X-Git-Url: https://git.karo-electronics.de/?a=blobdiff_plain;f=drivers%2Fgpu%2Fdrm%2Fradeon%2Fevergreen.c;fp=drivers%2Fgpu%2Fdrm%2Fradeon%2Fevergreen.c;h=6140ea1de45a6d1979c7dcbd048606adf8081cfc;hb=f2ec334db8d14ae3ec2e4bf8d974f75b8f772e26;hp=2f93d46ae69ad58dfb90ea5db02021a402506ae5;hpb=81e20d4d8d0317ecf1c7d193a52ab26cf74e1737;p=mv-sheeva.git diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 2f93d46ae69..6140ea1de45 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -32,26 +32,94 @@ #include "atom.h" #include "avivod.h" #include "evergreen_reg.h" +#include "evergreen_blit_shaders.h" #define EVERGREEN_PFP_UCODE_SIZE 1120 #define EVERGREEN_PM4_UCODE_SIZE 1376 static void evergreen_gpu_init(struct radeon_device *rdev); void evergreen_fini(struct radeon_device *rdev); +static void evergreen_pcie_gen2_enable(struct radeon_device *rdev); + +void evergreen_pre_page_flip(struct radeon_device *rdev, int crtc) +{ + struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc]; + u32 tmp; + + /* make sure flip is at vb rather than hb */ + tmp = RREG32(EVERGREEN_GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset); + tmp &= ~EVERGREEN_GRPH_SURFACE_UPDATE_H_RETRACE_EN; + WREG32(EVERGREEN_GRPH_FLIP_CONTROL + radeon_crtc->crtc_offset, tmp); + + /* set pageflip to happen anywhere in vblank interval */ + WREG32(EVERGREEN_MASTER_UPDATE_MODE + radeon_crtc->crtc_offset, 0); + + /* enable the pflip int */ + radeon_irq_kms_pflip_irq_get(rdev, crtc); +} + +void evergreen_post_page_flip(struct radeon_device *rdev, int crtc) +{ + /* disable the pflip int */ + radeon_irq_kms_pflip_irq_put(rdev, crtc); +} + +u32 evergreen_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base) +{ + struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id]; + u32 tmp = RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset); + + /* Lock the graphics update lock */ + tmp |= EVERGREEN_GRPH_UPDATE_LOCK; + WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp); + + /* update the scanout addresses */ + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset, + upper_32_bits(crtc_base)); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset, + (u32)crtc_base); + + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + radeon_crtc->crtc_offset, + upper_32_bits(crtc_base)); + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + radeon_crtc->crtc_offset, + (u32)crtc_base); + + /* Wait for update_pending to go high. */ + while (!(RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING)); + DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n"); + + /* Unlock the lock, so double-buffering can take place inside vblank */ + tmp &= ~EVERGREEN_GRPH_UPDATE_LOCK; + WREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset, tmp); + + /* Return current update_pending status: */ + return RREG32(EVERGREEN_GRPH_UPDATE + radeon_crtc->crtc_offset) & EVERGREEN_GRPH_SURFACE_UPDATE_PENDING; +} /* get temperature in millidegrees */ -u32 evergreen_get_temp(struct radeon_device *rdev) +int evergreen_get_temp(struct radeon_device *rdev) { u32 temp = (RREG32(CG_MULT_THERMAL_STATUS) & ASIC_T_MASK) >> ASIC_T_SHIFT; u32 actual_temp = 0; - if ((temp >> 10) & 1) - actual_temp = 0; - else if ((temp >> 9) & 1) + if (temp & 0x400) + actual_temp = -256; + else if (temp & 0x200) actual_temp = 255; - else - actual_temp = (temp >> 1) & 0xff; + else if (temp & 0x100) { + actual_temp = temp & 0x1ff; + actual_temp |= ~0x1ff; + } else + actual_temp = temp & 0xff; + + return (actual_temp * 1000) / 2; +} + +int sumo_get_temp(struct radeon_device *rdev) +{ + u32 temp = RREG32(CG_THERMAL_STATUS) & 0xff; + int actual_temp = temp - 49; return actual_temp * 1000; } @@ -284,9 +352,456 @@ void evergreen_hpd_fini(struct radeon_device *rdev) } } +/* watermark setup */ + +static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev, + struct radeon_crtc *radeon_crtc, + struct drm_display_mode *mode, + struct drm_display_mode *other_mode) +{ + u32 tmp = 0; + /* + * Line Buffer Setup + * There are 3 line buffers, each one shared by 2 display controllers. + * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between + * the display controllers. The paritioning is done via one of four + * preset allocations specified in bits 2:0: + * first display controller + * 0 - first half of lb (3840 * 2) + * 1 - first 3/4 of lb (5760 * 2) + * 2 - whole lb (7680 * 2) + * 3 - first 1/4 of lb (1920 * 2) + * second display controller + * 4 - second half of lb (3840 * 2) + * 5 - second 3/4 of lb (5760 * 2) + * 6 - whole lb (7680 * 2) + * 7 - last 1/4 of lb (1920 * 2) + */ + if (mode && other_mode) { + if (mode->hdisplay > other_mode->hdisplay) { + if (mode->hdisplay > 2560) + tmp = 1; /* 3/4 */ + else + tmp = 0; /* 1/2 */ + } else if (other_mode->hdisplay > mode->hdisplay) { + if (other_mode->hdisplay > 2560) + tmp = 3; /* 1/4 */ + else + tmp = 0; /* 1/2 */ + } else + tmp = 0; /* 1/2 */ + } else if (mode) + tmp = 2; /* whole */ + else if (other_mode) + tmp = 3; /* 1/4 */ + + /* second controller of the pair uses second half of the lb */ + if (radeon_crtc->crtc_id % 2) + tmp += 4; + WREG32(DC_LB_MEMORY_SPLIT + radeon_crtc->crtc_offset, tmp); + + switch (tmp) { + case 0: + case 4: + default: + if (ASIC_IS_DCE5(rdev)) + return 4096 * 2; + else + return 3840 * 2; + case 1: + case 5: + if (ASIC_IS_DCE5(rdev)) + return 6144 * 2; + else + return 5760 * 2; + case 2: + case 6: + if (ASIC_IS_DCE5(rdev)) + return 8192 * 2; + else + return 7680 * 2; + case 3: + case 7: + if (ASIC_IS_DCE5(rdev)) + return 2048 * 2; + else + return 1920 * 2; + } +} + +static u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev) +{ + u32 tmp = RREG32(MC_SHARED_CHMAP); + + switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { + case 0: + default: + return 1; + case 1: + return 2; + case 2: + return 4; + case 3: + return 8; + } +} + +struct evergreen_wm_params { + u32 dram_channels; /* number of dram channels */ + u32 yclk; /* bandwidth per dram data pin in kHz */ + u32 sclk; /* engine clock in kHz */ + u32 disp_clk; /* display clock in kHz */ + u32 src_width; /* viewport width */ + u32 active_time; /* active display time in ns */ + u32 blank_time; /* blank time in ns */ + bool interlaced; /* mode is interlaced */ + fixed20_12 vsc; /* vertical scale ratio */ + u32 num_heads; /* number of active crtcs */ + u32 bytes_per_pixel; /* bytes per pixel display + overlay */ + u32 lb_size; /* line buffer allocated to pipe */ + u32 vtaps; /* vertical scaler taps */ +}; + +static u32 evergreen_dram_bandwidth(struct evergreen_wm_params *wm) +{ + /* Calculate DRAM Bandwidth and the part allocated to display. */ + fixed20_12 dram_efficiency; /* 0.7 */ + fixed20_12 yclk, dram_channels, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + yclk.full = dfixed_const(wm->yclk); + yclk.full = dfixed_div(yclk, a); + dram_channels.full = dfixed_const(wm->dram_channels * 4); + a.full = dfixed_const(10); + dram_efficiency.full = dfixed_const(7); + dram_efficiency.full = dfixed_div(dram_efficiency, a); + bandwidth.full = dfixed_mul(dram_channels, yclk); + bandwidth.full = dfixed_mul(bandwidth, dram_efficiency); + + return dfixed_trunc(bandwidth); +} + +static u32 evergreen_dram_bandwidth_for_display(struct evergreen_wm_params *wm) +{ + /* Calculate DRAM Bandwidth and the part allocated to display. */ + fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */ + fixed20_12 yclk, dram_channels, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + yclk.full = dfixed_const(wm->yclk); + yclk.full = dfixed_div(yclk, a); + dram_channels.full = dfixed_const(wm->dram_channels * 4); + a.full = dfixed_const(10); + disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */ + disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a); + bandwidth.full = dfixed_mul(dram_channels, yclk); + bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation); + + return dfixed_trunc(bandwidth); +} + +static u32 evergreen_data_return_bandwidth(struct evergreen_wm_params *wm) +{ + /* Calculate the display Data return Bandwidth */ + fixed20_12 return_efficiency; /* 0.8 */ + fixed20_12 sclk, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + sclk.full = dfixed_const(wm->sclk); + sclk.full = dfixed_div(sclk, a); + a.full = dfixed_const(10); + return_efficiency.full = dfixed_const(8); + return_efficiency.full = dfixed_div(return_efficiency, a); + a.full = dfixed_const(32); + bandwidth.full = dfixed_mul(a, sclk); + bandwidth.full = dfixed_mul(bandwidth, return_efficiency); + + return dfixed_trunc(bandwidth); +} + +static u32 evergreen_dmif_request_bandwidth(struct evergreen_wm_params *wm) +{ + /* Calculate the DMIF Request Bandwidth */ + fixed20_12 disp_clk_request_efficiency; /* 0.8 */ + fixed20_12 disp_clk, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + disp_clk.full = dfixed_const(wm->disp_clk); + disp_clk.full = dfixed_div(disp_clk, a); + a.full = dfixed_const(10); + disp_clk_request_efficiency.full = dfixed_const(8); + disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a); + a.full = dfixed_const(32); + bandwidth.full = dfixed_mul(a, disp_clk); + bandwidth.full = dfixed_mul(bandwidth, disp_clk_request_efficiency); + + return dfixed_trunc(bandwidth); +} + +static u32 evergreen_available_bandwidth(struct evergreen_wm_params *wm) +{ + /* Calculate the Available bandwidth. Display can use this temporarily but not in average. */ + u32 dram_bandwidth = evergreen_dram_bandwidth(wm); + u32 data_return_bandwidth = evergreen_data_return_bandwidth(wm); + u32 dmif_req_bandwidth = evergreen_dmif_request_bandwidth(wm); + + return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth)); +} + +static u32 evergreen_average_bandwidth(struct evergreen_wm_params *wm) +{ + /* Calculate the display mode Average Bandwidth + * DisplayMode should contain the source and destination dimensions, + * timing, etc. + */ + fixed20_12 bpp; + fixed20_12 line_time; + fixed20_12 src_width; + fixed20_12 bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + line_time.full = dfixed_const(wm->active_time + wm->blank_time); + line_time.full = dfixed_div(line_time, a); + bpp.full = dfixed_const(wm->bytes_per_pixel); + src_width.full = dfixed_const(wm->src_width); + bandwidth.full = dfixed_mul(src_width, bpp); + bandwidth.full = dfixed_mul(bandwidth, wm->vsc); + bandwidth.full = dfixed_div(bandwidth, line_time); + + return dfixed_trunc(bandwidth); +} + +static u32 evergreen_latency_watermark(struct evergreen_wm_params *wm) +{ + /* First calcualte the latency in ns */ + u32 mc_latency = 2000; /* 2000 ns. */ + u32 available_bandwidth = evergreen_available_bandwidth(wm); + u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth; + u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth; + u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */ + u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) + + (wm->num_heads * cursor_line_pair_return_time); + u32 latency = mc_latency + other_heads_data_return_time + dc_latency; + u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time; + fixed20_12 a, b, c; + + if (wm->num_heads == 0) + return 0; + + a.full = dfixed_const(2); + b.full = dfixed_const(1); + if ((wm->vsc.full > a.full) || + ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) || + (wm->vtaps >= 5) || + ((wm->vsc.full >= a.full) && wm->interlaced)) + max_src_lines_per_dst_line = 4; + else + max_src_lines_per_dst_line = 2; + + a.full = dfixed_const(available_bandwidth); + b.full = dfixed_const(wm->num_heads); + a.full = dfixed_div(a, b); + + b.full = dfixed_const(1000); + c.full = dfixed_const(wm->disp_clk); + b.full = dfixed_div(c, b); + c.full = dfixed_const(wm->bytes_per_pixel); + b.full = dfixed_mul(b, c); + + lb_fill_bw = min(dfixed_trunc(a), dfixed_trunc(b)); + + a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); + b.full = dfixed_const(1000); + c.full = dfixed_const(lb_fill_bw); + b.full = dfixed_div(c, b); + a.full = dfixed_div(a, b); + line_fill_time = dfixed_trunc(a); + + if (line_fill_time < wm->active_time) + return latency; + else + return latency + (line_fill_time - wm->active_time); + +} + +static bool evergreen_average_bandwidth_vs_dram_bandwidth_for_display(struct evergreen_wm_params *wm) +{ + if (evergreen_average_bandwidth(wm) <= + (evergreen_dram_bandwidth_for_display(wm) / wm->num_heads)) + return true; + else + return false; +}; + +static bool evergreen_average_bandwidth_vs_available_bandwidth(struct evergreen_wm_params *wm) +{ + if (evergreen_average_bandwidth(wm) <= + (evergreen_available_bandwidth(wm) / wm->num_heads)) + return true; + else + return false; +}; + +static bool evergreen_check_latency_hiding(struct evergreen_wm_params *wm) +{ + u32 lb_partitions = wm->lb_size / wm->src_width; + u32 line_time = wm->active_time + wm->blank_time; + u32 latency_tolerant_lines; + u32 latency_hiding; + fixed20_12 a; + + a.full = dfixed_const(1); + if (wm->vsc.full > a.full) + latency_tolerant_lines = 1; + else { + if (lb_partitions <= (wm->vtaps + 1)) + latency_tolerant_lines = 1; + else + latency_tolerant_lines = 2; + } + + latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time); + + if (evergreen_latency_watermark(wm) <= latency_hiding) + return true; + else + return false; +} + +static void evergreen_program_watermarks(struct radeon_device *rdev, + struct radeon_crtc *radeon_crtc, + u32 lb_size, u32 num_heads) +{ + struct drm_display_mode *mode = &radeon_crtc->base.mode; + struct evergreen_wm_params wm; + u32 pixel_period; + u32 line_time = 0; + u32 latency_watermark_a = 0, latency_watermark_b = 0; + u32 priority_a_mark = 0, priority_b_mark = 0; + u32 priority_a_cnt = PRIORITY_OFF; + u32 priority_b_cnt = PRIORITY_OFF; + u32 pipe_offset = radeon_crtc->crtc_id * 16; + u32 tmp, arb_control3; + fixed20_12 a, b, c; + + if (radeon_crtc->base.enabled && num_heads && mode) { + pixel_period = 1000000 / (u32)mode->clock; + line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535); + priority_a_cnt = 0; + priority_b_cnt = 0; + + wm.yclk = rdev->pm.current_mclk * 10; + wm.sclk = rdev->pm.current_sclk * 10; + wm.disp_clk = mode->clock; + wm.src_width = mode->crtc_hdisplay; + wm.active_time = mode->crtc_hdisplay * pixel_period; + wm.blank_time = line_time - wm.active_time; + wm.interlaced = false; + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + wm.interlaced = true; + wm.vsc = radeon_crtc->vsc; + wm.vtaps = 1; + if (radeon_crtc->rmx_type != RMX_OFF) + wm.vtaps = 2; + wm.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm.lb_size = lb_size; + wm.dram_channels = evergreen_get_number_of_dram_channels(rdev); + wm.num_heads = num_heads; + + /* set for high clocks */ + latency_watermark_a = min(evergreen_latency_watermark(&wm), (u32)65535); + /* set for low clocks */ + /* wm.yclk = low clk; wm.sclk = low clk */ + latency_watermark_b = min(evergreen_latency_watermark(&wm), (u32)65535); + + /* possibly force display priority to high */ + /* should really do this at mode validation time... */ + if (!evergreen_average_bandwidth_vs_dram_bandwidth_for_display(&wm) || + !evergreen_average_bandwidth_vs_available_bandwidth(&wm) || + !evergreen_check_latency_hiding(&wm) || + (rdev->disp_priority == 2)) { + DRM_INFO("force priority to high\n"); + priority_a_cnt |= PRIORITY_ALWAYS_ON; + priority_b_cnt |= PRIORITY_ALWAYS_ON; + } + + a.full = dfixed_const(1000); + b.full = dfixed_const(mode->clock); + b.full = dfixed_div(b, a); + c.full = dfixed_const(latency_watermark_a); + c.full = dfixed_mul(c, b); + c.full = dfixed_mul(c, radeon_crtc->hsc); + c.full = dfixed_div(c, a); + a.full = dfixed_const(16); + c.full = dfixed_div(c, a); + priority_a_mark = dfixed_trunc(c); + priority_a_cnt |= priority_a_mark & PRIORITY_MARK_MASK; + + a.full = dfixed_const(1000); + b.full = dfixed_const(mode->clock); + b.full = dfixed_div(b, a); + c.full = dfixed_const(latency_watermark_b); + c.full = dfixed_mul(c, b); + c.full = dfixed_mul(c, radeon_crtc->hsc); + c.full = dfixed_div(c, a); + a.full = dfixed_const(16); + c.full = dfixed_div(c, a); + priority_b_mark = dfixed_trunc(c); + priority_b_cnt |= priority_b_mark & PRIORITY_MARK_MASK; + } + + /* select wm A */ + arb_control3 = RREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset); + tmp = arb_control3; + tmp &= ~LATENCY_WATERMARK_MASK(3); + tmp |= LATENCY_WATERMARK_MASK(1); + WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, tmp); + WREG32(PIPE0_LATENCY_CONTROL + pipe_offset, + (LATENCY_LOW_WATERMARK(latency_watermark_a) | + LATENCY_HIGH_WATERMARK(line_time))); + /* select wm B */ + tmp = RREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset); + tmp &= ~LATENCY_WATERMARK_MASK(3); + tmp |= LATENCY_WATERMARK_MASK(2); + WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, tmp); + WREG32(PIPE0_LATENCY_CONTROL + pipe_offset, + (LATENCY_LOW_WATERMARK(latency_watermark_b) | + LATENCY_HIGH_WATERMARK(line_time))); + /* restore original selection */ + WREG32(PIPE0_ARBITRATION_CONTROL3 + pipe_offset, arb_control3); + + /* write the priority marks */ + WREG32(PRIORITY_A_CNT + radeon_crtc->crtc_offset, priority_a_cnt); + WREG32(PRIORITY_B_CNT + radeon_crtc->crtc_offset, priority_b_cnt); + +} + void evergreen_bandwidth_update(struct radeon_device *rdev) { - /* XXX */ + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + u32 num_heads = 0, lb_size; + int i; + + radeon_update_display_priority(rdev); + + for (i = 0; i < rdev->num_crtc; i++) { + if (rdev->mode_info.crtcs[i]->base.enabled) + num_heads++; + } + for (i = 0; i < rdev->num_crtc; i += 2) { + mode0 = &rdev->mode_info.crtcs[i]->base.mode; + mode1 = &rdev->mode_info.crtcs[i+1]->base.mode; + lb_size = evergreen_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i], mode0, mode1); + evergreen_program_watermarks(rdev, rdev->mode_info.crtcs[i], lb_size, num_heads); + lb_size = evergreen_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i+1], mode1, mode0); + evergreen_program_watermarks(rdev, rdev->mode_info.crtcs[i+1], lb_size, num_heads); + } } static int evergreen_mc_wait_for_idle(struct radeon_device *rdev) @@ -312,6 +827,8 @@ void evergreen_pcie_gart_tlb_flush(struct radeon_device *rdev) unsigned i; u32 tmp; + WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1); + WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1)); for (i = 0; i < rdev->usec_timeout; i++) { /* read MC_STATUS */ @@ -452,31 +969,39 @@ static void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_sa save->vga_hdp_control = RREG32(VGA_HDP_CONTROL); save->crtc_control[0] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET); save->crtc_control[1] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET); - save->crtc_control[2] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET); - save->crtc_control[3] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET); - save->crtc_control[4] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET); - save->crtc_control[5] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET); + if (!(rdev->flags & RADEON_IS_IGP)) { + save->crtc_control[2] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET); + save->crtc_control[3] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET); + save->crtc_control[4] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET); + save->crtc_control[5] = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET); + } /* Stop all video */ WREG32(VGA_RENDER_CONTROL, 0); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1); + } WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } WREG32(D1VGA_CONTROL, 0); WREG32(D2VGA_CONTROL, 0); @@ -506,41 +1031,43 @@ static void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_ WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC1_REGISTER_OFFSET, (u32)rdev->mc.vram_start); - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET, - upper_32_bits(rdev->mc.vram_start)); - WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); - WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET, - (u32)rdev->mc.vram_start); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC2_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC2_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC3_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC3_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC4_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC4_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS_HIGH + EVERGREEN_CRTC5_REGISTER_OFFSET, + upper_32_bits(rdev->mc.vram_start)); + WREG32(EVERGREEN_GRPH_PRIMARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + WREG32(EVERGREEN_GRPH_SECONDARY_SURFACE_ADDRESS + EVERGREEN_CRTC5_REGISTER_OFFSET, + (u32)rdev->mc.vram_start); + } WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(rdev->mc.vram_start)); WREG32(EVERGREEN_VGA_MEMORY_BASE_ADDRESS, (u32)rdev->mc.vram_start); @@ -556,22 +1083,28 @@ static void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_ WREG32(EVERGREEN_D6VGA_CONTROL, save->vga_control[5]); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 1); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 1); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 1); + } WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, save->crtc_control[0]); WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, save->crtc_control[1]); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, save->crtc_control[2]); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, save->crtc_control[3]); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, save->crtc_control[4]); - WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, save->crtc_control[5]); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, save->crtc_control[2]); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, save->crtc_control[3]); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, save->crtc_control[4]); + WREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, save->crtc_control[5]); + } WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); - WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(EVERGREEN_CRTC_UPDATE_LOCK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } WREG32(VGA_RENDER_CONTROL, save->vga_render_control); } @@ -619,11 +1152,17 @@ static void evergreen_mc_program(struct radeon_device *rdev) rdev->mc.vram_end >> 12); } WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0); + if (rdev->flags & RADEON_IS_IGP) { + tmp = RREG32(MC_FUS_VM_FB_OFFSET) & 0x000FFFFF; + tmp |= ((rdev->mc.vram_end >> 20) & 0xF) << 24; + tmp |= ((rdev->mc.vram_start >> 20) & 0xF) << 20; + WREG32(MC_FUS_VM_FB_OFFSET, tmp); + } tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16; tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF); WREG32(MC_VM_FB_LOCATION, tmp); WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8)); - WREG32(HDP_NONSURFACE_INFO, (2 << 7)); + WREG32(HDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF); if (rdev->flags & RADEON_IS_AGP) { WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 16); @@ -646,6 +1185,22 @@ static void evergreen_mc_program(struct radeon_device *rdev) /* * CP. */ +void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + /* set to DX10/11 mode */ + radeon_ring_write(rdev, PACKET3(PACKET3_MODE_CONTROL, 0)); + radeon_ring_write(rdev, 1); + /* FIXME: implement */ + radeon_ring_write(rdev, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); + radeon_ring_write(rdev, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + radeon_ring_write(rdev, upper_32_bits(ib->gpu_addr) & 0xFF); + radeon_ring_write(rdev, ib->length_dw); +} + static int evergreen_cp_load_microcode(struct radeon_device *rdev) { @@ -656,7 +1211,11 @@ static int evergreen_cp_load_microcode(struct radeon_device *rdev) return -EINVAL; r700_cp_stop(rdev); - WREG32(CP_RB_CNTL, RB_NO_UPDATE | (15 << 8) | (3 << 0)); + WREG32(CP_RB_CNTL, +#ifdef __BIG_ENDIAN + BUF_SWAP_32BIT | +#endif + RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3)); fw_data = (const __be32 *)rdev->pfp_fw->data; WREG32(CP_PFP_UCODE_ADDR, 0); @@ -677,7 +1236,7 @@ static int evergreen_cp_load_microcode(struct radeon_device *rdev) static int evergreen_cp_start(struct radeon_device *rdev) { - int r; + int r, i; uint32_t cp_me; r = radeon_ring_lock(rdev, 7); @@ -697,16 +1256,44 @@ static int evergreen_cp_start(struct radeon_device *rdev) cp_me = 0xff; WREG32(CP_ME_CNTL, cp_me); - r = radeon_ring_lock(rdev, 4); + r = radeon_ring_lock(rdev, evergreen_default_size + 19); if (r) { DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); return r; } - /* init some VGT regs */ - radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(rdev, (VGT_VERTEX_REUSE_BLOCK_CNTL - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(rdev, 0xe); - radeon_ring_write(rdev, 0x10); + + /* setup clear context state */ + radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + radeon_ring_write(rdev, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); + + for (i = 0; i < evergreen_default_size; i++) + radeon_ring_write(rdev, evergreen_default_state[i]); + + radeon_ring_write(rdev, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + radeon_ring_write(rdev, PACKET3_PREAMBLE_END_CLEAR_STATE); + + /* set clear context state */ + radeon_ring_write(rdev, PACKET3(PACKET3_CLEAR_STATE, 0)); + radeon_ring_write(rdev, 0); + + /* SQ_VTX_BASE_VTX_LOC */ + radeon_ring_write(rdev, 0xc0026f00); + radeon_ring_write(rdev, 0x00000000); + radeon_ring_write(rdev, 0x00000000); + radeon_ring_write(rdev, 0x00000000); + + /* Clear consts */ + radeon_ring_write(rdev, 0xc0036f00); + radeon_ring_write(rdev, 0x00000bc4); + radeon_ring_write(rdev, 0xffffffff); + radeon_ring_write(rdev, 0xffffffff); + radeon_ring_write(rdev, 0xffffffff); + + radeon_ring_write(rdev, 0xc0026900); + radeon_ring_write(rdev, 0x00000316); + radeon_ring_write(rdev, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */ + radeon_ring_write(rdev, 0x00000010); /* */ + radeon_ring_unlock_commit(rdev); return 0; @@ -731,7 +1318,7 @@ int evergreen_cp_resume(struct radeon_device *rdev) /* Set ring buffer size */ rb_bufsz = drm_order(rdev->cp.ring_size / 8); - tmp = RB_NO_UPDATE | (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz; + tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz; #ifdef __BIG_ENDIAN tmp |= BUF_SWAP_32BIT; #endif @@ -745,8 +1332,23 @@ int evergreen_cp_resume(struct radeon_device *rdev) WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA); WREG32(CP_RB_RPTR_WR, 0); WREG32(CP_RB_WPTR, 0); - WREG32(CP_RB_RPTR_ADDR, rdev->cp.gpu_addr & 0xFFFFFFFF); - WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->cp.gpu_addr)); + + /* set the wb address wether it's enabled or not */ + WREG32(CP_RB_RPTR_ADDR, +#ifdef __BIG_ENDIAN + RB_RPTR_SWAP(2) | +#endif + ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC)); + WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF); + WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF); + + if (rdev->wb.enabled) + WREG32(SCRATCH_UMSK, 0xff); + else { + tmp |= RB_NO_UPDATE; + WREG32(SCRATCH_UMSK, 0); + } + mdelay(1); WREG32(CP_RB_CNTL, tmp); @@ -813,11 +1415,15 @@ static u32 evergreen_get_tile_pipe_to_backend_map(struct radeon_device *rdev, switch (rdev->family) { case CHIP_CEDAR: case CHIP_REDWOOD: + case CHIP_PALM: + case CHIP_TURKS: + case CHIP_CAICOS: force_no_swizzle = false; break; case CHIP_CYPRESS: case CHIP_HEMLOCK: case CHIP_JUNIPER: + case CHIP_BARTS: default: force_no_swizzle = true; break; @@ -912,6 +1518,46 @@ static u32 evergreen_get_tile_pipe_to_backend_map(struct radeon_device *rdev, return backend_map; } +static void evergreen_program_channel_remap(struct radeon_device *rdev) +{ + u32 tcp_chan_steer_lo, tcp_chan_steer_hi, mc_shared_chremap, tmp; + + tmp = RREG32(MC_SHARED_CHMAP); + switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { + case 0: + case 1: + case 2: + case 3: + default: + /* default mapping */ + mc_shared_chremap = 0x00fac688; + break; + } + + switch (rdev->family) { + case CHIP_HEMLOCK: + case CHIP_CYPRESS: + case CHIP_BARTS: + tcp_chan_steer_lo = 0x54763210; + tcp_chan_steer_hi = 0x0000ba98; + break; + case CHIP_JUNIPER: + case CHIP_REDWOOD: + case CHIP_CEDAR: + case CHIP_PALM: + case CHIP_TURKS: + case CHIP_CAICOS: + default: + tcp_chan_steer_lo = 0x76543210; + tcp_chan_steer_hi = 0x0000ba98; + break; + } + + WREG32(TCP_CHAN_STEER_LO, tcp_chan_steer_lo); + WREG32(TCP_CHAN_STEER_HI, tcp_chan_steer_hi); + WREG32(MC_SHARED_CHREMAP, mc_shared_chremap); +} + static void evergreen_gpu_init(struct radeon_device *rdev) { u32 cc_rb_backend_disable = 0; @@ -1019,6 +1665,90 @@ static void evergreen_gpu_init(struct radeon_device *rdev) rdev->config.evergreen.max_hw_contexts = 4; rdev->config.evergreen.sq_num_cf_insts = 1; + rdev->config.evergreen.sc_prim_fifo_size = 0x40; + rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30; + rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130; + break; + case CHIP_PALM: + rdev->config.evergreen.num_ses = 1; + rdev->config.evergreen.max_pipes = 2; + rdev->config.evergreen.max_tile_pipes = 2; + rdev->config.evergreen.max_simds = 2; + rdev->config.evergreen.max_backends = 1 * rdev->config.evergreen.num_ses; + rdev->config.evergreen.max_gprs = 256; + rdev->config.evergreen.max_threads = 192; + rdev->config.evergreen.max_gs_threads = 16; + rdev->config.evergreen.max_stack_entries = 256; + rdev->config.evergreen.sx_num_of_sets = 4; + rdev->config.evergreen.sx_max_export_size = 128; + rdev->config.evergreen.sx_max_export_pos_size = 32; + rdev->config.evergreen.sx_max_export_smx_size = 96; + rdev->config.evergreen.max_hw_contexts = 4; + rdev->config.evergreen.sq_num_cf_insts = 1; + + rdev->config.evergreen.sc_prim_fifo_size = 0x40; + rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30; + rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130; + break; + case CHIP_BARTS: + rdev->config.evergreen.num_ses = 2; + rdev->config.evergreen.max_pipes = 4; + rdev->config.evergreen.max_tile_pipes = 8; + rdev->config.evergreen.max_simds = 7; + rdev->config.evergreen.max_backends = 4 * rdev->config.evergreen.num_ses; + rdev->config.evergreen.max_gprs = 256; + rdev->config.evergreen.max_threads = 248; + rdev->config.evergreen.max_gs_threads = 32; + rdev->config.evergreen.max_stack_entries = 512; + rdev->config.evergreen.sx_num_of_sets = 4; + rdev->config.evergreen.sx_max_export_size = 256; + rdev->config.evergreen.sx_max_export_pos_size = 64; + rdev->config.evergreen.sx_max_export_smx_size = 192; + rdev->config.evergreen.max_hw_contexts = 8; + rdev->config.evergreen.sq_num_cf_insts = 2; + + rdev->config.evergreen.sc_prim_fifo_size = 0x100; + rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30; + rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130; + break; + case CHIP_TURKS: + rdev->config.evergreen.num_ses = 1; + rdev->config.evergreen.max_pipes = 4; + rdev->config.evergreen.max_tile_pipes = 4; + rdev->config.evergreen.max_simds = 6; + rdev->config.evergreen.max_backends = 2 * rdev->config.evergreen.num_ses; + rdev->config.evergreen.max_gprs = 256; + rdev->config.evergreen.max_threads = 248; + rdev->config.evergreen.max_gs_threads = 32; + rdev->config.evergreen.max_stack_entries = 256; + rdev->config.evergreen.sx_num_of_sets = 4; + rdev->config.evergreen.sx_max_export_size = 256; + rdev->config.evergreen.sx_max_export_pos_size = 64; + rdev->config.evergreen.sx_max_export_smx_size = 192; + rdev->config.evergreen.max_hw_contexts = 8; + rdev->config.evergreen.sq_num_cf_insts = 2; + + rdev->config.evergreen.sc_prim_fifo_size = 0x100; + rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30; + rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130; + break; + case CHIP_CAICOS: + rdev->config.evergreen.num_ses = 1; + rdev->config.evergreen.max_pipes = 4; + rdev->config.evergreen.max_tile_pipes = 2; + rdev->config.evergreen.max_simds = 2; + rdev->config.evergreen.max_backends = 1 * rdev->config.evergreen.num_ses; + rdev->config.evergreen.max_gprs = 256; + rdev->config.evergreen.max_threads = 192; + rdev->config.evergreen.max_gs_threads = 16; + rdev->config.evergreen.max_stack_entries = 256; + rdev->config.evergreen.sx_num_of_sets = 4; + rdev->config.evergreen.sx_max_export_size = 128; + rdev->config.evergreen.sx_max_export_pos_size = 32; + rdev->config.evergreen.sx_max_export_smx_size = 96; + rdev->config.evergreen.max_hw_contexts = 4; + rdev->config.evergreen.sq_num_cf_insts = 1; + rdev->config.evergreen.sc_prim_fifo_size = 0x40; rdev->config.evergreen.sc_hiz_tile_fifo_size = 0x30; rdev->config.evergreen.sc_earlyz_tile_fifo_size = 0x130; @@ -1164,6 +1894,7 @@ static void evergreen_gpu_init(struct radeon_device *rdev) switch (rdev->family) { case CHIP_CYPRESS: case CHIP_HEMLOCK: + case CHIP_BARTS: gb_backend_map = 0x66442200; break; case CHIP_JUNIPER: @@ -1180,12 +1911,43 @@ static void evergreen_gpu_init(struct radeon_device *rdev) } } - rdev->config.evergreen.tile_config = gb_addr_config; + /* setup tiling info dword. gb_addr_config is not adequate since it does + * not have bank info, so create a custom tiling dword. + * bits 3:0 num_pipes + * bits 7:4 num_banks + * bits 11:8 group_size + * bits 15:12 row_size + */ + rdev->config.evergreen.tile_config = 0; + switch (rdev->config.evergreen.max_tile_pipes) { + case 1: + default: + rdev->config.evergreen.tile_config |= (0 << 0); + break; + case 2: + rdev->config.evergreen.tile_config |= (1 << 0); + break; + case 4: + rdev->config.evergreen.tile_config |= (2 << 0); + break; + case 8: + rdev->config.evergreen.tile_config |= (3 << 0); + break; + } + rdev->config.evergreen.tile_config |= + ((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) << 4; + rdev->config.evergreen.tile_config |= + ((mc_arb_ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT) << 8; + rdev->config.evergreen.tile_config |= + ((gb_addr_config & 0x30000000) >> 28) << 12; + WREG32(GB_BACKEND_MAP, gb_backend_map); WREG32(GB_ADDR_CONFIG, gb_addr_config); WREG32(DMIF_ADDR_CONFIG, gb_addr_config); WREG32(HDP_ADDR_CONFIG, gb_addr_config); + evergreen_program_channel_remap(rdev); + num_shader_engines = ((RREG32(GB_ADDR_CONFIG) & NUM_SHADER_ENGINES(3)) >> 12) + 1; grbm_gfx_index = INSTANCE_BROADCAST_WRITES; @@ -1268,9 +2030,16 @@ static void evergreen_gpu_init(struct radeon_device *rdev) GS_PRIO(2) | ES_PRIO(3)); - if (rdev->family == CHIP_CEDAR) + switch (rdev->family) { + case CHIP_CEDAR: + case CHIP_PALM: + case CHIP_CAICOS: /* no vertex cache */ sq_config &= ~VC_ENABLE; + break; + default: + break; + } sq_lds_resource_mgmt = RREG32(SQ_LDS_RESOURCE_MGMT); @@ -1282,10 +2051,15 @@ static void evergreen_gpu_init(struct radeon_device *rdev) sq_gpr_resource_mgmt_3 = NUM_HS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 3 / 32); sq_gpr_resource_mgmt_3 |= NUM_LS_GPRS((rdev->config.evergreen.max_gprs - (4 * 2)) * 3 / 32); - if (rdev->family == CHIP_CEDAR) + switch (rdev->family) { + case CHIP_CEDAR: + case CHIP_PALM: ps_thread_count = 96; - else + break; + default: ps_thread_count = 128; + break; + } sq_thread_resource_mgmt = NUM_PS_THREADS(ps_thread_count); sq_thread_resource_mgmt |= NUM_VS_THREADS((((rdev->config.evergreen.max_threads - ps_thread_count) / 6) / 8) * 8); @@ -1316,14 +2090,21 @@ static void evergreen_gpu_init(struct radeon_device *rdev) WREG32(PA_SC_FORCE_EOV_MAX_CNTS, (FORCE_EOV_MAX_CLK_CNT(4095) | FORCE_EOV_MAX_REZ_CNT(255))); - if (rdev->family == CHIP_CEDAR) + switch (rdev->family) { + case CHIP_CEDAR: + case CHIP_PALM: + case CHIP_CAICOS: vgt_cache_invalidation = CACHE_INVALIDATION(TC_ONLY); - else + break; + default: vgt_cache_invalidation = CACHE_INVALIDATION(VC_AND_TC); + break; + } vgt_cache_invalidation |= AUTO_INVLD_EN(ES_AND_GS_AUTO); WREG32(VGT_CACHE_INVALIDATION, vgt_cache_invalidation); WREG32(VGT_GS_VERTEX_REUSE, 16); + WREG32(PA_SU_LINE_STIPPLE_VALUE, 0); WREG32(PA_SC_LINE_STIPPLE_STATE, 0); WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, 14); @@ -1403,12 +2184,17 @@ int evergreen_mc_init(struct radeon_device *rdev) rdev->mc.aper_base = pci_resource_start(rdev->pdev, 0); rdev->mc.aper_size = pci_resource_len(rdev->pdev, 0); /* Setup GPU memory space */ - /* size in MB on evergreen */ - rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; - rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; + if (rdev->flags & RADEON_IS_IGP) { + /* size in bytes on fusion */ + rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE); + rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE); + } else { + /* size in MB on evergreen */ + rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; + rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; + } rdev->mc.visible_vram_size = rdev->mc.aper_size; - rdev->mc.active_vram_size = rdev->mc.visible_vram_size; - r600_vram_gtt_location(rdev, &rdev->mc); + r700_vram_gtt_location(rdev, &rdev->mc); radeon_update_bandwidth_info(rdev); return 0; @@ -1416,16 +2202,40 @@ int evergreen_mc_init(struct radeon_device *rdev) bool evergreen_gpu_is_lockup(struct radeon_device *rdev) { - /* FIXME: implement for evergreen */ - return false; + u32 srbm_status; + u32 grbm_status; + u32 grbm_status_se0, grbm_status_se1; + struct r100_gpu_lockup *lockup = &rdev->config.evergreen.lockup; + int r; + + srbm_status = RREG32(SRBM_STATUS); + grbm_status = RREG32(GRBM_STATUS); + grbm_status_se0 = RREG32(GRBM_STATUS_SE0); + grbm_status_se1 = RREG32(GRBM_STATUS_SE1); + if (!(grbm_status & GUI_ACTIVE)) { + r100_gpu_lockup_update(lockup, &rdev->cp); + return false; + } + /* force CP activities */ + r = radeon_ring_lock(rdev, 2); + if (!r) { + /* PACKET2 NOP */ + radeon_ring_write(rdev, 0x80000000); + radeon_ring_write(rdev, 0x80000000); + radeon_ring_unlock_commit(rdev); + } + rdev->cp.rptr = RREG32(CP_RB_RPTR); + return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp); } static int evergreen_gpu_soft_reset(struct radeon_device *rdev) { struct evergreen_mc_save save; - u32 srbm_reset = 0; u32 grbm_reset = 0; + if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) + return 0; + dev_info(rdev->dev, "GPU softreset \n"); dev_info(rdev->dev, " GRBM_STATUS=0x%08X\n", RREG32(GRBM_STATUS)); @@ -1462,16 +2272,6 @@ static int evergreen_gpu_soft_reset(struct radeon_device *rdev) udelay(50); WREG32(GRBM_SOFT_RESET, 0); (void)RREG32(GRBM_SOFT_RESET); - - /* reset all the system blocks */ - srbm_reset = SRBM_SOFT_RESET_ALL_MASK; - - dev_info(rdev->dev, " SRBM_SOFT_RESET=0x%08X\n", srbm_reset); - WREG32(SRBM_SOFT_RESET, srbm_reset); - (void)RREG32(SRBM_SOFT_RESET); - udelay(50); - WREG32(SRBM_SOFT_RESET, 0); - (void)RREG32(SRBM_SOFT_RESET); /* Wait a little for things to settle down */ udelay(50); dev_info(rdev->dev, " GRBM_STATUS=0x%08X\n", @@ -1482,10 +2282,6 @@ static int evergreen_gpu_soft_reset(struct radeon_device *rdev) RREG32(GRBM_STATUS_SE1)); dev_info(rdev->dev, " SRBM_STATUS=0x%08X\n", RREG32(SRBM_STATUS)); - /* After reset we need to reinit the asic as GPU often endup in an - * incoherent state. - */ - atom_asic_init(rdev->mode_info.atom_context); evergreen_mc_resume(rdev, &save); return 0; } @@ -1525,17 +2321,21 @@ void evergreen_disable_interrupt_state(struct radeon_device *rdev) WREG32(GRBM_INT_CNTL, 0); WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); - WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); - WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); - WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); - WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); - WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); - WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); - WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); - WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } WREG32(DACA_AUTODETECT_INT_CONTROL, 0); WREG32(DACB_AUTODETECT_INT_CONTROL, 0); @@ -1561,9 +2361,10 @@ int evergreen_irq_set(struct radeon_device *rdev) u32 crtc1 = 0, crtc2 = 0, crtc3 = 0, crtc4 = 0, crtc5 = 0, crtc6 = 0; u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6; u32 grbm_int_cntl = 0; + u32 grph1 = 0, grph2 = 0, grph3 = 0, grph4 = 0, grph5 = 0, grph6 = 0; if (!rdev->irq.installed) { - WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n"); + WARN(1, "Can't enable IRQ/MSI because no handler is installed\n"); return -EINVAL; } /* don't enable anything if the ih is disabled */ @@ -1584,28 +2385,35 @@ int evergreen_irq_set(struct radeon_device *rdev) if (rdev->irq.sw_int) { DRM_DEBUG("evergreen_irq_set: sw int\n"); cp_int_cntl |= RB_INT_ENABLE; + cp_int_cntl |= TIME_STAMP_INT_ENABLE; } - if (rdev->irq.crtc_vblank_int[0]) { + if (rdev->irq.crtc_vblank_int[0] || + rdev->irq.pflip[0]) { DRM_DEBUG("evergreen_irq_set: vblank 0\n"); crtc1 |= VBLANK_INT_MASK; } - if (rdev->irq.crtc_vblank_int[1]) { + if (rdev->irq.crtc_vblank_int[1] || + rdev->irq.pflip[1]) { DRM_DEBUG("evergreen_irq_set: vblank 1\n"); crtc2 |= VBLANK_INT_MASK; } - if (rdev->irq.crtc_vblank_int[2]) { + if (rdev->irq.crtc_vblank_int[2] || + rdev->irq.pflip[2]) { DRM_DEBUG("evergreen_irq_set: vblank 2\n"); crtc3 |= VBLANK_INT_MASK; } - if (rdev->irq.crtc_vblank_int[3]) { + if (rdev->irq.crtc_vblank_int[3] || + rdev->irq.pflip[3]) { DRM_DEBUG("evergreen_irq_set: vblank 3\n"); crtc4 |= VBLANK_INT_MASK; } - if (rdev->irq.crtc_vblank_int[4]) { + if (rdev->irq.crtc_vblank_int[4] || + rdev->irq.pflip[4]) { DRM_DEBUG("evergreen_irq_set: vblank 4\n"); crtc5 |= VBLANK_INT_MASK; } - if (rdev->irq.crtc_vblank_int[5]) { + if (rdev->irq.crtc_vblank_int[5] || + rdev->irq.pflip[5]) { DRM_DEBUG("evergreen_irq_set: vblank 5\n"); crtc6 |= VBLANK_INT_MASK; } @@ -1643,10 +2451,19 @@ int evergreen_irq_set(struct radeon_device *rdev) WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, crtc1); WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, crtc2); - WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, crtc3); - WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, crtc4); - WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, crtc5); - WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, crtc6); + if (!(rdev->flags & RADEON_IS_IGP)) { + WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, crtc3); + WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, crtc4); + WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, crtc5); + WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, crtc6); + } + + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, grph1); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, grph2); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, grph3); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, grph4); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, grph5); + WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, grph6); WREG32(DC_HPD1_INT_CONTROL, hpd1); WREG32(DC_HPD2_INT_CONTROL, hpd2); @@ -1658,79 +2475,92 @@ int evergreen_irq_set(struct radeon_device *rdev) return 0; } -static inline void evergreen_irq_ack(struct radeon_device *rdev, - u32 *disp_int, - u32 *disp_int_cont, - u32 *disp_int_cont2, - u32 *disp_int_cont3, - u32 *disp_int_cont4, - u32 *disp_int_cont5) +static inline void evergreen_irq_ack(struct radeon_device *rdev) { u32 tmp; - *disp_int = RREG32(DISP_INTERRUPT_STATUS); - *disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE); - *disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2); - *disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3); - *disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4); - *disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5); - - if (*disp_int & LB_D1_VBLANK_INTERRUPT) + rdev->irq.stat_regs.evergreen.disp_int = RREG32(DISP_INTERRUPT_STATUS); + rdev->irq.stat_regs.evergreen.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE); + rdev->irq.stat_regs.evergreen.disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2); + rdev->irq.stat_regs.evergreen.disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3); + rdev->irq.stat_regs.evergreen.disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4); + rdev->irq.stat_regs.evergreen.disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5); + rdev->irq.stat_regs.evergreen.d1grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET); + rdev->irq.stat_regs.evergreen.d2grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET); + rdev->irq.stat_regs.evergreen.d3grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET); + rdev->irq.stat_regs.evergreen.d4grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET); + rdev->irq.stat_regs.evergreen.d5grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET); + rdev->irq.stat_regs.evergreen.d6grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET); + + if (rdev->irq.stat_regs.evergreen.d1grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + if (rdev->irq.stat_regs.evergreen.d2grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + if (rdev->irq.stat_regs.evergreen.d3grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + if (rdev->irq.stat_regs.evergreen.d4grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + if (rdev->irq.stat_regs.evergreen.d5grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + if (rdev->irq.stat_regs.evergreen.d6grph_int & GRPH_PFLIP_INT_OCCURRED) + WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR); + + if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int & LB_D1_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int_cont & LB_D2_VBLANK_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int_cont & LB_D2_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int_cont2 & LB_D3_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int_cont3 & LB_D4_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int_cont4 & LB_D5_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) WREG32(VBLANK_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VBLANK_ACK); - if (*disp_int_cont5 & LB_D6_VLINE_INTERRUPT) + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT) WREG32(VLINE_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VLINE_ACK); - if (*disp_int & DC_HPD1_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT) { tmp = RREG32(DC_HPD1_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD1_INT_CONTROL, tmp); } - if (*disp_int_cont & DC_HPD2_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT) { tmp = RREG32(DC_HPD2_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD2_INT_CONTROL, tmp); } - if (*disp_int_cont2 & DC_HPD3_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT) { tmp = RREG32(DC_HPD3_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD3_INT_CONTROL, tmp); } - if (*disp_int_cont3 & DC_HPD4_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT) { tmp = RREG32(DC_HPD4_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD4_INT_CONTROL, tmp); } - if (*disp_int_cont4 & DC_HPD5_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT) { tmp = RREG32(DC_HPD5_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD5_INT_CONTROL, tmp); } - if (*disp_int_cont5 & DC_HPD6_INTERRUPT) { + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT) { tmp = RREG32(DC_HPD5_INT_CONTROL); tmp |= DC_HPDx_INT_ACK; WREG32(DC_HPD6_INT_CONTROL, tmp); @@ -1739,14 +2569,10 @@ static inline void evergreen_irq_ack(struct radeon_device *rdev, void evergreen_irq_disable(struct radeon_device *rdev) { - u32 disp_int, disp_int_cont, disp_int_cont2; - u32 disp_int_cont3, disp_int_cont4, disp_int_cont5; - r600_disable_interrupts(rdev); /* Wait and acknowledge irq */ mdelay(1); - evergreen_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2, - &disp_int_cont3, &disp_int_cont4, &disp_int_cont5); + evergreen_irq_ack(rdev); evergreen_disable_interrupt_state(rdev); } @@ -1760,8 +2586,10 @@ static inline u32 evergreen_get_ih_wptr(struct radeon_device *rdev) { u32 wptr, tmp; - /* XXX use writeback */ - wptr = RREG32(IH_RB_WPTR); + if (rdev->wb.enabled) + wptr = rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4]; + else + wptr = RREG32(IH_RB_WPTR); if (wptr & RB_OVERFLOW) { /* When a ring buffer overflow happen start parsing interrupt @@ -1784,8 +2612,6 @@ int evergreen_irq_process(struct radeon_device *rdev) u32 rptr = rdev->ih.rptr; u32 src_id, src_data; u32 ring_index; - u32 disp_int, disp_int_cont, disp_int_cont2; - u32 disp_int_cont3, disp_int_cont4, disp_int_cont5; unsigned long flags; bool queue_hotplug = false; @@ -1806,30 +2632,34 @@ int evergreen_irq_process(struct radeon_device *rdev) restart_ih: /* display interrupts */ - evergreen_irq_ack(rdev, &disp_int, &disp_int_cont, &disp_int_cont2, - &disp_int_cont3, &disp_int_cont4, &disp_int_cont5); + evergreen_irq_ack(rdev); rdev->ih.wptr = wptr; while (rptr != wptr) { /* wptr/rptr are in bytes! */ ring_index = rptr / 4; - src_id = rdev->ih.ring[ring_index] & 0xff; - src_data = rdev->ih.ring[ring_index + 1] & 0xfffffff; + src_id = le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff; + src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff; switch (src_id) { case 1: /* D1 vblank/vline */ switch (src_data) { case 0: /* D1 vblank */ - if (disp_int & LB_D1_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 0); - wake_up(&rdev->irq.vblank_queue); - disp_int &= ~LB_D1_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[0]) { + drm_handle_vblank(rdev->ddev, 0); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[0]) + radeon_crtc_handle_flip(rdev, 0); + rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VBLANK_INTERRUPT; DRM_DEBUG("IH: D1 vblank\n"); } break; case 1: /* D1 vline */ - if (disp_int & LB_D1_VLINE_INTERRUPT) { - disp_int &= ~LB_D1_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VLINE_INTERRUPT; DRM_DEBUG("IH: D1 vline\n"); } break; @@ -1841,16 +2671,21 @@ restart_ih: case 2: /* D2 vblank/vline */ switch (src_data) { case 0: /* D2 vblank */ - if (disp_int_cont & LB_D2_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 1); - wake_up(&rdev->irq.vblank_queue); - disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[1]) { + drm_handle_vblank(rdev->ddev, 1); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[1]) + radeon_crtc_handle_flip(rdev, 1); + rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT; DRM_DEBUG("IH: D2 vblank\n"); } break; case 1: /* D2 vline */ - if (disp_int_cont & LB_D2_VLINE_INTERRUPT) { - disp_int_cont &= ~LB_D2_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VLINE_INTERRUPT; DRM_DEBUG("IH: D2 vline\n"); } break; @@ -1862,16 +2697,21 @@ restart_ih: case 3: /* D3 vblank/vline */ switch (src_data) { case 0: /* D3 vblank */ - if (disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 2); - wake_up(&rdev->irq.vblank_queue); - disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[2]) { + drm_handle_vblank(rdev->ddev, 2); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[2]) + radeon_crtc_handle_flip(rdev, 2); + rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT; DRM_DEBUG("IH: D3 vblank\n"); } break; case 1: /* D3 vline */ - if (disp_int_cont2 & LB_D3_VLINE_INTERRUPT) { - disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT; DRM_DEBUG("IH: D3 vline\n"); } break; @@ -1883,16 +2723,21 @@ restart_ih: case 4: /* D4 vblank/vline */ switch (src_data) { case 0: /* D4 vblank */ - if (disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 3); - wake_up(&rdev->irq.vblank_queue); - disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[3]) { + drm_handle_vblank(rdev->ddev, 3); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[3]) + radeon_crtc_handle_flip(rdev, 3); + rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT; DRM_DEBUG("IH: D4 vblank\n"); } break; case 1: /* D4 vline */ - if (disp_int_cont3 & LB_D4_VLINE_INTERRUPT) { - disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT; DRM_DEBUG("IH: D4 vline\n"); } break; @@ -1904,16 +2749,21 @@ restart_ih: case 5: /* D5 vblank/vline */ switch (src_data) { case 0: /* D5 vblank */ - if (disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 4); - wake_up(&rdev->irq.vblank_queue); - disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[4]) { + drm_handle_vblank(rdev->ddev, 4); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[4]) + radeon_crtc_handle_flip(rdev, 4); + rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT; DRM_DEBUG("IH: D5 vblank\n"); } break; case 1: /* D5 vline */ - if (disp_int_cont4 & LB_D5_VLINE_INTERRUPT) { - disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT; DRM_DEBUG("IH: D5 vline\n"); } break; @@ -1925,16 +2775,21 @@ restart_ih: case 6: /* D6 vblank/vline */ switch (src_data) { case 0: /* D6 vblank */ - if (disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) { - drm_handle_vblank(rdev->ddev, 5); - wake_up(&rdev->irq.vblank_queue); - disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[5]) { + drm_handle_vblank(rdev->ddev, 5); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (rdev->irq.pflip[5]) + radeon_crtc_handle_flip(rdev, 5); + rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT; DRM_DEBUG("IH: D6 vblank\n"); } break; case 1: /* D6 vline */ - if (disp_int_cont5 & LB_D6_VLINE_INTERRUPT) { - disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT; DRM_DEBUG("IH: D6 vline\n"); } break; @@ -1946,43 +2801,43 @@ restart_ih: case 42: /* HPD hotplug */ switch (src_data) { case 0: - if (disp_int & DC_HPD1_INTERRUPT) { - disp_int &= ~DC_HPD1_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int &= ~DC_HPD1_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD1\n"); } break; case 1: - if (disp_int_cont & DC_HPD2_INTERRUPT) { - disp_int_cont &= ~DC_HPD2_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont &= ~DC_HPD2_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD2\n"); } break; case 2: - if (disp_int_cont2 & DC_HPD3_INTERRUPT) { - disp_int_cont2 &= ~DC_HPD3_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~DC_HPD3_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD3\n"); } break; case 3: - if (disp_int_cont3 & DC_HPD4_INTERRUPT) { - disp_int_cont3 &= ~DC_HPD4_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~DC_HPD4_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD4\n"); } break; case 4: - if (disp_int_cont4 & DC_HPD5_INTERRUPT) { - disp_int_cont4 &= ~DC_HPD5_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~DC_HPD5_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD5\n"); } break; case 5: - if (disp_int_cont5 & DC_HPD6_INTERRUPT) { - disp_int_cont5 &= ~DC_HPD6_INTERRUPT; + if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT) { + rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~DC_HPD6_INTERRUPT; queue_hotplug = true; DRM_DEBUG("IH: HPD6\n"); } @@ -2000,6 +2855,7 @@ restart_ih: break; case 181: /* CP EOP event */ DRM_DEBUG("IH: CP EOP\n"); + radeon_fence_process(rdev); break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: CP EOP\n"); @@ -2020,7 +2876,7 @@ restart_ih: if (wptr != rdev->ih.wptr) goto restart_ih; if (queue_hotplug) - queue_work(rdev->wq, &rdev->hotplug_work); + schedule_work(&rdev->hotplug_work); rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); spin_unlock_irqrestore(&rdev->ih.lock, flags); @@ -2031,12 +2887,31 @@ static int evergreen_startup(struct radeon_device *rdev) { int r; - if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { - r = r600_init_microcode(rdev); + /* enable pcie gen2 link */ + if (!ASIC_IS_DCE5(rdev)) + evergreen_pcie_gen2_enable(rdev); + + if (ASIC_IS_DCE5(rdev)) { + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw || !rdev->mc_fw) { + r = ni_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + r = btc_mc_load_microcode(rdev); if (r) { - DRM_ERROR("Failed to load firmware!\n"); + DRM_ERROR("Failed to load MC firmware!\n"); return r; } + } else { + 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; + } + } } evergreen_mc_program(rdev); @@ -2048,26 +2923,23 @@ static int evergreen_startup(struct radeon_device *rdev) return r; } evergreen_gpu_init(rdev); -#if 0 - if (!rdev->r600_blit.shader_obj) { - r = r600_blit_init(rdev); - if (r) { - DRM_ERROR("radeon: failed blitter (%d).\n", r); - return r; - } - } - r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); - if (unlikely(r != 0)) - return r; - r = radeon_bo_pin(rdev->r600_blit.shader_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->r600_blit.shader_gpu_addr); - radeon_bo_unreserve(rdev->r600_blit.shader_obj); + r = evergreen_blit_init(rdev); if (r) { - DRM_ERROR("failed to pin blit object %d\n", r); - return r; + evergreen_blit_fini(rdev); + rdev->asic->copy = NULL; + dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); + } + /* XXX: ontario has problems blitting to gart at the moment */ + if (rdev->family == CHIP_PALM) { + rdev->asic->copy = NULL; + radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); } -#endif + + /* allocate wb buffer */ + r = radeon_wb_init(rdev); + if (r) + return r; /* Enable IRQ */ r = r600_irq_init(rdev); @@ -2087,8 +2959,6 @@ static int evergreen_startup(struct radeon_device *rdev) r = evergreen_cp_resume(rdev); if (r) return r; - /* write back buffer are not vital so don't worry about failure */ - r600_wb_enable(rdev); return 0; } @@ -2097,6 +2967,11 @@ int evergreen_resume(struct radeon_device *rdev) { int r; + /* reset the asic, the gfx blocks are often in a bad state + * after the driver is unloaded or after a resume + */ + if (radeon_asic_reset(rdev)) + dev_warn(rdev->dev, "GPU reset failed !\n"); /* Do not reset GPU before posting, on rv770 hw unlike on r500 hw, * posting will perform necessary task to bring back GPU into good * shape. @@ -2122,45 +2997,44 @@ int evergreen_resume(struct radeon_device *rdev) int evergreen_suspend(struct radeon_device *rdev) { -#if 0 int r; -#endif + /* FIXME: we should wait for ring to be empty */ r700_cp_stop(rdev); rdev->cp.ready = false; evergreen_irq_suspend(rdev); - r600_wb_disable(rdev); + radeon_wb_disable(rdev); evergreen_pcie_gart_disable(rdev); -#if 0 + /* unpin shaders bo */ r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); if (likely(r == 0)) { radeon_bo_unpin(rdev->r600_blit.shader_obj); radeon_bo_unreserve(rdev->r600_blit.shader_obj); } -#endif + return 0; } -static bool evergreen_card_posted(struct radeon_device *rdev) +int evergreen_copy_blit(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_pages, struct radeon_fence *fence) { - u32 reg; - - /* first check CRTCs */ - reg = RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET) | - RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET) | - RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET) | - RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET) | - RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET) | - RREG32(EVERGREEN_CRTC_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET); - if (reg & EVERGREEN_CRTC_MASTER_EN) - return true; - - /* then check MEM_SIZE, in case the crtcs are off */ - if (RREG32(CONFIG_MEMSIZE)) - return true; + int r; - return false; + mutex_lock(&rdev->r600_blit.mutex); + rdev->r600_blit.vb_ib = NULL; + r = evergreen_blit_prepare_copy(rdev, num_pages * RADEON_GPU_PAGE_SIZE); + if (r) { + if (rdev->r600_blit.vb_ib) + radeon_ib_free(rdev, &rdev->r600_blit.vb_ib); + mutex_unlock(&rdev->r600_blit.mutex); + return r; + } + evergreen_kms_blit_copy(rdev, src_offset, dst_offset, num_pages * RADEON_GPU_PAGE_SIZE); + evergreen_blit_done_copy(rdev, fence); + mutex_unlock(&rdev->r600_blit.mutex); + return 0; } /* Plan is to move initialization in that function and use @@ -2193,8 +3067,13 @@ int evergreen_init(struct radeon_device *rdev) r = radeon_atombios_init(rdev); if (r) return r; + /* reset the asic, the gfx blocks are often in a bad state + * after the driver is unloaded or after a resume + */ + if (radeon_asic_reset(rdev)) + dev_warn(rdev->dev, "GPU reset failed !\n"); /* Post card if necessary */ - if (!evergreen_card_posted(rdev)) { + if (!radeon_card_posted(rdev)) { if (!rdev->bios) { dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n"); return -EINVAL; @@ -2246,8 +3125,8 @@ int evergreen_init(struct radeon_device *rdev) if (r) { dev_err(rdev->dev, "disabling GPU acceleration\n"); r700_cp_fini(rdev); - r600_wb_fini(rdev); r600_irq_fini(rdev); + radeon_wb_fini(rdev); radeon_irq_kms_fini(rdev); evergreen_pcie_gart_fini(rdev); rdev->accel_working = false; @@ -2269,10 +3148,10 @@ int evergreen_init(struct radeon_device *rdev) void evergreen_fini(struct radeon_device *rdev) { - /*r600_blit_fini(rdev);*/ + evergreen_blit_fini(rdev); r700_cp_fini(rdev); - r600_wb_fini(rdev); r600_irq_fini(rdev); + radeon_wb_fini(rdev); radeon_irq_kms_fini(rdev); evergreen_pcie_gart_fini(rdev); radeon_gem_fini(rdev); @@ -2284,3 +3163,55 @@ void evergreen_fini(struct radeon_device *rdev) rdev->bios = NULL; radeon_dummy_page_fini(rdev); } + +static void evergreen_pcie_gen2_enable(struct radeon_device *rdev) +{ + u32 link_width_cntl, speed_cntl; + + if (radeon_pcie_gen2 == 0) + return; + + if (rdev->flags & RADEON_IS_IGP) + return; + + if (!(rdev->flags & RADEON_IS_PCIE)) + return; + + /* x2 cards have a special sequence */ + if (ASIC_IS_X2(rdev)) + return; + + speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); + if ((speed_cntl & LC_OTHER_SIDE_EVER_SENT_GEN2) || + (speed_cntl & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + + link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL); + link_width_cntl &= ~LC_UPCONFIGURE_DIS; + WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); + + speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); + speed_cntl &= ~LC_TARGET_LINK_SPEED_OVERRIDE_EN; + WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); + + speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); + speed_cntl |= LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); + + speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); + speed_cntl &= ~LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); + + speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); + speed_cntl |= LC_GEN2_EN_STRAP; + WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); + + } else { + link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL); + /* XXX: only disable it if gen1 bridge vendor == 0x111d or 0x1106 */ + if (1) + link_width_cntl |= LC_UPCONFIGURE_DIS; + else + link_width_cntl &= ~LC_UPCONFIGURE_DIS; + WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); + } +}