evict_vram(struct radeon_device *rdev) @@ -326,21 +330,17 @@ int radeon_bo_list_validate(struct list_head *head, void *fence) bo = lobj->bo; if (!bo->pin_count) { if (lobj->wdomain) { - bo->tbo.proposed_placement = - radeon_ttm_flags_from_domain(lobj->wdomain); + radeon_ttm_placement_from_domain(bo, + lobj->wdomain); } else { - bo->tbo.proposed_placement = - radeon_ttm_flags_from_domain(lobj->rdomain); + radeon_ttm_placement_from_domain(bo, + lobj->rdomain); } -retry: r = ttm_buffer_object_validate(&bo->tbo, - bo->tbo.proposed_placement, + &bo->placement, true, false); - if (unlikely(r)) { - if (r == -ERESTART) - goto retry; + if (unlikely(r)) return r; - } } lobj->gpu_offset = radeon_bo_gpu_offset(bo); lobj->tiling_flags = bo->tiling_flags; @@ -378,7 +378,7 @@ int radeon_bo_fbdev_mmap(struct radeon_bo *bo, return ttm_fbdev_mmap(vma, &bo->tbo); } -static int radeon_bo_get_surface_reg(struct radeon_bo *bo) +int radeon_bo_get_surface_reg(struct radeon_bo *bo) { struct radeon_device *rdev = bo->rdev; struct radeon_surface_reg *reg; diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index e9da130..f6b69c2 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -175,5 +175,5 @@ extern int radeon_bo_check_tiling(struct radeon_bo *bo, bool has_moved, extern void radeon_bo_move_notify(struct ttm_buffer_object *bo, struct ttm_mem_reg *mem); extern void radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo); - +extern int radeon_bo_get_surface_reg(struct radeon_bo *bo); #endif diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index bdb46c8..d2ed896 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -197,16 +197,19 @@ static int radeon_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, return 0; } -static uint32_t radeon_evict_flags(struct ttm_buffer_object *bo) +static void radeon_evict_flags(struct ttm_buffer_object *bo, + struct ttm_placement *placement) { - uint32_t cur_placement = bo->mem.placement & ~TTM_PL_MASK_MEMTYPE; - + struct radeon_bo *rbo = container_of(bo, struct radeon_bo, tbo); switch (bo->mem.mem_type) { + case TTM_PL_VRAM: + radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_GTT); + break; + case TTM_PL_TT: default: - return (cur_placement & ~TTM_PL_MASK_CACHING) | - TTM_PL_FLAG_SYSTEM | - TTM_PL_FLAG_CACHED; + radeon_ttm_placement_from_domain(rbo, RADEON_GEM_DOMAIN_CPU); } + *placement = rbo->placement; } static int radeon_verify_access(struct ttm_buffer_object *bo, struct file *filp) @@ -283,14 +286,21 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo, struct radeon_device *rdev; struct ttm_mem_reg *old_mem = &bo->mem; struct ttm_mem_reg tmp_mem; - uint32_t proposed_placement; + u32 placements; + struct ttm_placement placement; int r; rdev = radeon_get_rdev(bo->bdev); tmp_mem = *new_mem; tmp_mem.mm_node = NULL; - proposed_placement = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING; - r = ttm_bo_mem_space(bo, proposed_placement, &tmp_mem, + placement.fpfn = 0; + placement.lpfn = 0; + placement.num_placement = 1; + placement.placement = &placements; + placement.num_busy_placement = 1; + placement.busy_placement = &placements; + placements = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; + r = ttm_bo_mem_space(bo, &placement, &tmp_mem, interruptible, no_wait); if (unlikely(r)) { return r; @@ -329,15 +339,21 @@ static int radeon_move_ram_vram(struct ttm_buffer_object *bo, struct radeon_device *rdev; struct ttm_mem_reg *old_mem = &bo->mem; struct ttm_mem_reg tmp_mem; - uint32_t proposed_flags; + struct ttm_placement placement; + u32 placements; int r; rdev = radeon_get_rdev(bo->bdev); tmp_mem = *new_mem; tmp_mem.mm_node = NULL; - proposed_flags = TTM_PL_FLAG_TT | TTM_PL_MASK_CACHING; - r = ttm_bo_mem_space(bo, proposed_flags, &tmp_mem, - interruptible, no_wait); + placement.fpfn = 0; + placement.lpfn = 0; + placement.num_placement = 1; + placement.placement = &placements; + placement.num_busy_placement = 1; + placement.busy_placement = &placements; + placements = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; + r = ttm_bo_mem_space(bo, &placement, &tmp_mem, interruptible, no_wait); if (unlikely(r)) { return r; } @@ -407,18 +423,6 @@ memcpy: return r; } -const uint32_t radeon_mem_prios[] = { - TTM_PL_VRAM, - TTM_PL_TT, - TTM_PL_SYSTEM, -}; - -const uint32_t radeon_busy_prios[] = { - TTM_PL_TT, - TTM_PL_VRAM, - TTM_PL_SYSTEM, -}; - static int radeon_sync_obj_wait(void *sync_obj, void *sync_arg, bool lazy, bool interruptible) { @@ -446,10 +450,6 @@ static bool radeon_sync_obj_signaled(void *sync_obj, void *sync_arg) } static struct ttm_bo_driver radeon_bo_driver = { - .mem_type_prio = radeon_mem_prios, - .mem_busy_prio = radeon_busy_prios, - .num_mem_type_prio = ARRAY_SIZE(radeon_mem_prios), - .num_mem_busy_prio = ARRAY_SIZE(radeon_busy_prios), .create_ttm_backend_entry = &radeon_create_ttm_backend_entry, .invalidate_caches = &radeon_invalidate_caches, .init_mem_type = &radeon_init_mem_type, @@ -483,7 +483,7 @@ int radeon_ttm_init(struct radeon_device *rdev) return r; } r = ttm_bo_init_mm(&rdev->mman.bdev, TTM_PL_VRAM, - 0, rdev->mc.real_vram_size >> PAGE_SHIFT); + rdev->mc.real_vram_size >> PAGE_SHIFT); if (r) { DRM_ERROR("Failed initializing VRAM heap.\n"); return r; @@ -506,7 +506,7 @@ int radeon_ttm_init(struct radeon_device *rdev) DRM_INFO("radeon: %uM of VRAM memory ready\n", (unsigned)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); + rdev->mc.gtt_size >> PAGE_SHIFT); if (r) { DRM_ERROR("Failed initializing GTT heap.\n"); return r; diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c index eda6d757..c1fcddd 100644 --- a/drivers/gpu/drm/radeon/rs400.c +++ b/drivers/gpu/drm/radeon/rs400.c @@ -430,6 +430,8 @@ int rs400_resume(struct radeon_device *rdev) radeon_combios_asic_init(rdev->ddev); /* Resume clock after posting */ r300_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return rs400_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c index fd5ab01..4f8ea42 100644 --- a/drivers/gpu/drm/radeon/rs600.c +++ b/drivers/gpu/drm/radeon/rs600.c @@ -586,6 +586,8 @@ int rs600_resume(struct radeon_device *rdev) atom_asic_init(rdev->mode_info.atom_context); /* Resume clock after posting */ rv515_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return rs600_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/rs690.c b/drivers/gpu/drm/radeon/rs690.c index eb486ee..1e22f52 100644 --- a/drivers/gpu/drm/radeon/rs690.c +++ b/drivers/gpu/drm/radeon/rs690.c @@ -260,8 +260,9 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, 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); + a.full = rfixed_div(b, c); + request_fifo_depth.full = rfixed_mul(a, wm->num_line_pair); + request_fifo_depth.full = rfixed_ceil(request_fifo_depth); if (a.full < rfixed_const(4)) { wm->lb_request_fifo_depth = 4; } else { @@ -390,6 +391,7 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, 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); + wm->priority_mark_max.full = rfixed_ceil(wm->priority_mark_max); /* Determine estimated width */ estimated_width.full = tolerable_latency.full - wm->worst_case_latency.full; @@ -399,6 +401,7 @@ void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, } else { a.full = rfixed_const(16); wm->priority_mark.full = rfixed_div(estimated_width, a); + wm->priority_mark.full = rfixed_ceil(wm->priority_mark); wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full; } } @@ -655,6 +658,8 @@ int rs690_resume(struct radeon_device *rdev) atom_asic_init(rdev->mode_info.atom_context); /* Resume clock after posting */ rv515_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return rs690_startup(rdev); } diff --git a/drivers/gpu/drm/radeon/rv515.c b/drivers/gpu/drm/radeon/rv515.c index 7793239..59632a5 100644 --- a/drivers/gpu/drm/radeon/rv515.c +++ b/drivers/gpu/drm/radeon/rv515.c @@ -513,6 +513,8 @@ int rv515_resume(struct radeon_device *rdev) atom_asic_init(rdev->mode_info.atom_context); /* Resume clock after posting */ rv515_clock_startup(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); return rv515_startup(rdev); } @@ -889,8 +891,9 @@ void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, 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); + a.full = rfixed_div(b, c); + request_fifo_depth.full = rfixed_mul(a, wm->num_line_pair); + request_fifo_depth.full = rfixed_ceil(request_fifo_depth); if (a.full < rfixed_const(4)) { wm->lb_request_fifo_depth = 4; } else { @@ -992,15 +995,17 @@ void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, 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); + wm->priority_mark_max.full = rfixed_ceil(wm->priority_mark_max); /* 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); + wm->priority_mark.full = wm->priority_mark_max.full; } else { a.full = rfixed_const(16); wm->priority_mark.full = rfixed_div(estimated_width, a); + wm->priority_mark.full = rfixed_ceil(wm->priority_mark); wm->priority_mark.full = wm->priority_mark_max.full - wm->priority_mark.full; } } diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index dd4f020..2d124bb 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -874,6 +874,14 @@ static int rv770_startup(struct radeon_device *rdev) { int r; + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { + r = r600_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + rv770_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { rv770_agp_enable(rdev); @@ -1039,25 +1047,17 @@ int rv770_init(struct radeon_device *rdev) rdev->ih.ring_obj = NULL; r600_ih_ring_init(rdev, 64 * 1024); - if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { - r = r600_init_microcode(rdev); - if (r) { - DRM_ERROR("Failed to load firmware!\n"); - return r; - } - } - r = r600_pcie_gart_init(rdev); if (r) return r; - rdev->accel_working = true; r = r600_blit_init(rdev); if (r) { - DRM_ERROR("radeon: failled blitter (%d).\n", r); - rdev->accel_working = false; + DRM_ERROR("radeon: failed blitter (%d).\n", r); + return r; } + rdev->accel_working = true; r = rv770_startup(rdev); if (r) { rv770_suspend(rdev); @@ -1069,12 +1069,12 @@ int rv770_init(struct radeon_device *rdev) if (rdev->accel_working) { r = radeon_ib_pool_init(rdev); if (r) { - DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r); + DRM_ERROR("radeon: failed initializing IB pool (%d).\n", r); rdev->accel_working = false; } r = r600_ib_test(rdev); if (r) { - DRM_ERROR("radeon: failled testing IB (%d).\n", r); + DRM_ERROR("radeon: failed testing IB (%d).\n", r); rdev->accel_working = false; } } diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index e13fd23..a835b6f 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -27,6 +27,14 @@ /* * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com> */ +/* Notes: + * + * We store bo pointer in drm_mm_node struct so we know which bo own a + * specific node. There is no protection on the pointer, thus to make + * sure things don't go berserk you have to access this pointer while + * holding the global lru lock and make sure anytime you free a node you + * reset the pointer to NULL. + */ #include "ttm/ttm_module.h" #include "ttm/ttm_bo_driver.h" @@ -51,6 +59,60 @@ static struct attribute ttm_bo_count = { .mode = S_IRUGO }; +static inline int ttm_mem_type_from_flags(uint32_t flags, uint32_t *mem_type) +{ + int i; + + for (i = 0; i <= TTM_PL_PRIV5; i++) + if (flags & (1 << i)) { + *mem_type = i; + return 0; + } + return -EINVAL; +} + +static void ttm_mem_type_manager_debug(struct ttm_bo_global *glob, + struct ttm_mem_type_manager *man) +{ + printk(KERN_ERR TTM_PFX " has_type: %d\n", man->has_type); + printk(KERN_ERR TTM_PFX " use_type: %d\n", man->use_type); + printk(KERN_ERR TTM_PFX " flags: 0x%08X\n", man->flags); + printk(KERN_ERR TTM_PFX " gpu_offset: 0x%08lX\n", man->gpu_offset); + printk(KERN_ERR TTM_PFX " io_offset: 0x%08lX\n", man->io_offset); + printk(KERN_ERR TTM_PFX " io_size: %ld\n", man->io_size); + printk(KERN_ERR TTM_PFX " size: %ld\n", (unsigned long)man->size); + printk(KERN_ERR TTM_PFX " available_caching: 0x%08X\n", + man->available_caching); + printk(KERN_ERR TTM_PFX " default_caching: 0x%08X\n", + man->default_caching); + spin_lock(&glob->lru_lock); + drm_mm_debug_table(&man->manager, TTM_PFX); + spin_unlock(&glob->lru_lock); +} + +static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo, + struct ttm_placement *placement) +{ + struct ttm_bo_device *bdev = bo->bdev; + struct ttm_bo_global *glob = bo->glob; + struct ttm_mem_type_manager *man; + int i, ret, mem_type; + + printk(KERN_ERR TTM_PFX "No space for %p (%ld pages, %ldK, %ldM)\n", + bo, bo->mem.num_pages, bo->mem.size >> 10, + bo->mem.size >> 20); + for (i = 0; i < placement->num_placement; i++) { + ret = ttm_mem_type_from_flags(placement->placement[i], + &mem_type); + if (ret) + return; + man = &bdev->man[mem_type]; + printk(KERN_ERR TTM_PFX " placement[%d]=0x%08X (%d)\n", + i, placement->placement[i], mem_type); + ttm_mem_type_manager_debug(glob, man); + } +} + static ssize_t ttm_bo_global_show(struct kobject *kobj, struct attribute *attr, char *buffer) @@ -117,7 +179,7 @@ int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo, bool interruptible) ret = wait_event_interruptible(bo->event_queue, atomic_read(&bo->reserved) == 0); if (unlikely(ret != 0)) - return -ERESTART; + return ret; } else { wait_event(bo->event_queue, atomic_read(&bo->reserved) == 0); } @@ -247,7 +309,6 @@ EXPORT_SYMBOL(ttm_bo_unreserve); /* * Call bo->mutex locked. */ - static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc) { struct ttm_bo_device *bdev = bo->bdev; @@ -329,14 +390,8 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo, } if (bo->mem.mem_type == TTM_PL_SYSTEM) { - - struct ttm_mem_reg *old_mem = &bo->mem; - uint32_t save_flags = old_mem->placement; - - *old_mem = *mem; + bo->mem = *mem; mem->mm_node = NULL; - ttm_flag_masked(&save_flags, mem->placement, - TTM_PL_MASK_MEMTYPE); goto moved; } @@ -419,6 +474,7 @@ static int ttm_bo_cleanup_refs(struct ttm_buffer_object *bo, bool remove_all) kref_put(&bo->list_kref, ttm_bo_ref_bug); } if (bo->mem.mm_node) { + bo->mem.mm_node->private = NULL; drm_mm_put_block(bo->mem.mm_node); bo->mem.mm_node = NULL; } @@ -555,24 +611,21 @@ void ttm_bo_unref(struct ttm_buffer_object **p_bo) } EXPORT_SYMBOL(ttm_bo_unref); -static int ttm_bo_evict(struct ttm_buffer_object *bo, unsigned mem_type, - bool interruptible, bool no_wait) +static int ttm_bo_evict(struct ttm_buffer_object *bo, bool interruptible, + bool no_wait) { - int ret = 0; struct ttm_bo_device *bdev = bo->bdev; struct ttm_bo_global *glob = bo->glob; struct ttm_mem_reg evict_mem; - uint32_t proposed_placement; - - if (bo->mem.mem_type != mem_type) - goto out; + struct ttm_placement placement; + int ret = 0; spin_lock(&bo->lock); ret = ttm_bo_wait(bo, false, interruptible, no_wait); spin_unlock(&bo->lock); if (unlikely(ret != 0)) { - if (ret != -ERESTART) { + if (ret != -ERESTARTSYS) { printk(KERN_ERR TTM_PFX "Failed to expire sync object before " "buffer eviction.\n"); @@ -585,116 +638,139 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo, unsigned mem_type, evict_mem = bo->mem; evict_mem.mm_node = NULL; - proposed_placement = bdev->driver->evict_flags(bo); - - ret = ttm_bo_mem_space(bo, proposed_placement, - &evict_mem, interruptible, no_wait); - if (unlikely(ret != 0 && ret != -ERESTART)) - ret = ttm_bo_mem_space(bo, TTM_PL_FLAG_SYSTEM, - &evict_mem, interruptible, no_wait); - + placement.fpfn = 0; + placement.lpfn = 0; + placement.num_placement = 0; + placement.num_busy_placement = 0; + bdev->driver->evict_flags(bo, &placement); + ret = ttm_bo_mem_space(bo, &placement, &evict_mem, interruptible, + no_wait); if (ret) { - if (ret != -ERESTART) + if (ret != -ERESTARTSYS) { printk(KERN_ERR TTM_PFX "Failed to find memory space for " "buffer 0x%p eviction.\n", bo); + ttm_bo_mem_space_debug(bo, &placement); + } goto out; } ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, interruptible, no_wait); if (ret) { - if (ret != -ERESTART) + if (ret != -ERESTARTSYS) printk(KERN_ERR TTM_PFX "Buffer eviction failed\n"); + spin_lock(&glob->lru_lock); + if (evict_mem.mm_node) { + evict_mem.mm_node->private = NULL; + drm_mm_put_block(evict_mem.mm_node); + evict_mem.mm_node = NULL; + } + spin_unlock(&glob->lru_lock); goto out; } + bo->evicted = true; +out: + return ret; +} + +static int ttm_mem_evict_first(struct ttm_bo_device *bdev, + uint32_t mem_type, + bool interruptible, bool no_wait) +{ + struct ttm_bo_global *glob = bdev->glob; + struct ttm_mem_type_manager *man = &bdev->man[mem_type]; + struct ttm_buffer_object *bo; + int ret, put_count = 0; spin_lock(&glob->lru_lock); - if (evict_mem.mm_node) { - drm_mm_put_block(evict_mem.mm_node); - evict_mem.mm_node = NULL; - } + bo = list_first_entry(&man->lru, struct ttm_buffer_object, lru); + kref_get(&bo->list_kref); + ret = ttm_bo_reserve_locked(bo, interruptible, no_wait, false, 0); + if (likely(ret == 0)) + put_count = ttm_bo_del_from_lru(bo); spin_unlock(&glob->lru_lock); - bo->evicted = true; -out: + if (unlikely(ret != 0)) + return ret; + while (put_count--) + kref_put(&bo->list_kref, ttm_bo_ref_bug); + ret = ttm_bo_evict(bo, interruptible, no_wait); + ttm_bo_unreserve(bo); + kref_put(&bo->list_kref, ttm_bo_release_list); return ret; } +static int ttm_bo_man_get_node(struct ttm_buffer_object *bo, + struct ttm_mem_type_manager *man, + struct ttm_placement *placement, + struct ttm_mem_reg *mem, + struct drm_mm_node **node) +{ + struct ttm_bo_global *glob = bo->glob; + unsigned long lpfn; + int ret; + + lpfn = placement->lpfn; + if (!lpfn) + lpfn = man->size; + *node = NULL; + do { + ret = drm_mm_pre_get(&man->manager); + if (unlikely(ret)) + return ret; + + spin_lock(&glob->lru_lock); + *node = drm_mm_search_free_in_range(&man->manager, + mem->num_pages, mem->page_alignment, + placement->fpfn, lpfn, 1); + if (unlikely(*node == NULL)) { + spin_unlock(&glob->lru_lock); + return 0; + } + *node = drm_mm_get_block_atomic_range(*node, mem->num_pages, + mem->page_alignment, + placement->fpfn, + lpfn); + spin_unlock(&glob->lru_lock); + } while (*node == NULL); + return 0; +} + /** * Repeatedly evict memory from the LRU for @mem_type until we create enough * space, or we've evicted everything and there isn't enough space. */ -static int ttm_bo_mem_force_space(struct ttm_bo_device *bdev, - struct ttm_mem_reg *mem, - uint32_t mem_type, - bool interruptible, bool no_wait) +static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo, + uint32_t mem_type, + struct ttm_placement *placement, + struct ttm_mem_reg *mem, + bool interruptible, bool no_wait) { + struct ttm_bo_device *bdev = bo->bdev; struct ttm_bo_global *glob = bdev->glob; - struct drm_mm_node *node; - struct ttm_buffer_object *entry; struct ttm_mem_type_manager *man = &bdev->man[mem_type]; - struct list_head *lru; - unsigned long num_pages = mem->num_pages; - int put_count = 0; + struct drm_mm_node *node; int ret; -retry_pre_get: - ret = drm_mm_pre_get(&man->manager); - if (unlikely(ret != 0)) - return ret; - - spin_lock(&glob->lru_lock); do { - node = drm_mm_search_free(&man->manager, num_pages, - mem->page_alignment, 1); + ret = ttm_bo_man_get_node(bo, man, placement, mem, &node); + if (unlikely(ret != 0)) + return ret; if (node) break; - - lru = &man->lru; - if (list_empty(lru)) + spin_lock(&glob->lru_lock); + if (list_empty(&man->lru)) { + spin_unlock(&glob->lru_lock); break; - - entry = list_first_entry(lru, struct ttm_buffer_object, lru); - kref_get(&entry->list_kref); - - ret = - ttm_bo_reserve_locked(entry, interruptible, no_wait, - false, 0); - - if (likely(ret == 0)) - put_count = ttm_bo_del_from_lru(entry); - + } spin_unlock(&glob->lru_lock); - + ret = ttm_mem_evict_first(bdev, mem_type, interruptible, + no_wait); if (unlikely(ret != 0)) return ret; - - while (put_count--) - kref_put(&entry->list_kref, ttm_bo_ref_bug); - - ret = ttm_bo_evict(entry, mem_type, interruptible, no_wait); - - ttm_bo_unreserve(entry); - - kref_put(&entry->list_kref, ttm_bo_release_list); - if (ret) - return ret; - - spin_lock(&glob->lru_lock); } while (1); - - if (!node) { - spin_unlock(&glob->lru_lock); + if (node == NULL) return -ENOMEM; - } - - node = drm_mm_get_block_atomic(node, num_pages, mem->page_alignment); - if (unlikely(!node)) { - spin_unlock(&glob->lru_lock); - goto retry_pre_get; - } - - spin_unlock(&glob->lru_lock); mem->mm_node = node; mem->mem_type = mem_type; return 0; @@ -725,7 +801,6 @@ static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man, return result; } - static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man, bool disallow_fixed, uint32_t mem_type, @@ -758,66 +833,55 @@ static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man, * space. */ int ttm_bo_mem_space(struct ttm_buffer_object *bo, - uint32_t proposed_placement, - struct ttm_mem_reg *mem, - bool interruptible, bool no_wait) + struct ttm_placement *placement, + struct ttm_mem_reg *mem, + bool interruptible, bool no_wait) { struct ttm_bo_device *bdev = bo->bdev; - struct ttm_bo_global *glob = bo->glob; struct ttm_mem_type_manager *man; - - uint32_t num_prios = bdev->driver->num_mem_type_prio; - const uint32_t *prios = bdev->driver->mem_type_prio; - uint32_t i; uint32_t mem_type = TTM_PL_SYSTEM; uint32_t cur_flags = 0; bool type_found = false; bool type_ok = false; - bool has_eagain = false; + bool has_erestartsys = false; struct drm_mm_node *node = NULL; - int ret; + int i, ret; mem->mm_node = NULL; - for (i = 0; i < num_prios; ++i) { - mem_type = prios[i]; + for (i = 0; i <= placement->num_placement; ++i) { + ret = ttm_mem_type_from_flags(placement->placement[i], + &mem_type); + if (ret) + return ret; man = &bdev->man[mem_type]; type_ok = ttm_bo_mt_compatible(man, - bo->type == ttm_bo_type_user, - mem_type, proposed_placement, - &cur_flags); + bo->type == ttm_bo_type_user, + mem_type, + placement->placement[i], + &cur_flags); if (!type_ok) continue; cur_flags = ttm_bo_select_caching(man, bo->mem.placement, cur_flags); + /* + * Use the access and other non-mapping-related flag bits from + * the memory placement flags to the current flags + */ + ttm_flag_masked(&cur_flags, placement->placement[i], + ~TTM_PL_MASK_MEMTYPE); if (mem_type == TTM_PL_SYSTEM) break; if (man->has_type && man->use_type) { type_found = true; - do { - ret = drm_mm_pre_get(&man->manager); - if (unlikely(ret)) - return ret; - - spin_lock(&glob->lru_lock); - node = drm_mm_search_free(&man->manager, - mem->num_pages, - mem->page_alignment, - 1); - if (unlikely(!node)) { - spin_unlock(&glob->lru_lock); - break; - } - node = drm_mm_get_block_atomic(node, - mem->num_pages, - mem-> - page_alignment); - spin_unlock(&glob->lru_lock); - } while (!node); + ret = ttm_bo_man_get_node(bo, man, placement, mem, + &node); + if (unlikely(ret)) + return ret; } if (node) break; @@ -827,67 +891,65 @@ int ttm_bo_mem_space(struct ttm_buffer_object *bo, mem->mm_node = node; mem->mem_type = mem_type; mem->placement = cur_flags; + if (node) + node->private = bo; return 0; } if (!type_found) return -EINVAL; - num_prios = bdev->driver->num_mem_busy_prio; - prios = bdev->driver->mem_busy_prio; - - for (i = 0; i < num_prios; ++i) { - mem_type = prios[i]; + for (i = 0; i <= placement->num_busy_placement; ++i) { + ret = ttm_mem_type_from_flags(placement->placement[i], + &mem_type); + if (ret) + return ret; man = &bdev->man[mem_type]; - if (!man->has_type) continue; - if (!ttm_bo_mt_compatible(man, - bo->type == ttm_bo_type_user, - mem_type, - proposed_placement, &cur_flags)) + bo->type == ttm_bo_type_user, + mem_type, + placement->placement[i], + &cur_flags)) continue; cur_flags = ttm_bo_select_caching(man, bo->mem.placement, cur_flags); + /* + * Use the access and other non-mapping-related flag bits from + * the memory placement flags to the current flags + */ + ttm_flag_masked(&cur_flags, placement->placement[i], + ~TTM_PL_MASK_MEMTYPE); - ret = ttm_bo_mem_force_space(bdev, mem, mem_type, - interruptible, no_wait); - + ret = ttm_bo_mem_force_space(bo, mem_type, placement, mem, + interruptible, no_wait); if (ret == 0 && mem->mm_node) { mem->placement = cur_flags; + mem->mm_node->private = bo; return 0; } - - if (ret == -ERESTART) - has_eagain = true; + if (ret == -ERESTARTSYS) + has_erestartsys = true; } - - ret = (has_eagain) ? -ERESTART : -ENOMEM; + ret = (has_erestartsys) ? -ERESTARTSYS : -ENOMEM; return ret; } EXPORT_SYMBOL(ttm_bo_mem_space); int ttm_bo_wait_cpu(struct ttm_buffer_object *bo, bool no_wait) { - int ret = 0; - if ((atomic_read(&bo->cpu_writers) > 0) && no_wait) return -EBUSY; - ret = wait_event_interruptible(bo->event_queue, - atomic_read(&bo->cpu_writers) == 0); - - if (ret == -ERESTARTSYS) - ret = -ERESTART; - - return ret; + return wait_event_interruptible(bo->event_queue, + atomic_read(&bo->cpu_writers) == 0); } int ttm_bo_move_buffer(struct ttm_buffer_object *bo, - uint32_t proposed_placement, - bool interruptible, bool no_wait) + struct ttm_placement *placement, + bool interruptible, bool no_wait) { struct ttm_bo_global *glob = bo->glob; int ret = 0; @@ -900,101 +962,82 @@ int ttm_bo_move_buffer(struct ttm_buffer_object *bo, * Have the driver move function wait for idle when necessary, * instead of doing it here. */ - spin_lock(&bo->lock); ret = ttm_bo_wait(bo, false, interruptible, no_wait); spin_unlock(&bo->lock); - if (ret) return ret; - mem.num_pages = bo->num_pages; mem.size = mem.num_pages << PAGE_SHIFT; mem.page_alignment = bo->mem.page_alignment; - /* * Determine where to move the buffer. */ - - ret = ttm_bo_mem_space(bo, proposed_placement, &mem, - interruptible, no_wait); + ret = ttm_bo_mem_space(bo, placement, &mem, interruptible, no_wait); if (ret) goto out_unlock; - ret = ttm_bo_handle_move_mem(bo, &mem, false, interruptible, no_wait); - out_unlock: if (ret && mem.mm_node) { spin_lock(&glob->lru_lock); + mem.mm_node->private = NULL; drm_mm_put_block(mem.mm_node); spin_unlock(&glob->lru_lock); } return ret; } -static int ttm_bo_mem_compat(uint32_t proposed_placement, +static int ttm_bo_mem_compat(struct ttm_placement *placement, struct ttm_mem_reg *mem) { - if ((proposed_placement & mem->placement & TTM_PL_MASK_MEM) == 0) - return 0; - if ((proposed_placement & mem->placement & TTM_PL_MASK_CACHING) == 0) - return 0; - - return 1; + int i; + + for (i = 0; i < placement->num_placement; i++) { + if ((placement->placement[i] & mem->placement & + TTM_PL_MASK_CACHING) && + (placement->placement[i] & mem->placement & + TTM_PL_MASK_MEM)) + return i; + } + return -1; } int ttm_buffer_object_validate(struct ttm_buffer_object *bo, - uint32_t proposed_placement, - bool interruptible, bool no_wait) + struct ttm_placement *placement, + bool interruptible, bool no_wait) { int ret; BUG_ON(!atomic_read(&bo->reserved)); - bo->proposed_placement = proposed_placement; - - TTM_DEBUG("Proposed placement 0x%08lx, Old flags 0x%08lx\n", - (unsigned long)proposed_placement, - (unsigned long)bo->mem.placement); - + /* Check that range is valid */ + if (placement->lpfn || placement->fpfn) + if (placement->fpfn > placement->lpfn || + (placement->lpfn - placement->fpfn) < bo->num_pages) + return -EINVAL; /* * Check whether we need to move buffer. */ - - if (!ttm_bo_mem_compat(bo->proposed_placement, &bo->mem)) { - ret = ttm_bo_move_buffer(bo, bo->proposed_placement, - interruptible, no_wait); - if (ret) { - if (ret != -ERESTART) - printk(KERN_ERR TTM_PFX - "Failed moving buffer. " - "Proposed placement 0x%08x\n", - bo->proposed_placement); - if (ret == -ENOMEM) - printk(KERN_ERR TTM_PFX - "Out of aperture space or " - "DRM memory quota.\n"); + ret = ttm_bo_mem_compat(placement, &bo->mem); + if (ret < 0) { + ret = ttm_bo_move_buffer(bo, placement, interruptible, no_wait); + if (ret) return ret; - } + } else { + /* + * Use the access and other non-mapping-related flag bits from + * the compatible memory placement flags to the active flags + */ + ttm_flag_masked(&bo->mem.placement, placement->placement[ret], + ~TTM_PL_MASK_MEMTYPE); } - /* * We might need to add a TTM. */ - if (bo->mem.mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) { ret = ttm_bo_add_ttm(bo, true); if (ret) return ret; } - /* - * Validation has succeeded, move the access and other - * non-mapping-related flag bits from the proposed flags to - * the active flags - */ - - ttm_flag_masked(&bo->mem.placement, bo->proposed_placement, - ~TTM_PL_MASK_MEMTYPE); - return 0; } EXPORT_SYMBOL(ttm_buffer_object_validate); @@ -1042,8 +1085,10 @@ int ttm_buffer_object_init(struct ttm_bo_device *bdev, size_t acc_size, void (*destroy) (struct ttm_buffer_object *)) { - int ret = 0; + int i, c, ret = 0; unsigned long num_pages; + uint32_t placements[8]; + struct ttm_placement placement; size += buffer_start & ~PAGE_MASK; num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; @@ -1100,7 +1145,16 @@ int ttm_buffer_object_init(struct ttm_bo_device *bdev, goto out_err; } - ret = ttm_buffer_object_validate(bo, flags, interruptible, false); + placement.fpfn = 0; + placement.lpfn = 0; + for (i = 0, c = 0; i <= TTM_PL_PRIV5; i++) + if (flags & (1 << i)) + placements[c++] = (flags & ~TTM_PL_MASK_MEM) | (1 << i); + placement.placement = placements; + placement.num_placement = c; + placement.busy_placement = placements; + placement.num_busy_placement = c; + ret = ttm_buffer_object_validate(bo, &placement, interruptible, false); if (ret) goto out_err; @@ -1135,8 +1189,8 @@ int ttm_buffer_object_create(struct ttm_bo_device *bdev, struct ttm_buffer_object **p_bo) { struct ttm_buffer_object *bo; - int ret; struct ttm_mem_global *mem_glob = bdev->glob->mem_glob; + int ret; size_t acc_size = ttm_bo_size(bdev->glob, (size + PAGE_SIZE - 1) >> PAGE_SHIFT); @@ -1161,66 +1215,32 @@ int ttm_buffer_object_create(struct ttm_bo_device *bdev, return ret; } -static int ttm_bo_leave_list(struct ttm_buffer_object *bo, - uint32_t mem_type, bool allow_errors) -{ - int ret; - - spin_lock(&bo->lock); - ret = ttm_bo_wait(bo, false, false, false); - spin_unlock(&bo->lock); - - if (ret && allow_errors) - goto out; - - if (bo->mem.mem_type == mem_type) - ret = ttm_bo_evict(bo, mem_type, false, false); - - if (ret) { - if (allow_errors) { - goto out; - } else { - ret = 0; - printk(KERN_ERR TTM_PFX "Cleanup eviction failed\n"); - } - } - -out: - return ret; -} - static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev, - struct list_head *head, - unsigned mem_type, bool allow_errors) + unsigned mem_type, bool allow_errors) { + struct ttm_mem_type_manager *man = &bdev->man[mem_type]; struct ttm_bo_global *glob = bdev->glob; - struct ttm_buffer_object *entry; int ret; - int put_count; /* * Can't use standard list traversal since we're unlocking. */ spin_lock(&glob->lru_lock); - - while (!list_empty(head)) { - entry = list_first_entry(head, struct ttm_buffer_object, lru); - kref_get(&entry->list_kref); - ret = ttm_bo_reserve_locked(entry, false, false, false, 0); - put_count = ttm_bo_del_from_lru(entry); + while (!list_empty(&man->lru)) { spin_unlock(&glob->lru_lock); - while (put_count--) - kref_put(&entry->list_kref, ttm_bo_ref_bug); - BUG_ON(ret); - ret = ttm_bo_leave_list(entry, mem_type, allow_errors); - ttm_bo_unreserve(entry); - kref_put(&entry->list_kref, ttm_bo_release_list); + ret = ttm_mem_evict_first(bdev, mem_type, false, false); + if (ret) { + if (allow_errors) { + return ret; + } else { + printk(KERN_ERR TTM_PFX + "Cleanup eviction failed\n"); + } + } spin_lock(&glob->lru_lock); } - spin_unlock(&glob->lru_lock); - return 0; } @@ -1247,7 +1267,7 @@ int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type) ret = 0; if (mem_type > 0) { - ttm_bo_force_list_clean(bdev, &man->lru, mem_type, false); + ttm_bo_force_list_clean(bdev, mem_type, false); spin_lock(&glob->lru_lock); if (drm_mm_clean(&man->manager)) @@ -1280,12 +1300,12 @@ int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type) return 0; } - return ttm_bo_force_list_clean(bdev, &man->lru, mem_type, true); + return ttm_bo_force_list_clean(bdev, mem_type, true); } EXPORT_SYMBOL(ttm_bo_evict_mm); int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type, - unsigned long p_offset, unsigned long p_size) + unsigned long p_size) { int ret = -EINVAL; struct ttm_mem_type_manager *man; @@ -1315,7 +1335,7 @@ int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type, type); return ret; } - ret = drm_mm_init(&man->manager, p_offset, p_size); + ret = drm_mm_init(&man->manager, 0, p_size); if (ret) return ret; } @@ -1464,7 +1484,7 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev, * Initialize the system memory buffer type. * Other types need to be driver / IOCTL initialized. */ - ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0, 0); + ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0); if (unlikely(ret != 0)) goto out_no_sys; @@ -1694,7 +1714,7 @@ int ttm_bo_block_reservation(struct ttm_buffer_object *bo, bool interruptible, ret = wait_event_interruptible (bo->event_queue, atomic_read(&bo->reserved) == 0); if (unlikely(ret != 0)) - return -ERESTART; + return ret; } else { wait_event(bo->event_queue, atomic_read(&bo->reserved) == 0); diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c index 1c040d0..609a85a 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_vm.c +++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c @@ -114,7 +114,7 @@ static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) ret = ttm_bo_wait(bo, false, true, false); spin_unlock(&bo->lock); if (unlikely(ret != 0)) { - retval = (ret != -ERESTART) ? + retval = (ret != -ERESTARTSYS) ? VM_FAULT_SIGBUS : VM_FAULT_NOPAGE; goto out_unlock; } @@ -349,9 +349,6 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp, switch (ret) { case 0: break; - case -ERESTART: - ret = -EINTR; - goto out_unref; case -EBUSY: ret = -EAGAIN; goto out_unref; @@ -421,8 +418,6 @@ ssize_t ttm_bo_fbdev_io(struct ttm_buffer_object *bo, const char __user *wbuf, switch (ret) { case 0: break; - case -ERESTART: - return -EINTR; case -EBUSY: return -EAGAIN; default: diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c index 8bfde5f..f5245c0 100644 --- a/drivers/gpu/drm/ttm/ttm_memory.c +++ b/drivers/gpu/drm/ttm/ttm_memory.c @@ -323,8 +323,10 @@ static int ttm_mem_init_dma32_zone(struct ttm_mem_global *glob, * No special dma32 zone needed. */ - if (mem <= ((uint64_t) 1ULL << 32)) + if (mem <= ((uint64_t) 1ULL << 32)) { + kfree(zone); return 0; + } /* * Limit max dma32 memory to 4GB for now diff --git a/include/drm/drm_mm.h b/include/drm/drm_mm.h index 62329f9..4c10be3 100644 --- a/include/drm/drm_mm.h +++ b/include/drm/drm_mm.h @@ -66,6 +66,13 @@ extern struct drm_mm_node *drm_mm_get_block_generic(struct drm_mm_node *node, unsigned long size, unsigned alignment, int atomic); +extern struct drm_mm_node *drm_mm_get_block_range_generic( + struct drm_mm_node *node, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end, + int atomic); static inline struct drm_mm_node *drm_mm_get_block(struct drm_mm_node *parent, unsigned long size, unsigned alignment) @@ -78,11 +85,38 @@ static inline struct drm_mm_node *drm_mm_get_block_atomic(struct drm_mm_node *pa { return drm_mm_get_block_generic(parent, size, alignment, 1); } +static inline struct drm_mm_node *drm_mm_get_block_range( + struct drm_mm_node *parent, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end) +{ + return drm_mm_get_block_range_generic(parent, size, alignment, + start, end, 0); +} +static inline struct drm_mm_node *drm_mm_get_block_atomic_range( + struct drm_mm_node *parent, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end) +{ + return drm_mm_get_block_range_generic(parent, size, alignment, + start, end, 1); +} extern void drm_mm_put_block(struct drm_mm_node *cur); extern struct drm_mm_node *drm_mm_search_free(const struct drm_mm *mm, unsigned long size, unsigned alignment, int best_match); +extern struct drm_mm_node *drm_mm_search_free_in_range( + const struct drm_mm *mm, + unsigned long size, + unsigned alignment, + unsigned long start, + unsigned long end, + int best_match); extern int drm_mm_init(struct drm_mm *mm, unsigned long start, unsigned long size); extern void drm_mm_takedown(struct drm_mm *mm); @@ -99,6 +133,7 @@ static inline struct drm_mm *drm_get_mm(struct drm_mm_node *block) return block->mm; } +extern void drm_mm_debug_table(struct drm_mm *mm, const char *prefix); #ifdef CONFIG_DEBUG_FS int drm_mm_dump_table(struct seq_file *m, struct drm_mm *mm); #endif diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h index 4911461..4fd4985 100644 --- a/include/drm/ttm/ttm_bo_api.h +++ b/include/drm/ttm/ttm_bo_api.h @@ -44,6 +44,29 @@ struct ttm_bo_device; struct drm_mm_node; + +/** + * struct ttm_placement + * + * @fpfn: first valid page frame number to put the object + * @lpfn: last valid page frame number to put the object + * @num_placement: number of prefered placements + * @placement: prefered placements + * @num_busy_placement: number of prefered placements when need to evict buffer + * @busy_placement: prefered placements when need to evict buffer + * + * Structure indicating the placement you request for an object. + */ +struct ttm_placement { + unsigned fpfn; + unsigned lpfn; + unsigned num_placement; + const uint32_t *placement; + unsigned num_busy_placement; + const uint32_t *busy_placement; +}; + + /** * struct ttm_mem_reg * @@ -109,10 +132,6 @@ struct ttm_tt; * the object is destroyed. * @event_queue: Queue for processes waiting on buffer object status change. * @lock: spinlock protecting mostly synchronization members. - * @proposed_placement: Proposed placement for the buffer. Changed only by the - * creator prior to validation as opposed to bo->mem.proposed_flags which is - * changed by the implementation prior to a buffer move if it wants to outsmart - * the buffer creator / user. This latter happens, for example, at eviction. * @mem: structure describing current placement. * @persistant_swap_storage: Usually the swap storage is deleted for buffers * pinned in physical memory. If this behaviour is not desired, this member @@ -177,7 +196,6 @@ struct ttm_buffer_object { * Members protected by the bo::reserved lock. */ - uint32_t proposed_placement; struct ttm_mem_reg mem; struct file *persistant_swap_storage; struct ttm_tt *ttm; @@ -285,7 +303,7 @@ ttm_bo_reference(struct ttm_buffer_object *bo) * Note: It might be necessary to block validations before the * wait by reserving the buffer. * Returns -EBUSY if no_wait is true and the buffer is busy. - * Returns -ERESTART if interrupted by a signal. + * Returns -ERESTARTSYS if interrupted by a signal. */ extern int ttm_bo_wait(struct ttm_buffer_object *bo, bool lazy, bool interruptible, bool no_wait); @@ -293,21 +311,22 @@ extern int ttm_bo_wait(struct ttm_buffer_object *bo, bool lazy, * ttm_buffer_object_validate * * @bo: The buffer object. - * @proposed_placement: Proposed_placement for the buffer object. + * @placement: Proposed placement for the buffer object. * @interruptible: Sleep interruptible if sleeping. * @no_wait: Return immediately if the buffer is busy. * * Changes placement and caching policy of the buffer object - * according to bo::proposed_flags. + * according proposed placement. * Returns - * -EINVAL on invalid proposed_flags. + * -EINVAL on invalid proposed placement. * -ENOMEM on out-of-memory condition. * -EBUSY if no_wait is true and buffer busy. - * -ERESTART if interrupted by a signal. + * -ERESTARTSYS if interrupted by a signal. */ extern int ttm_buffer_object_validate(struct ttm_buffer_object *bo, - uint32_t proposed_placement, - bool interruptible, bool no_wait); + struct ttm_placement *placement, + bool interruptible, bool no_wait); + /** * ttm_bo_unref * @@ -328,7 +347,7 @@ extern void ttm_bo_unref(struct ttm_buffer_object **bo); * waiting for buffer idle. This lock is recursive. * Returns * -EBUSY if the buffer is busy and no_wait is true. - * -ERESTART if interrupted by a signal. + * -ERESTARTSYS if interrupted by a signal. */ extern int @@ -371,7 +390,7 @@ extern void ttm_bo_synccpu_write_release(struct ttm_buffer_object *bo); * Returns * -ENOMEM: Out of memory. * -EINVAL: Invalid placement flags. - * -ERESTART: Interrupted by signal while sleeping waiting for resources. + * -ERESTARTSYS: Interrupted by signal while sleeping waiting for resources. */ extern int ttm_buffer_object_init(struct ttm_bo_device *bdev, @@ -411,7 +430,7 @@ extern int ttm_buffer_object_init(struct ttm_bo_device *bdev, * Returns * -ENOMEM: Out of memory. * -EINVAL: Invalid placement flags. - * -ERESTART: Interrupted by signal while waiting for resources. + * -ERESTARTSYS: Interrupted by signal while waiting for resources. */ extern int ttm_buffer_object_create(struct ttm_bo_device *bdev, @@ -445,7 +464,6 @@ extern int ttm_bo_check_placement(struct ttm_buffer_object *bo, * * @bdev: Pointer to a ttm_bo_device struct. * @mem_type: The memory type. - * @p_offset: offset for managed area in pages. * @p_size: size managed area in pages. * * Initialize a manager for a given memory type. @@ -458,7 +476,7 @@ extern int ttm_bo_check_placement(struct ttm_buffer_object *bo, */ extern int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type, - unsigned long p_offset, unsigned long p_size); + unsigned long p_size); /** * ttm_bo_clean_mm * @@ -503,7 +521,7 @@ extern int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type); * * Returns: * -EINVAL: Invalid or uninitialized memory type. - * -ERESTART: The call was interrupted by a signal while waiting to + * -ERESTARTSYS: The call was interrupted by a signal while waiting to * evict a buffer. */ @@ -606,7 +624,7 @@ extern int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma, * be called from the fops::read and fops::write method. * Returns: * See man (2) write, man(2) read. In particular, - * the function may return -EINTR if + * the function may return -ERESTARTSYS if * interrupted by a signal. */ diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index 7a39ab9..ff7664e 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -242,12 +242,6 @@ struct ttm_mem_type_manager { /** * struct ttm_bo_driver * - * @mem_type_prio: Priority array of memory types to place a buffer object in - * if it fits without evicting buffers from any of these memory types. - * @mem_busy_prio: Priority array of memory types to place a buffer object in - * if it needs to evict buffers to make room. - * @num_mem_type_prio: Number of elements in the @mem_type_prio array. - * @num_mem_busy_prio: Number of elements in the @num_mem_busy_prio array. * @create_ttm_backend_entry: Callback to create a struct ttm_backend. * @invalidate_caches: Callback to invalidate read caches when a buffer object * has been evicted. @@ -265,11 +259,6 @@ struct ttm_mem_type_manager { */ struct ttm_bo_driver { - const uint32_t *mem_type_prio; - const uint32_t *mem_busy_prio; - uint32_t num_mem_type_prio; - uint32_t num_mem_busy_prio; - /** * struct ttm_bo_driver member create_ttm_backend_entry * @@ -306,7 +295,8 @@ struct ttm_bo_driver { * finished, they'll end up in bo->mem.flags */ - uint32_t(*evict_flags) (struct ttm_buffer_object *bo); + void(*evict_flags) (struct ttm_buffer_object *bo, + struct ttm_placement *placement); /** * struct ttm_bo_driver member move: * @@ -648,12 +638,12 @@ extern bool ttm_mem_reg_is_pci(struct ttm_bo_device *bdev, * -EBUSY: No space available (only if no_wait == 1). * -ENOMEM: Could not allocate memory for the buffer object, either due to * fragmentation or concurrent allocators. - * -ERESTART: An interruptible sleep was interrupted by a signal. + * -ERESTARTSYS: An interruptible sleep was interrupted by a signal. */ extern int ttm_bo_mem_space(struct ttm_buffer_object *bo, - uint32_t proposed_placement, - struct ttm_mem_reg *mem, - bool interruptible, bool no_wait); + struct ttm_placement *placement, + struct ttm_mem_reg *mem, + bool interruptible, bool no_wait); /** * ttm_bo_wait_for_cpu * @@ -663,7 +653,7 @@ extern int ttm_bo_mem_space(struct ttm_buffer_object *bo, * Wait until a buffer object is no longer sync'ed for CPU access. * Returns: * -EBUSY: Buffer object was sync'ed for CPU access. (only if no_wait == 1). - * -ERESTART: An interruptible sleep was interrupted by a signal. + * -ERESTARTSYS: An interruptible sleep was interrupted by a signal. */ extern int ttm_bo_wait_cpu(struct ttm_buffer_object *bo, bool no_wait); @@ -767,7 +757,7 @@ extern void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo); * -EAGAIN: The reservation may cause a deadlock. * Release all buffer reservations, wait for @bo to become unreserved and * try again. (only if use_sequence == 1). - * -ERESTART: A wait for the buffer to become unreserved was interrupted by + * -ERESTARTSYS: A wait for the buffer to become unreserved was interrupted by * a signal. Release all buffer reservations and return to user-space. */ extern int ttm_bo_reserve(struct ttm_buffer_object *bo, @@ -808,7 +798,7 @@ extern int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo, * * Returns: * -EBUSY: If no_wait == 1 and the buffer is already reserved. - * -ERESTART: If interruptible == 1 and the process received a signal + * -ERESTARTSYS: If interruptible == 1 and the process received a signal * while sleeping. */ extern int ttm_bo_block_reservation(struct ttm_buffer_object *bo, |
