From 2858c00d2823c83acce2a1175dbabb2cebee8678 Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 1 Aug 2013 17:34:07 +0200 Subject: drm/radeon: fix halting UVD MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Removing the clock/power or resetting the VCPU can cause hangs if that happens in the middle of a register write. Stall the memory and register bus before putting the VCPU into reset. Keep it in reset when unloading the module or suspending. Signed-off-by: Christian König Cc: stable@vger.kernel.org Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/evergreen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 038dcac7670c..5b98e573d60a 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5291,10 +5291,10 @@ int evergreen_resume(struct radeon_device *rdev) int evergreen_suspend(struct radeon_device *rdev) { r600_audio_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_suspend(rdev); r700_cp_stop(rdev); r600_dma_stop(rdev); - r600_uvd_rbc_stop(rdev); evergreen_irq_suspend(rdev); radeon_wb_disable(rdev); evergreen_pcie_gart_disable(rdev); @@ -5429,6 +5429,7 @@ void evergreen_fini(struct radeon_device *rdev) radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); evergreen_pcie_gart_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); -- cgit v1.2.3 From 6fab3febf6d949b0a12b1e4e73db38e4a177a79e Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sun, 4 Aug 2013 12:13:17 -0400 Subject: drm/radeon: always program the MC on startup For r6xx+ asics. This mirrors the behavior of pre-r6xx asics. We need to program the MC even if something else in startup() fails. Failure to do so results in an unusable GPU. Based on a fix from: Mark Kettenis Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/radeon/cik.c | 3 ++- drivers/gpu/drm/radeon/evergreen.c | 3 ++- drivers/gpu/drm/radeon/ni.c | 3 ++- drivers/gpu/drm/radeon/r600.c | 3 ++- drivers/gpu/drm/radeon/rv770.c | 3 ++- drivers/gpu/drm/radeon/si.c | 3 ++- 6 files changed, 12 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 524db70aaf6e..5b587876c7f5 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -5954,6 +5954,8 @@ static int cik_startup(struct radeon_device *rdev) struct radeon_ring *ring; int r; + cik_mc_program(rdev); + if (rdev->flags & RADEON_IS_IGP) { if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || !rdev->mec_fw || !rdev->sdma_fw || !rdev->rlc_fw) { @@ -5985,7 +5987,6 @@ static int cik_startup(struct radeon_device *rdev) if (r) return r; - cik_mc_program(rdev); r = cik_pcie_gart_enable(rdev); if (r) return r; diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 5b98e573d60a..d5b49e33315e 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5106,6 +5106,8 @@ static int evergreen_startup(struct radeon_device *rdev) /* enable aspm */ evergreen_program_aspm(rdev); + evergreen_mc_program(rdev); + if (ASIC_IS_DCE5(rdev)) { if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw || !rdev->mc_fw) { r = ni_init_microcode(rdev); @@ -5133,7 +5135,6 @@ static int evergreen_startup(struct radeon_device *rdev) if (r) return r; - evergreen_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { evergreen_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 16e91b08bf57..12cebe46e34f 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2079,6 +2079,8 @@ static int cayman_startup(struct radeon_device *rdev) /* enable aspm */ evergreen_program_aspm(rdev); + evergreen_mc_program(rdev); + if (rdev->flags & RADEON_IS_IGP) { if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { r = ni_init_microcode(rdev); @@ -2107,7 +2109,6 @@ static int cayman_startup(struct radeon_device *rdev) if (r) return r; - evergreen_mc_program(rdev); r = cayman_pcie_gart_enable(rdev); if (r) return r; diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 0a9553abec54..af848045670c 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3330,6 +3330,8 @@ static int r600_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ r600_pcie_gen2_enable(rdev); + r600_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { r = r600_init_microcode(rdev); if (r) { @@ -3342,7 +3344,6 @@ static int r600_startup(struct radeon_device *rdev) if (r) return r; - r600_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { r600_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index f1010131bac0..3cc08a4d99d9 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1829,6 +1829,8 @@ static int rv770_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ rv770_pcie_gen2_enable(rdev); + rv770_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { r = r600_init_microcode(rdev); if (r) { @@ -1841,7 +1843,6 @@ static int rv770_startup(struct radeon_device *rdev) if (r) return r; - rv770_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { rv770_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 242c1ac83e23..6a2dca4acfc1 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6418,6 +6418,8 @@ static int si_startup(struct radeon_device *rdev) /* enable aspm */ si_program_aspm(rdev); + si_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || !rdev->rlc_fw || !rdev->mc_fw) { r = si_init_microcode(rdev); @@ -6437,7 +6439,6 @@ static int si_startup(struct radeon_device *rdev) if (r) return r; - si_mc_program(rdev); r = si_pcie_gart_enable(rdev); if (r) return r; -- cgit v1.2.3 From 4f8629675800505c274bf7d17baefed197a76cd9 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 5 Aug 2013 12:37:32 -0400 Subject: drm/radeon/kms: remove r6xx+ blit copy routines No longer used now that we use the async dma engines or CP DMA for bo copies. Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/Makefile | 4 +- drivers/gpu/drm/radeon/cayman_blit_shaders.c | 54 -- drivers/gpu/drm/radeon/evergreen.c | 8 - drivers/gpu/drm/radeon/evergreen_blit_kms.c | 729 ---------------------- drivers/gpu/drm/radeon/evergreen_blit_shaders.c | 54 -- drivers/gpu/drm/radeon/ni.c | 8 - drivers/gpu/drm/radeon/r600.c | 26 - drivers/gpu/drm/radeon/r600_blit.c | 31 + drivers/gpu/drm/radeon/r600_blit_kms.c | 785 ------------------------ drivers/gpu/drm/radeon/r600_blit_shaders.h | 1 - drivers/gpu/drm/radeon/radeon.h | 30 - drivers/gpu/drm/radeon/radeon_asic.h | 16 - drivers/gpu/drm/radeon/radeon_blit_common.h | 44 -- drivers/gpu/drm/radeon/rv770.c | 7 - 14 files changed, 33 insertions(+), 1764 deletions(-) delete mode 100644 drivers/gpu/drm/radeon/evergreen_blit_kms.c delete mode 100644 drivers/gpu/drm/radeon/r600_blit_kms.c delete mode 100644 drivers/gpu/drm/radeon/radeon_blit_common.h (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index c3df52c1a60c..bfabd69b4e39 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -72,8 +72,8 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \ radeon_cs.o radeon_bios.o radeon_benchmark.o r100.o r300.o r420.o \ rs400.o rs600.o rs690.o rv515.o r520.o r600.o rv770.o radeon_test.o \ r200.o radeon_legacy_tv.o r600_cs.o r600_blit_shaders.o \ - r600_blit_kms.o radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \ - evergreen.o evergreen_cs.o evergreen_blit_shaders.o evergreen_blit_kms.o \ + radeon_pm.o atombios_dp.o r600_audio.o r600_hdmi.o \ + evergreen.o evergreen_cs.o evergreen_blit_shaders.o \ evergreen_hdmi.o radeon_trace_points.o ni.o cayman_blit_shaders.o \ atombios_encoders.o radeon_semaphore.o radeon_sa.o atombios_i2c.o si.o \ si_blit_shaders.o radeon_prime.o radeon_uvd.o cik.o cik_blit_shaders.o \ diff --git a/drivers/gpu/drm/radeon/cayman_blit_shaders.c b/drivers/gpu/drm/radeon/cayman_blit_shaders.c index 19a0114d2e3b..98d009e154bf 100644 --- a/drivers/gpu/drm/radeon/cayman_blit_shaders.c +++ b/drivers/gpu/drm/radeon/cayman_blit_shaders.c @@ -317,58 +317,4 @@ const u32 cayman_default_state[] = 0x00000010, /* */ }; -const u32 cayman_vs[] = -{ - 0x00000004, - 0x80400400, - 0x0000a03c, - 0x95000688, - 0x00004000, - 0x15000688, - 0x00000000, - 0x88000000, - 0x04000000, - 0x67961001, -#ifdef __BIG_ENDIAN - 0x00020000, -#else - 0x00000000, -#endif - 0x00000000, - 0x04000000, - 0x67961000, -#ifdef __BIG_ENDIAN - 0x00020008, -#else - 0x00000008, -#endif - 0x00000000, -}; - -const u32 cayman_ps[] = -{ - 0x00000004, - 0xa00c0000, - 0x00000008, - 0x80400000, - 0x00000000, - 0x95000688, - 0x00000000, - 0x88000000, - 0x00380400, - 0x00146b10, - 0x00380000, - 0x20146b10, - 0x00380400, - 0x40146b00, - 0x80380000, - 0x60146b00, - 0x00000010, - 0x000d1000, - 0xb0800000, - 0x00000000, -}; - -const u32 cayman_ps_size = ARRAY_SIZE(cayman_ps); -const u32 cayman_vs_size = ARRAY_SIZE(cayman_vs); const u32 cayman_default_size = ARRAY_SIZE(cayman_default_state); diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index d5b49e33315e..a5ab5693eb2a 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5144,13 +5144,6 @@ static int evergreen_startup(struct radeon_device *rdev) } evergreen_gpu_init(rdev); - r = evergreen_blit_init(rdev); - if (r) { - r600_blit_fini(rdev); - rdev->asic->copy.copy = NULL; - dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); - } - /* allocate rlc buffers */ if (rdev->flags & RADEON_IS_IGP) { rdev->rlc.reg_list = sumo_rlc_save_restore_register_list; @@ -5420,7 +5413,6 @@ int evergreen_init(struct radeon_device *rdev) void evergreen_fini(struct radeon_device *rdev) { r600_audio_fini(rdev); - r600_blit_fini(rdev); r700_cp_fini(rdev); r600_dma_fini(rdev); r600_irq_fini(rdev); diff --git a/drivers/gpu/drm/radeon/evergreen_blit_kms.c b/drivers/gpu/drm/radeon/evergreen_blit_kms.c deleted file mode 100644 index 057c87b6515a..000000000000 --- a/drivers/gpu/drm/radeon/evergreen_blit_kms.c +++ /dev/null @@ -1,729 +0,0 @@ -/* - * Copyright 2010 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - * Authors: - * Alex Deucher - */ - -#include -#include -#include "radeon.h" - -#include "evergreend.h" -#include "evergreen_blit_shaders.h" -#include "cayman_blit_shaders.h" -#include "radeon_blit_common.h" - -/* emits 17 */ -static void -set_render_target(struct radeon_device *rdev, int format, - int w, int h, u64 gpu_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 cb_color_info; - int pitch, slice; - - h = ALIGN(h, 8); - if (h < 8) - h = 8; - - cb_color_info = CB_FORMAT(format) | - CB_SOURCE_FORMAT(CB_SF_EXPORT_NORM) | - CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); - pitch = (w / 8) - 1; - slice = ((w * h) / 64) - 1; - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 15)); - radeon_ring_write(ring, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, pitch); - radeon_ring_write(ring, slice); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, cb_color_info); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, (w - 1) | ((h - 1) << 16)); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); -} - -/* emits 5dw */ -static void -cp_set_surface_sync(struct radeon_device *rdev, - u32 sync_type, u32 size, - u64 mc_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 cp_coher_size; - - if (size == 0xffffffff) - cp_coher_size = 0xffffffff; - else - cp_coher_size = ((size + 255) >> 8); - - if (rdev->family >= CHIP_CAYMAN) { - /* CP_COHER_CNTL2 has to be set manually when submitting a surface_sync - * to the RB directly. For IBs, the CP programs this as part of the - * surface_sync packet. - */ - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(ring, (0x85e8 - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, 0); /* CP_COHER_CNTL2 */ - } - radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)); - radeon_ring_write(ring, sync_type); - radeon_ring_write(ring, cp_coher_size); - radeon_ring_write(ring, mc_addr >> 8); - radeon_ring_write(ring, 10); /* poll interval */ -} - -/* emits 11dw + 1 surface sync = 16dw */ -static void -set_shaders(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u64 gpu_addr; - - /* VS */ - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 3)); - radeon_ring_write(ring, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, 2); - radeon_ring_write(ring, 0); - - /* PS */ - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 4)); - radeon_ring_write(ring, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, 1); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 2); - - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset; - cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr); -} - -/* emits 10 + 1 sync (5) = 15 */ -static void -set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 sq_vtx_constant_word2, sq_vtx_constant_word3; - - /* high addr, stride */ - sq_vtx_constant_word2 = SQ_VTXC_BASE_ADDR_HI(upper_32_bits(gpu_addr) & 0xff) | - SQ_VTXC_STRIDE(16); -#ifdef __BIG_ENDIAN - sq_vtx_constant_word2 |= SQ_VTXC_ENDIAN_SWAP(SQ_ENDIAN_8IN32); -#endif - /* xyzw swizzles */ - sq_vtx_constant_word3 = SQ_VTCX_SEL_X(SQ_SEL_X) | - SQ_VTCX_SEL_Y(SQ_SEL_Y) | - SQ_VTCX_SEL_Z(SQ_SEL_Z) | - SQ_VTCX_SEL_W(SQ_SEL_W); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 8)); - radeon_ring_write(ring, 0x580); - radeon_ring_write(ring, gpu_addr & 0xffffffff); - radeon_ring_write(ring, 48 - 1); /* size */ - radeon_ring_write(ring, sq_vtx_constant_word2); - radeon_ring_write(ring, sq_vtx_constant_word3); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_BUFFER)); - - if ((rdev->family == CHIP_CEDAR) || - (rdev->family == CHIP_PALM) || - (rdev->family == CHIP_SUMO) || - (rdev->family == CHIP_SUMO2) || - (rdev->family == CHIP_CAICOS)) - cp_set_surface_sync(rdev, - PACKET3_TC_ACTION_ENA, 48, gpu_addr); - else - cp_set_surface_sync(rdev, - PACKET3_VC_ACTION_ENA, 48, gpu_addr); - -} - -/* emits 10 */ -static void -set_tex_resource(struct radeon_device *rdev, - int format, int w, int h, int pitch, - u64 gpu_addr, u32 size) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 sq_tex_resource_word0, sq_tex_resource_word1; - u32 sq_tex_resource_word4, sq_tex_resource_word7; - - if (h < 1) - h = 1; - - sq_tex_resource_word0 = TEX_DIM(SQ_TEX_DIM_2D); - sq_tex_resource_word0 |= ((((pitch >> 3) - 1) << 6) | - ((w - 1) << 18)); - sq_tex_resource_word1 = ((h - 1) << 0) | - TEX_ARRAY_MODE(ARRAY_1D_TILED_THIN1); - /* xyzw swizzles */ - sq_tex_resource_word4 = TEX_DST_SEL_X(SQ_SEL_X) | - TEX_DST_SEL_Y(SQ_SEL_Y) | - TEX_DST_SEL_Z(SQ_SEL_Z) | - TEX_DST_SEL_W(SQ_SEL_W); - - sq_tex_resource_word7 = format | - S__SQ_CONSTANT_TYPE(SQ_TEX_VTX_VALID_TEXTURE); - - cp_set_surface_sync(rdev, - PACKET3_TC_ACTION_ENA, size, gpu_addr); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 8)); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, sq_tex_resource_word0); - radeon_ring_write(ring, sq_tex_resource_word1); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, sq_tex_resource_word4); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, sq_tex_resource_word7); -} - -/* emits 12 */ -static void -set_scissors(struct radeon_device *rdev, int x1, int y1, - int x2, int y2) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - /* workaround some hw bugs */ - if (x2 == 0) - x1 = 1; - if (y2 == 0) - y1 = 1; - if (rdev->family >= CHIP_CAYMAN) { - if ((x2 == 1) && (y2 == 1)) - x2 = 2; - } - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_START) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); -} - -/* emits 10 */ -static void -draw_auto(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(ring, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, DI_PT_RECTLIST); - - radeon_ring_write(ring, PACKET3(PACKET3_INDEX_TYPE, 0)); - radeon_ring_write(ring, -#ifdef __BIG_ENDIAN - (2 << 2) | -#endif - DI_INDEX_SIZE_16_BIT); - - radeon_ring_write(ring, PACKET3(PACKET3_NUM_INSTANCES, 0)); - radeon_ring_write(ring, 1); - - radeon_ring_write(ring, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1)); - radeon_ring_write(ring, 3); - radeon_ring_write(ring, DI_SRC_SEL_AUTO_INDEX); - -} - -/* emits 39 */ -static void -set_default_state(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2, sq_gpr_resource_mgmt_3; - u32 sq_thread_resource_mgmt, sq_thread_resource_mgmt_2; - u32 sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2, sq_stack_resource_mgmt_3; - int num_ps_gprs, num_vs_gprs, num_temp_gprs; - int num_gs_gprs, num_es_gprs, num_hs_gprs, num_ls_gprs; - int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads; - int num_hs_threads, num_ls_threads; - int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries; - int num_hs_stack_entries, num_ls_stack_entries; - u64 gpu_addr; - int dwords; - - /* set clear context state */ - radeon_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0)); - radeon_ring_write(ring, 0); - - if (rdev->family < CHIP_CAYMAN) { - switch (rdev->family) { - case CHIP_CEDAR: - default: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 96; - num_vs_threads = 16; - num_gs_threads = 16; - num_es_threads = 16; - num_hs_threads = 16; - num_ls_threads = 16; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - case CHIP_REDWOOD: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 20; - num_gs_threads = 20; - num_es_threads = 20; - num_hs_threads = 20; - num_ls_threads = 20; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - case CHIP_JUNIPER: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 20; - num_gs_threads = 20; - num_es_threads = 20; - num_hs_threads = 20; - num_ls_threads = 20; - num_ps_stack_entries = 85; - num_vs_stack_entries = 85; - num_gs_stack_entries = 85; - num_es_stack_entries = 85; - num_hs_stack_entries = 85; - num_ls_stack_entries = 85; - break; - case CHIP_CYPRESS: - case CHIP_HEMLOCK: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 20; - num_gs_threads = 20; - num_es_threads = 20; - num_hs_threads = 20; - num_ls_threads = 20; - num_ps_stack_entries = 85; - num_vs_stack_entries = 85; - num_gs_stack_entries = 85; - num_es_stack_entries = 85; - num_hs_stack_entries = 85; - num_ls_stack_entries = 85; - break; - case CHIP_PALM: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 96; - num_vs_threads = 16; - num_gs_threads = 16; - num_es_threads = 16; - num_hs_threads = 16; - num_ls_threads = 16; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - case CHIP_SUMO: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 96; - num_vs_threads = 25; - num_gs_threads = 25; - num_es_threads = 25; - num_hs_threads = 25; - num_ls_threads = 25; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - case CHIP_SUMO2: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 96; - num_vs_threads = 25; - num_gs_threads = 25; - num_es_threads = 25; - num_hs_threads = 25; - num_ls_threads = 25; - num_ps_stack_entries = 85; - num_vs_stack_entries = 85; - num_gs_stack_entries = 85; - num_es_stack_entries = 85; - num_hs_stack_entries = 85; - num_ls_stack_entries = 85; - break; - case CHIP_BARTS: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 20; - num_gs_threads = 20; - num_es_threads = 20; - num_hs_threads = 20; - num_ls_threads = 20; - num_ps_stack_entries = 85; - num_vs_stack_entries = 85; - num_gs_stack_entries = 85; - num_es_stack_entries = 85; - num_hs_stack_entries = 85; - num_ls_stack_entries = 85; - break; - case CHIP_TURKS: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 20; - num_gs_threads = 20; - num_es_threads = 20; - num_hs_threads = 20; - num_ls_threads = 20; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - case CHIP_CAICOS: - num_ps_gprs = 93; - num_vs_gprs = 46; - num_temp_gprs = 4; - num_gs_gprs = 31; - num_es_gprs = 31; - num_hs_gprs = 23; - num_ls_gprs = 23; - num_ps_threads = 128; - num_vs_threads = 10; - num_gs_threads = 10; - num_es_threads = 10; - num_hs_threads = 10; - num_ls_threads = 10; - num_ps_stack_entries = 42; - num_vs_stack_entries = 42; - num_gs_stack_entries = 42; - num_es_stack_entries = 42; - num_hs_stack_entries = 42; - num_ls_stack_entries = 42; - break; - } - - if ((rdev->family == CHIP_CEDAR) || - (rdev->family == CHIP_PALM) || - (rdev->family == CHIP_SUMO) || - (rdev->family == CHIP_SUMO2) || - (rdev->family == CHIP_CAICOS)) - sq_config = 0; - else - sq_config = VC_ENABLE; - - sq_config |= (EXPORT_SRC_C | - CS_PRIO(0) | - LS_PRIO(0) | - HS_PRIO(0) | - PS_PRIO(0) | - VS_PRIO(1) | - GS_PRIO(2) | - ES_PRIO(3)); - - sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) | - NUM_VS_GPRS(num_vs_gprs) | - NUM_CLAUSE_TEMP_GPRS(num_temp_gprs)); - sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) | - NUM_ES_GPRS(num_es_gprs)); - sq_gpr_resource_mgmt_3 = (NUM_HS_GPRS(num_hs_gprs) | - NUM_LS_GPRS(num_ls_gprs)); - sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) | - NUM_VS_THREADS(num_vs_threads) | - NUM_GS_THREADS(num_gs_threads) | - NUM_ES_THREADS(num_es_threads)); - sq_thread_resource_mgmt_2 = (NUM_HS_THREADS(num_hs_threads) | - NUM_LS_THREADS(num_ls_threads)); - sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) | - NUM_VS_STACK_ENTRIES(num_vs_stack_entries)); - sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) | - NUM_ES_STACK_ENTRIES(num_es_stack_entries)); - sq_stack_resource_mgmt_3 = (NUM_HS_STACK_ENTRIES(num_hs_stack_entries) | - NUM_LS_STACK_ENTRIES(num_ls_stack_entries)); - - /* disable dyn gprs */ - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(ring, (SQ_DYN_GPR_CNTL_PS_FLUSH_REQ - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, 0); - - /* setup LDS */ - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(ring, (SQ_LDS_RESOURCE_MGMT - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, 0x10001000); - - /* SQ config */ - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 11)); - radeon_ring_write(ring, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_START) >> 2); - radeon_ring_write(ring, sq_config); - radeon_ring_write(ring, sq_gpr_resource_mgmt_1); - radeon_ring_write(ring, sq_gpr_resource_mgmt_2); - radeon_ring_write(ring, sq_gpr_resource_mgmt_3); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, sq_thread_resource_mgmt); - radeon_ring_write(ring, sq_thread_resource_mgmt_2); - radeon_ring_write(ring, sq_stack_resource_mgmt_1); - radeon_ring_write(ring, sq_stack_resource_mgmt_2); - radeon_ring_write(ring, sq_stack_resource_mgmt_3); - } - - /* CONTEXT_CONTROL */ - radeon_ring_write(ring, 0xc0012800); - radeon_ring_write(ring, 0x80000000); - radeon_ring_write(ring, 0x80000000); - - /* SQ_VTX_BASE_VTX_LOC */ - radeon_ring_write(ring, 0xc0026f00); - radeon_ring_write(ring, 0x00000000); - radeon_ring_write(ring, 0x00000000); - radeon_ring_write(ring, 0x00000000); - - /* SET_SAMPLER */ - radeon_ring_write(ring, 0xc0036e00); - radeon_ring_write(ring, 0x00000000); - radeon_ring_write(ring, 0x00000012); - radeon_ring_write(ring, 0x00000000); - radeon_ring_write(ring, 0x00000000); - - /* set to DX10/11 mode */ - radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0)); - radeon_ring_write(ring, 1); - - /* emit an IB pointing at default state */ - dwords = ALIGN(rdev->r600_blit.state_len, 0x10); - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset; - radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); - radeon_ring_write(ring, gpu_addr & 0xFFFFFFFC); - radeon_ring_write(ring, upper_32_bits(gpu_addr) & 0xFF); - radeon_ring_write(ring, dwords); - -} - -int evergreen_blit_init(struct radeon_device *rdev) -{ - u32 obj_size; - int i, r, dwords; - void *ptr; - u32 packet2s[16]; - int num_packet2s = 0; - - rdev->r600_blit.primitives.set_render_target = set_render_target; - rdev->r600_blit.primitives.cp_set_surface_sync = cp_set_surface_sync; - rdev->r600_blit.primitives.set_shaders = set_shaders; - rdev->r600_blit.primitives.set_vtx_resource = set_vtx_resource; - rdev->r600_blit.primitives.set_tex_resource = set_tex_resource; - rdev->r600_blit.primitives.set_scissors = set_scissors; - rdev->r600_blit.primitives.draw_auto = draw_auto; - rdev->r600_blit.primitives.set_default_state = set_default_state; - - rdev->r600_blit.ring_size_common = 8; /* sync semaphore */ - rdev->r600_blit.ring_size_common += 55; /* shaders + def state */ - rdev->r600_blit.ring_size_common += 16; /* fence emit for VB IB */ - rdev->r600_blit.ring_size_common += 5; /* done copy */ - rdev->r600_blit.ring_size_common += 16; /* fence emit for done copy */ - - rdev->r600_blit.ring_size_per_loop = 74; - if (rdev->family >= CHIP_CAYMAN) - rdev->r600_blit.ring_size_per_loop += 9; /* additional DWs for surface sync */ - - rdev->r600_blit.max_dim = 16384; - - rdev->r600_blit.state_offset = 0; - - if (rdev->family < CHIP_CAYMAN) - rdev->r600_blit.state_len = evergreen_default_size; - else - rdev->r600_blit.state_len = cayman_default_size; - - dwords = rdev->r600_blit.state_len; - while (dwords & 0xf) { - packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0)); - dwords++; - } - - obj_size = dwords * 4; - obj_size = ALIGN(obj_size, 256); - - rdev->r600_blit.vs_offset = obj_size; - if (rdev->family < CHIP_CAYMAN) - obj_size += evergreen_vs_size * 4; - else - obj_size += cayman_vs_size * 4; - obj_size = ALIGN(obj_size, 256); - - rdev->r600_blit.ps_offset = obj_size; - if (rdev->family < CHIP_CAYMAN) - obj_size += evergreen_ps_size * 4; - else - obj_size += cayman_ps_size * 4; - obj_size = ALIGN(obj_size, 256); - - /* pin copy shader into vram if not already initialized */ - if (!rdev->r600_blit.shader_obj) { - r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, - NULL, &rdev->r600_blit.shader_obj); - if (r) { - DRM_ERROR("evergreen failed to allocate shader\n"); - 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); - if (r) { - dev_err(rdev->dev, "(%d) pin blit object failed\n", r); - return r; - } - } - - DRM_DEBUG("evergreen blit allocated bo %08x vs %08x ps %08x\n", - obj_size, - rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset); - - r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); - if (unlikely(r != 0)) - return r; - r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr); - if (r) { - DRM_ERROR("failed to map blit object %d\n", r); - return r; - } - - if (rdev->family < CHIP_CAYMAN) { - memcpy_toio(ptr + rdev->r600_blit.state_offset, - evergreen_default_state, rdev->r600_blit.state_len * 4); - - if (num_packet2s) - memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4), - packet2s, num_packet2s * 4); - for (i = 0; i < evergreen_vs_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(evergreen_vs[i]); - for (i = 0; i < evergreen_ps_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(evergreen_ps[i]); - } else { - memcpy_toio(ptr + rdev->r600_blit.state_offset, - cayman_default_state, rdev->r600_blit.state_len * 4); - - if (num_packet2s) - memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4), - packet2s, num_packet2s * 4); - for (i = 0; i < cayman_vs_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(cayman_vs[i]); - for (i = 0; i < cayman_ps_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(cayman_ps[i]); - } - radeon_bo_kunmap(rdev->r600_blit.shader_obj); - radeon_bo_unreserve(rdev->r600_blit.shader_obj); - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); - return 0; -} diff --git a/drivers/gpu/drm/radeon/evergreen_blit_shaders.c b/drivers/gpu/drm/radeon/evergreen_blit_shaders.c index f85c0af115b5..d43383470cdf 100644 --- a/drivers/gpu/drm/radeon/evergreen_blit_shaders.c +++ b/drivers/gpu/drm/radeon/evergreen_blit_shaders.c @@ -300,58 +300,4 @@ const u32 evergreen_default_state[] = 0x00000010, /* */ }; -const u32 evergreen_vs[] = -{ - 0x00000004, - 0x80800400, - 0x0000a03c, - 0x95000688, - 0x00004000, - 0x15200688, - 0x00000000, - 0x00000000, - 0x3c000000, - 0x67961001, -#ifdef __BIG_ENDIAN - 0x000a0000, -#else - 0x00080000, -#endif - 0x00000000, - 0x1c000000, - 0x67961000, -#ifdef __BIG_ENDIAN - 0x00020008, -#else - 0x00000008, -#endif - 0x00000000, -}; - -const u32 evergreen_ps[] = -{ - 0x00000003, - 0xa00c0000, - 0x00000008, - 0x80400000, - 0x00000000, - 0x95200688, - 0x00380400, - 0x00146b10, - 0x00380000, - 0x20146b10, - 0x00380400, - 0x40146b00, - 0x80380000, - 0x60146b00, - 0x00000000, - 0x00000000, - 0x00000010, - 0x000d1000, - 0xb0800000, - 0x00000000, -}; - -const u32 evergreen_ps_size = ARRAY_SIZE(evergreen_ps); -const u32 evergreen_vs_size = ARRAY_SIZE(evergreen_vs); const u32 evergreen_default_size = ARRAY_SIZE(evergreen_default_state); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index ccb4f8b54852..acdd6039ef14 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2118,13 +2118,6 @@ static int cayman_startup(struct radeon_device *rdev) return r; cayman_gpu_init(rdev); - r = evergreen_blit_init(rdev); - if (r) { - r600_blit_fini(rdev); - rdev->asic->copy.copy = NULL; - dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); - } - /* allocate rlc buffers */ if (rdev->flags & RADEON_IS_IGP) { rdev->rlc.reg_list = tn_rlc_save_restore_register_list; @@ -2413,7 +2406,6 @@ int cayman_init(struct radeon_device *rdev) void cayman_fini(struct radeon_device *rdev) { - r600_blit_fini(rdev); cayman_cp_fini(rdev); cayman_dma_fini(rdev); r600_irq_fini(rdev); diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index e66e72077350..3db2e4ddb2d6 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -3136,25 +3136,6 @@ void r600_uvd_semaphore_emit(struct radeon_device *rdev, radeon_ring_write(ring, emit_wait ? 1 : 0); } -int r600_copy_blit(struct radeon_device *rdev, - uint64_t src_offset, - uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - struct radeon_sa_bo *vb = NULL; - int r; - - r = r600_blit_prepare_copy(rdev, num_gpu_pages, fence, &vb, &sem); - if (r) { - return r; - } - r600_kms_blit_copy(rdev, src_offset, dst_offset, num_gpu_pages, vb); - r600_blit_done_copy(rdev, fence, vb, sem); - return 0; -} - /** * r600_copy_cpdma - copy pages using the CP DMA engine * @@ -3356,12 +3337,6 @@ static int r600_startup(struct radeon_device *rdev) return r; } r600_gpu_init(rdev); - r = r600_blit_init(rdev); - if (r) { - r600_blit_fini(rdev); - rdev->asic->copy.copy = NULL; - dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); - } /* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -3574,7 +3549,6 @@ int r600_init(struct radeon_device *rdev) void r600_fini(struct radeon_device *rdev) { r600_audio_fini(rdev); - r600_blit_fini(rdev); r600_cp_fini(rdev); r600_dma_fini(rdev); r600_irq_fini(rdev); diff --git a/drivers/gpu/drm/radeon/r600_blit.c b/drivers/gpu/drm/radeon/r600_blit.c index f651881eb0ae..daf7572be976 100644 --- a/drivers/gpu/drm/radeon/r600_blit.c +++ b/drivers/gpu/drm/radeon/r600_blit.c @@ -31,6 +31,37 @@ #include "r600_blit_shaders.h" +/* 23 bits of float fractional data */ +#define I2F_FRAC_BITS 23 +#define I2F_MASK ((1 << I2F_FRAC_BITS) - 1) + +/* + * Converts unsigned integer into 32-bit IEEE floating point representation. + * Will be exact from 0 to 2^24. Above that, we round towards zero + * as the fractional bits will not fit in a float. (It would be better to + * round towards even as the fpu does, but that is slower.) + */ +static __pure uint32_t int2float(uint32_t x) +{ + uint32_t msb, exponent, fraction; + + /* Zero is special */ + if (!x) return 0; + + /* Get location of the most significant bit */ + msb = __fls(x); + + /* + * Use a rotate instead of a shift because that works both leftwards + * and rightwards due to the mod(32) behaviour. This means we don't + * need to check to see if we are above 2^24 or not. + */ + fraction = ror32(x, (msb - I2F_FRAC_BITS) & 0x1f) & I2F_MASK; + exponent = (127 + msb) << I2F_FRAC_BITS; + + return fraction + exponent; +} + #define DI_PT_RECTLIST 0x11 #define DI_INDEX_SIZE_16_BIT 0x0 #define DI_SRC_SEL_AUTO_INDEX 0x2 diff --git a/drivers/gpu/drm/radeon/r600_blit_kms.c b/drivers/gpu/drm/radeon/r600_blit_kms.c deleted file mode 100644 index 9fb5780a552f..000000000000 --- a/drivers/gpu/drm/radeon/r600_blit_kms.c +++ /dev/null @@ -1,785 +0,0 @@ -/* - * Copyright 2009 Advanced Micro Devices, Inc. - * Copyright 2009 Red Hat Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - */ - -#include -#include -#include "radeon.h" - -#include "r600d.h" -#include "r600_blit_shaders.h" -#include "radeon_blit_common.h" - -/* 23 bits of float fractional data */ -#define I2F_FRAC_BITS 23 -#define I2F_MASK ((1 << I2F_FRAC_BITS) - 1) - -/* - * Converts unsigned integer into 32-bit IEEE floating point representation. - * Will be exact from 0 to 2^24. Above that, we round towards zero - * as the fractional bits will not fit in a float. (It would be better to - * round towards even as the fpu does, but that is slower.) - */ -__pure uint32_t int2float(uint32_t x) -{ - uint32_t msb, exponent, fraction; - - /* Zero is special */ - if (!x) return 0; - - /* Get location of the most significant bit */ - msb = __fls(x); - - /* - * Use a rotate instead of a shift because that works both leftwards - * and rightwards due to the mod(32) behaviour. This means we don't - * need to check to see if we are above 2^24 or not. - */ - fraction = ror32(x, (msb - I2F_FRAC_BITS) & 0x1f) & I2F_MASK; - exponent = (127 + msb) << I2F_FRAC_BITS; - - return fraction + exponent; -} - -/* emits 21 on rv770+, 23 on r600 */ -static void -set_render_target(struct radeon_device *rdev, int format, - int w, int h, u64 gpu_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 cb_color_info; - int pitch, slice; - - h = ALIGN(h, 8); - if (h < 8) - h = 8; - - cb_color_info = CB_FORMAT(format) | - CB_SOURCE_FORMAT(CB_SF_EXPORT_NORM) | - CB_ARRAY_MODE(ARRAY_1D_TILED_THIN1); - pitch = (w / 8) - 1; - slice = ((w * h) / 64) - 1; - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_BASE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - - if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) { - radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_BASE_UPDATE, 0)); - radeon_ring_write(ring, 2 << 0); - } - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_SIZE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, (pitch << 0) | (slice << 10)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_VIEW - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_INFO - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, cb_color_info); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_TILE - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_FRAG - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (CB_COLOR0_MASK - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); -} - -/* emits 5dw */ -static void -cp_set_surface_sync(struct radeon_device *rdev, - u32 sync_type, u32 size, - u64 mc_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 cp_coher_size; - - if (size == 0xffffffff) - cp_coher_size = 0xffffffff; - else - cp_coher_size = ((size + 255) >> 8); - - radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3)); - radeon_ring_write(ring, sync_type); - radeon_ring_write(ring, cp_coher_size); - radeon_ring_write(ring, mc_addr >> 8); - radeon_ring_write(ring, 10); /* poll interval */ -} - -/* emits 21dw + 1 surface sync = 26dw */ -static void -set_shaders(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u64 gpu_addr; - u32 sq_pgm_resources; - - /* setup shader regs */ - sq_pgm_resources = (1 << 0); - - /* VS */ - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_START_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_RESOURCES_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, sq_pgm_resources); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_CF_OFFSET_VS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); - - /* PS */ - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.ps_offset; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_START_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, gpu_addr >> 8); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_RESOURCES_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, sq_pgm_resources | (1 << 28)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_EXPORTS_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 2); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 1)); - radeon_ring_write(ring, (SQ_PGM_CF_OFFSET_PS - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, 0); - - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.vs_offset; - cp_set_surface_sync(rdev, PACKET3_SH_ACTION_ENA, 512, gpu_addr); -} - -/* emits 9 + 1 sync (5) = 14*/ -static void -set_vtx_resource(struct radeon_device *rdev, u64 gpu_addr) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 sq_vtx_constant_word2; - - sq_vtx_constant_word2 = SQ_VTXC_BASE_ADDR_HI(upper_32_bits(gpu_addr) & 0xff) | - SQ_VTXC_STRIDE(16); -#ifdef __BIG_ENDIAN - sq_vtx_constant_word2 |= SQ_VTXC_ENDIAN_SWAP(SQ_ENDIAN_8IN32); -#endif - - radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 7)); - radeon_ring_write(ring, 0x460); - radeon_ring_write(ring, gpu_addr & 0xffffffff); - radeon_ring_write(ring, 48 - 1); - radeon_ring_write(ring, sq_vtx_constant_word2); - radeon_ring_write(ring, 1 << 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, SQ_TEX_VTX_VALID_BUFFER << 30); - - if ((rdev->family == CHIP_RV610) || - (rdev->family == CHIP_RV620) || - (rdev->family == CHIP_RS780) || - (rdev->family == CHIP_RS880) || - (rdev->family == CHIP_RV710)) - cp_set_surface_sync(rdev, - PACKET3_TC_ACTION_ENA, 48, gpu_addr); - else - cp_set_surface_sync(rdev, - PACKET3_VC_ACTION_ENA, 48, gpu_addr); -} - -/* emits 9 */ -static void -set_tex_resource(struct radeon_device *rdev, - int format, int w, int h, int pitch, - u64 gpu_addr, u32 size) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - uint32_t sq_tex_resource_word0, sq_tex_resource_word1, sq_tex_resource_word4; - - if (h < 1) - h = 1; - - sq_tex_resource_word0 = S_038000_DIM(V_038000_SQ_TEX_DIM_2D) | - S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1); - sq_tex_resource_word0 |= S_038000_PITCH((pitch >> 3) - 1) | - S_038000_TEX_WIDTH(w - 1); - - sq_tex_resource_word1 = S_038004_DATA_FORMAT(format); - sq_tex_resource_word1 |= S_038004_TEX_HEIGHT(h - 1); - - sq_tex_resource_word4 = S_038010_REQUEST_SIZE(1) | - S_038010_DST_SEL_X(SQ_SEL_X) | - S_038010_DST_SEL_Y(SQ_SEL_Y) | - S_038010_DST_SEL_Z(SQ_SEL_Z) | - S_038010_DST_SEL_W(SQ_SEL_W); - - cp_set_surface_sync(rdev, - PACKET3_TC_ACTION_ENA, size, gpu_addr); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_RESOURCE, 7)); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, sq_tex_resource_word0); - radeon_ring_write(ring, sq_tex_resource_word1); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, gpu_addr >> 8); - radeon_ring_write(ring, sq_tex_resource_word4); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, SQ_TEX_VTX_VALID_TEXTURE << 30); -} - -/* emits 12 */ -static void -set_scissors(struct radeon_device *rdev, int x1, int y1, - int x2, int y2) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_SCREEN_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_GENERIC_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); - - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); - radeon_ring_write(ring, (PA_SC_WINDOW_SCISSOR_TL - PACKET3_SET_CONTEXT_REG_OFFSET) >> 2); - radeon_ring_write(ring, (x1 << 0) | (y1 << 16) | (1 << 31)); - radeon_ring_write(ring, (x2 << 0) | (y2 << 16)); -} - -/* emits 10 */ -static void -draw_auto(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); - radeon_ring_write(ring, (VGT_PRIMITIVE_TYPE - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); - radeon_ring_write(ring, DI_PT_RECTLIST); - - radeon_ring_write(ring, PACKET3(PACKET3_INDEX_TYPE, 0)); - radeon_ring_write(ring, -#ifdef __BIG_ENDIAN - (2 << 2) | -#endif - DI_INDEX_SIZE_16_BIT); - - radeon_ring_write(ring, PACKET3(PACKET3_NUM_INSTANCES, 0)); - radeon_ring_write(ring, 1); - - radeon_ring_write(ring, PACKET3(PACKET3_DRAW_INDEX_AUTO, 1)); - radeon_ring_write(ring, 3); - radeon_ring_write(ring, DI_SRC_SEL_AUTO_INDEX); - -} - -/* emits 14 */ -static void -set_default_state(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - u32 sq_config, sq_gpr_resource_mgmt_1, sq_gpr_resource_mgmt_2; - u32 sq_thread_resource_mgmt, sq_stack_resource_mgmt_1, sq_stack_resource_mgmt_2; - int num_ps_gprs, num_vs_gprs, num_temp_gprs, num_gs_gprs, num_es_gprs; - int num_ps_threads, num_vs_threads, num_gs_threads, num_es_threads; - int num_ps_stack_entries, num_vs_stack_entries, num_gs_stack_entries, num_es_stack_entries; - u64 gpu_addr; - int dwords; - - switch (rdev->family) { - case CHIP_R600: - num_ps_gprs = 192; - num_vs_gprs = 56; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 136; - num_vs_threads = 48; - num_gs_threads = 4; - num_es_threads = 4; - num_ps_stack_entries = 128; - num_vs_stack_entries = 128; - num_gs_stack_entries = 0; - num_es_stack_entries = 0; - break; - case CHIP_RV630: - case CHIP_RV635: - num_ps_gprs = 84; - num_vs_gprs = 36; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 144; - num_vs_threads = 40; - num_gs_threads = 4; - num_es_threads = 4; - num_ps_stack_entries = 40; - num_vs_stack_entries = 40; - num_gs_stack_entries = 32; - num_es_stack_entries = 16; - break; - case CHIP_RV610: - case CHIP_RV620: - case CHIP_RS780: - case CHIP_RS880: - default: - num_ps_gprs = 84; - num_vs_gprs = 36; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 136; - num_vs_threads = 48; - num_gs_threads = 4; - num_es_threads = 4; - num_ps_stack_entries = 40; - num_vs_stack_entries = 40; - num_gs_stack_entries = 32; - num_es_stack_entries = 16; - break; - case CHIP_RV670: - num_ps_gprs = 144; - num_vs_gprs = 40; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 136; - num_vs_threads = 48; - num_gs_threads = 4; - num_es_threads = 4; - num_ps_stack_entries = 40; - num_vs_stack_entries = 40; - num_gs_stack_entries = 32; - num_es_stack_entries = 16; - break; - case CHIP_RV770: - num_ps_gprs = 192; - num_vs_gprs = 56; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 188; - num_vs_threads = 60; - num_gs_threads = 0; - num_es_threads = 0; - num_ps_stack_entries = 256; - num_vs_stack_entries = 256; - num_gs_stack_entries = 0; - num_es_stack_entries = 0; - break; - case CHIP_RV730: - case CHIP_RV740: - num_ps_gprs = 84; - num_vs_gprs = 36; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 188; - num_vs_threads = 60; - num_gs_threads = 0; - num_es_threads = 0; - num_ps_stack_entries = 128; - num_vs_stack_entries = 128; - num_gs_stack_entries = 0; - num_es_stack_entries = 0; - break; - case CHIP_RV710: - num_ps_gprs = 192; - num_vs_gprs = 56; - num_temp_gprs = 4; - num_gs_gprs = 0; - num_es_gprs = 0; - num_ps_threads = 144; - num_vs_threads = 48; - num_gs_threads = 0; - num_es_threads = 0; - num_ps_stack_entries = 128; - num_vs_stack_entries = 128; - num_gs_stack_entries = 0; - num_es_stack_entries = 0; - break; - } - - if ((rdev->family == CHIP_RV610) || - (rdev->family == CHIP_RV620) || - (rdev->family == CHIP_RS780) || - (rdev->family == CHIP_RS880) || - (rdev->family == CHIP_RV710)) - sq_config = 0; - else - sq_config = VC_ENABLE; - - sq_config |= (DX9_CONSTS | - ALU_INST_PREFER_VECTOR | - PS_PRIO(0) | - VS_PRIO(1) | - GS_PRIO(2) | - ES_PRIO(3)); - - sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(num_ps_gprs) | - NUM_VS_GPRS(num_vs_gprs) | - NUM_CLAUSE_TEMP_GPRS(num_temp_gprs)); - sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(num_gs_gprs) | - NUM_ES_GPRS(num_es_gprs)); - sq_thread_resource_mgmt = (NUM_PS_THREADS(num_ps_threads) | - NUM_VS_THREADS(num_vs_threads) | - NUM_GS_THREADS(num_gs_threads) | - NUM_ES_THREADS(num_es_threads)); - sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(num_ps_stack_entries) | - NUM_VS_STACK_ENTRIES(num_vs_stack_entries)); - sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(num_gs_stack_entries) | - NUM_ES_STACK_ENTRIES(num_es_stack_entries)); - - /* emit an IB pointing at default state */ - dwords = ALIGN(rdev->r600_blit.state_len, 0x10); - gpu_addr = rdev->r600_blit.shader_gpu_addr + rdev->r600_blit.state_offset; - radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); - radeon_ring_write(ring, -#ifdef __BIG_ENDIAN - (2 << 0) | -#endif - (gpu_addr & 0xFFFFFFFC)); - radeon_ring_write(ring, upper_32_bits(gpu_addr) & 0xFF); - radeon_ring_write(ring, dwords); - - /* SQ config */ - radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 6)); - radeon_ring_write(ring, (SQ_CONFIG - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); - radeon_ring_write(ring, sq_config); - radeon_ring_write(ring, sq_gpr_resource_mgmt_1); - radeon_ring_write(ring, sq_gpr_resource_mgmt_2); - radeon_ring_write(ring, sq_thread_resource_mgmt); - radeon_ring_write(ring, sq_stack_resource_mgmt_1); - radeon_ring_write(ring, sq_stack_resource_mgmt_2); -} - -int r600_blit_init(struct radeon_device *rdev) -{ - u32 obj_size; - int i, r, dwords; - void *ptr; - u32 packet2s[16]; - int num_packet2s = 0; - - rdev->r600_blit.primitives.set_render_target = set_render_target; - rdev->r600_blit.primitives.cp_set_surface_sync = cp_set_surface_sync; - rdev->r600_blit.primitives.set_shaders = set_shaders; - rdev->r600_blit.primitives.set_vtx_resource = set_vtx_resource; - rdev->r600_blit.primitives.set_tex_resource = set_tex_resource; - rdev->r600_blit.primitives.set_scissors = set_scissors; - rdev->r600_blit.primitives.draw_auto = draw_auto; - rdev->r600_blit.primitives.set_default_state = set_default_state; - - rdev->r600_blit.ring_size_common = 8; /* sync semaphore */ - rdev->r600_blit.ring_size_common += 40; /* shaders + def state */ - rdev->r600_blit.ring_size_common += 5; /* done copy */ - rdev->r600_blit.ring_size_common += 16; /* fence emit for done copy */ - - rdev->r600_blit.ring_size_per_loop = 76; - /* set_render_target emits 2 extra dwords on rv6xx */ - if (rdev->family > CHIP_R600 && rdev->family < CHIP_RV770) - rdev->r600_blit.ring_size_per_loop += 2; - - rdev->r600_blit.max_dim = 8192; - - rdev->r600_blit.state_offset = 0; - - if (rdev->family >= CHIP_RV770) - rdev->r600_blit.state_len = r7xx_default_size; - else - rdev->r600_blit.state_len = r6xx_default_size; - - dwords = rdev->r600_blit.state_len; - while (dwords & 0xf) { - packet2s[num_packet2s++] = cpu_to_le32(PACKET2(0)); - dwords++; - } - - obj_size = dwords * 4; - obj_size = ALIGN(obj_size, 256); - - rdev->r600_blit.vs_offset = obj_size; - obj_size += r6xx_vs_size * 4; - obj_size = ALIGN(obj_size, 256); - - rdev->r600_blit.ps_offset = obj_size; - obj_size += r6xx_ps_size * 4; - obj_size = ALIGN(obj_size, 256); - - /* pin copy shader into vram if not already initialized */ - if (rdev->r600_blit.shader_obj == NULL) { - r = radeon_bo_create(rdev, obj_size, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, - NULL, &rdev->r600_blit.shader_obj); - if (r) { - DRM_ERROR("r600 failed to allocate shader\n"); - 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); - if (r) { - dev_err(rdev->dev, "(%d) pin blit object failed\n", r); - return r; - } - } - - DRM_DEBUG("r6xx blit allocated bo %08x vs %08x ps %08x\n", - obj_size, - rdev->r600_blit.vs_offset, rdev->r600_blit.ps_offset); - - r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); - if (unlikely(r != 0)) - return r; - r = radeon_bo_kmap(rdev->r600_blit.shader_obj, &ptr); - if (r) { - DRM_ERROR("failed to map blit object %d\n", r); - return r; - } - if (rdev->family >= CHIP_RV770) - memcpy_toio(ptr + rdev->r600_blit.state_offset, - r7xx_default_state, rdev->r600_blit.state_len * 4); - else - memcpy_toio(ptr + rdev->r600_blit.state_offset, - r6xx_default_state, rdev->r600_blit.state_len * 4); - if (num_packet2s) - memcpy_toio(ptr + rdev->r600_blit.state_offset + (rdev->r600_blit.state_len * 4), - packet2s, num_packet2s * 4); - for (i = 0; i < r6xx_vs_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.vs_offset + i * 4) = cpu_to_le32(r6xx_vs[i]); - for (i = 0; i < r6xx_ps_size; i++) - *(u32 *)((unsigned long)ptr + rdev->r600_blit.ps_offset + i * 4) = cpu_to_le32(r6xx_ps[i]); - radeon_bo_kunmap(rdev->r600_blit.shader_obj); - radeon_bo_unreserve(rdev->r600_blit.shader_obj); - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); - return 0; -} - -void r600_blit_fini(struct radeon_device *rdev) -{ - int r; - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); - if (rdev->r600_blit.shader_obj == NULL) - return; - /* If we can't reserve the bo, unref should be enough to destroy - * it when it becomes idle. - */ - r = radeon_bo_reserve(rdev->r600_blit.shader_obj, false); - if (!r) { - radeon_bo_unpin(rdev->r600_blit.shader_obj); - radeon_bo_unreserve(rdev->r600_blit.shader_obj); - } - radeon_bo_unref(&rdev->r600_blit.shader_obj); -} - -static unsigned r600_blit_create_rect(unsigned num_gpu_pages, - int *width, int *height, int max_dim) -{ - unsigned max_pages; - unsigned pages = num_gpu_pages; - int w, h; - - if (num_gpu_pages == 0) { - /* not supposed to be called with no pages, but just in case */ - h = 0; - w = 0; - pages = 0; - WARN_ON(1); - } else { - int rect_order = 2; - h = RECT_UNIT_H; - while (num_gpu_pages / rect_order) { - h *= 2; - rect_order *= 4; - if (h >= max_dim) { - h = max_dim; - break; - } - } - max_pages = (max_dim * h) / (RECT_UNIT_W * RECT_UNIT_H); - if (pages > max_pages) - pages = max_pages; - w = (pages * RECT_UNIT_W * RECT_UNIT_H) / h; - w = (w / RECT_UNIT_W) * RECT_UNIT_W; - pages = (w * h) / (RECT_UNIT_W * RECT_UNIT_H); - BUG_ON(pages == 0); - } - - - DRM_DEBUG("blit_rectangle: h=%d, w=%d, pages=%d\n", h, w, pages); - - /* return width and height only of the caller wants it */ - if (height) - *height = h; - if (width) - *width = w; - - return pages; -} - - -int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_pages, - struct radeon_fence **fence, struct radeon_sa_bo **vb, - struct radeon_semaphore **sem) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - int r; - int ring_size; - int num_loops = 0; - int dwords_per_loop = rdev->r600_blit.ring_size_per_loop; - - /* num loops */ - while (num_gpu_pages) { - num_gpu_pages -= - r600_blit_create_rect(num_gpu_pages, NULL, NULL, - rdev->r600_blit.max_dim); - num_loops++; - } - - /* 48 bytes for vertex per loop */ - r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, vb, - (num_loops*48)+256, 256, true); - if (r) { - return r; - } - - r = radeon_semaphore_create(rdev, sem); - if (r) { - radeon_sa_bo_free(rdev, vb, NULL); - return r; - } - - /* calculate number of loops correctly */ - ring_size = num_loops * dwords_per_loop; - ring_size += rdev->r600_blit.ring_size_common; - r = radeon_ring_lock(rdev, ring, ring_size); - if (r) { - radeon_sa_bo_free(rdev, vb, NULL); - radeon_semaphore_free(rdev, sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, RADEON_RING_TYPE_GFX_INDEX)) { - radeon_semaphore_sync_rings(rdev, *sem, (*fence)->ring, - RADEON_RING_TYPE_GFX_INDEX); - radeon_fence_note_sync(*fence, RADEON_RING_TYPE_GFX_INDEX); - } else { - radeon_semaphore_free(rdev, sem, NULL); - } - - rdev->r600_blit.primitives.set_default_state(rdev); - rdev->r600_blit.primitives.set_shaders(rdev); - return 0; -} - -void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence **fence, - struct radeon_sa_bo *vb, struct radeon_semaphore *sem) -{ - struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; - int r; - - r = radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_sa_bo_free(rdev, &vb, *fence); - radeon_semaphore_free(rdev, &sem, *fence); -} - -void r600_kms_blit_copy(struct radeon_device *rdev, - u64 src_gpu_addr, u64 dst_gpu_addr, - unsigned num_gpu_pages, - struct radeon_sa_bo *vb) -{ - u64 vb_gpu_addr; - u32 *vb_cpu_addr; - - DRM_DEBUG("emitting copy %16llx %16llx %d\n", - src_gpu_addr, dst_gpu_addr, num_gpu_pages); - vb_cpu_addr = (u32 *)radeon_sa_bo_cpu_addr(vb); - vb_gpu_addr = radeon_sa_bo_gpu_addr(vb); - - while (num_gpu_pages) { - int w, h; - unsigned size_in_bytes; - unsigned pages_per_loop = - r600_blit_create_rect(num_gpu_pages, &w, &h, - rdev->r600_blit.max_dim); - - size_in_bytes = pages_per_loop * RADEON_GPU_PAGE_SIZE; - DRM_DEBUG("rectangle w=%d h=%d\n", w, h); - - vb_cpu_addr[0] = 0; - vb_cpu_addr[1] = 0; - vb_cpu_addr[2] = 0; - vb_cpu_addr[3] = 0; - - vb_cpu_addr[4] = 0; - vb_cpu_addr[5] = int2float(h); - vb_cpu_addr[6] = 0; - vb_cpu_addr[7] = int2float(h); - - vb_cpu_addr[8] = int2float(w); - vb_cpu_addr[9] = int2float(h); - vb_cpu_addr[10] = int2float(w); - vb_cpu_addr[11] = int2float(h); - - rdev->r600_blit.primitives.set_tex_resource(rdev, FMT_8_8_8_8, - w, h, w, src_gpu_addr, size_in_bytes); - rdev->r600_blit.primitives.set_render_target(rdev, COLOR_8_8_8_8, - w, h, dst_gpu_addr); - rdev->r600_blit.primitives.set_scissors(rdev, 0, 0, w, h); - rdev->r600_blit.primitives.set_vtx_resource(rdev, vb_gpu_addr); - rdev->r600_blit.primitives.draw_auto(rdev); - rdev->r600_blit.primitives.cp_set_surface_sync(rdev, - PACKET3_CB_ACTION_ENA | PACKET3_CB0_DEST_BASE_ENA, - size_in_bytes, dst_gpu_addr); - - vb_cpu_addr += 12; - vb_gpu_addr += 4*12; - src_gpu_addr += size_in_bytes; - dst_gpu_addr += size_in_bytes; - num_gpu_pages -= pages_per_loop; - } -} diff --git a/drivers/gpu/drm/radeon/r600_blit_shaders.h b/drivers/gpu/drm/radeon/r600_blit_shaders.h index 2f3ce7a75976..f437d36dd98c 100644 --- a/drivers/gpu/drm/radeon/r600_blit_shaders.h +++ b/drivers/gpu/drm/radeon/r600_blit_shaders.h @@ -35,5 +35,4 @@ extern const u32 r6xx_default_state[]; extern const u32 r6xx_ps_size, r6xx_vs_size; extern const u32 r6xx_default_size, r7xx_default_size; -__pure uint32_t int2float(uint32_t x); #endif diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 9f19259667df..83be8fdceab1 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -844,35 +844,6 @@ struct r600_ih { bool enabled; }; -struct r600_blit_cp_primitives { - void (*set_render_target)(struct radeon_device *rdev, int format, - int w, int h, u64 gpu_addr); - void (*cp_set_surface_sync)(struct radeon_device *rdev, - u32 sync_type, u32 size, - u64 mc_addr); - void (*set_shaders)(struct radeon_device *rdev); - void (*set_vtx_resource)(struct radeon_device *rdev, u64 gpu_addr); - void (*set_tex_resource)(struct radeon_device *rdev, - int format, int w, int h, int pitch, - u64 gpu_addr, u32 size); - void (*set_scissors)(struct radeon_device *rdev, int x1, int y1, - int x2, int y2); - void (*draw_auto)(struct radeon_device *rdev); - void (*set_default_state)(struct radeon_device *rdev); -}; - -struct r600_blit { - struct radeon_bo *shader_obj; - struct r600_blit_cp_primitives primitives; - int max_dim; - int ring_size_common; - int ring_size_per_loop; - u64 shader_gpu_addr; - u32 vs_offset, ps_offset; - u32 state_offset; - u32 state_len; -}; - /* * RLC stuff */ @@ -2066,7 +2037,6 @@ struct radeon_device { const struct firmware *sdma_fw; /* CIK SDMA firmware */ const struct firmware *smc_fw; /* SMC firmware */ const struct firmware *uvd_fw; /* UVD firmware */ - struct r600_blit r600_blit; struct r600_vram_scratch vram_scratch; int msi_enabled; /* msi enabled */ struct r600_ih ih; /* r6/700 interrupt ring */ diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 3d61d5aac18f..3cf7d89c1bd8 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -337,9 +337,6 @@ void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); int r600_dma_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); int r600_uvd_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); -int r600_copy_blit(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, struct radeon_fence **fence); int r600_copy_cpdma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, struct radeon_fence **fence); @@ -371,8 +368,6 @@ int r600_count_pipe_bits(uint32_t val); int r600_mc_wait_for_idle(struct radeon_device *rdev); int r600_pcie_gart_init(struct radeon_device *rdev); void r600_scratch_init(struct radeon_device *rdev); -int r600_blit_init(struct radeon_device *rdev); -void r600_blit_fini(struct radeon_device *rdev); int r600_init_microcode(struct radeon_device *rdev); /* r600 irq */ int r600_irq_process(struct radeon_device *rdev); @@ -391,16 +386,6 @@ int r600_hdmi_buffer_status_changed(struct drm_encoder *encoder); void r600_hdmi_update_audio_settings(struct drm_encoder *encoder); void r600_hdmi_enable(struct drm_encoder *encoder, bool enable); void r600_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mode); -/* r600 blit */ -int r600_blit_prepare_copy(struct radeon_device *rdev, unsigned num_gpu_pages, - struct radeon_fence **fence, struct radeon_sa_bo **vb, - struct radeon_semaphore **sem); -void r600_blit_done_copy(struct radeon_device *rdev, struct radeon_fence **fence, - struct radeon_sa_bo *vb, struct radeon_semaphore *sem); -void r600_kms_blit_copy(struct radeon_device *rdev, - u64 src_gpu_addr, u64 dst_gpu_addr, - unsigned num_gpu_pages, - struct radeon_sa_bo *vb); int r600_mc_wait_for_idle(struct radeon_device *rdev); u32 r600_get_xclk(struct radeon_device *rdev); uint64_t r600_get_gpu_clock_counter(struct radeon_device *rdev); @@ -530,7 +515,6 @@ extern u32 evergreen_page_flip(struct radeon_device *rdev, int crtc, u64 crtc_ba extern void evergreen_post_page_flip(struct radeon_device *rdev, int crtc); extern void dce4_wait_for_vblank(struct radeon_device *rdev, int crtc); void evergreen_disable_interrupt_state(struct radeon_device *rdev); -int evergreen_blit_init(struct radeon_device *rdev); int evergreen_mc_wait_for_idle(struct radeon_device *rdev); void evergreen_dma_fence_ring_emit(struct radeon_device *rdev, struct radeon_fence *fence); diff --git a/drivers/gpu/drm/radeon/radeon_blit_common.h b/drivers/gpu/drm/radeon/radeon_blit_common.h deleted file mode 100644 index 4ecbe72c9d2d..000000000000 --- a/drivers/gpu/drm/radeon/radeon_blit_common.h +++ /dev/null @@ -1,44 +0,0 @@ -/* - * Copyright 2009 Advanced Micro Devices, Inc. - * Copyright 2009 Red Hat Inc. - * Copyright 2012 Alcatel-Lucent, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - */ - -#ifndef __RADEON_BLIT_COMMON_H__ - -#define DI_PT_RECTLIST 0x11 -#define DI_INDEX_SIZE_16_BIT 0x0 -#define DI_SRC_SEL_AUTO_INDEX 0x2 - -#define FMT_8 0x1 -#define FMT_5_6_5 0x8 -#define FMT_8_8_8_8 0x1a -#define COLOR_8 0x1 -#define COLOR_5_6_5 0x8 -#define COLOR_8_8_8_8 0x1a - -#define RECT_UNIT_H 32 -#define RECT_UNIT_W (RADEON_GPU_PAGE_SIZE / 4 / RECT_UNIT_H) - -#define __RADEON_BLIT_COMMON_H__ -#endif diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index f5e92cfcc140..95590bd07afb 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1852,12 +1852,6 @@ static int rv770_startup(struct radeon_device *rdev) } rv770_gpu_init(rdev); - r = r600_blit_init(rdev); - if (r) { - r600_blit_fini(rdev); - rdev->asic->copy.copy = NULL; - dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); - } /* allocate wb buffer */ r = radeon_wb_init(rdev); @@ -2092,7 +2086,6 @@ int rv770_init(struct radeon_device *rdev) void rv770_fini(struct radeon_device *rdev) { - r600_blit_fini(rdev); r700_cp_fini(rdev); r600_dma_fini(rdev); r600_irq_fini(rdev); -- cgit v1.2.3 From 10b7ca7e09aa0d5f90265f130aee2b9270bfaadc Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 17 Apr 2013 17:22:05 -0400 Subject: drm/radeon: clean up sumo_rlc_init() for code sharing This will eventually be shared with newer asics to reduce code duplication. Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/evergreen.c | 221 +++++++++++++++++++------------------ 1 file changed, 113 insertions(+), 108 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index a5ab5693eb2a..617840829114 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -3910,131 +3910,136 @@ int sumo_rlc_init(struct radeon_device *rdev) dws = rdev->rlc.reg_list_size; cs_data = rdev->rlc.cs_data; - /* save restore block */ - if (rdev->rlc.save_restore_obj == NULL) { - r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.save_restore_obj); + if (src_ptr) { + /* save restore block */ + if (rdev->rlc.save_restore_obj == NULL) { + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.save_restore_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r); + return r; + } + } + + r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); + if (unlikely(r != 0)) { + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->rlc.save_restore_gpu_addr); if (r) { - dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r); + radeon_bo_unreserve(rdev->rlc.save_restore_obj); + dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r); + sumo_rlc_fini(rdev); return r; } - } - r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); - if (unlikely(r != 0)) { - sumo_rlc_fini(rdev); - return r; - } - r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->rlc.save_restore_gpu_addr); - if (r) { + r = radeon_bo_kmap(rdev->rlc.save_restore_obj, (void **)&rdev->rlc.sr_ptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC sr bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + /* write the sr buffer */ + dst_ptr = rdev->rlc.sr_ptr; + /* format: + * dw0: (reg2 << 16) | reg1 + * dw1: reg1 save space + * dw2: reg2 save space + */ + for (i = 0; i < dws; i++) { + data = src_ptr[i] >> 2; + i++; + if (i < dws) + data |= (src_ptr[i] >> 2) << 16; + j = (((i - 1) * 3) / 2); + dst_ptr[j] = data; + } + j = ((i * 3) / 2); + dst_ptr[j] = RLC_SAVE_RESTORE_LIST_END_MARKER; + + radeon_bo_kunmap(rdev->rlc.save_restore_obj); radeon_bo_unreserve(rdev->rlc.save_restore_obj); - dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r); - sumo_rlc_fini(rdev); - return r; - } - r = radeon_bo_kmap(rdev->rlc.save_restore_obj, (void **)&rdev->rlc.sr_ptr); - if (r) { - dev_warn(rdev->dev, "(%d) map RLC sr bo failed\n", r); - sumo_rlc_fini(rdev); - return r; - } - /* write the sr buffer */ - dst_ptr = rdev->rlc.sr_ptr; - /* format: - * dw0: (reg2 << 16) | reg1 - * dw1: reg1 save space - * dw2: reg2 save space - */ - for (i = 0; i < dws; i++) { - data = src_ptr[i] >> 2; - i++; - if (i < dws) - data |= (src_ptr[i] >> 2) << 16; - j = (((i - 1) * 3) / 2); - dst_ptr[j] = data; } - j = ((i * 3) / 2); - dst_ptr[j] = RLC_SAVE_RESTORE_LIST_END_MARKER; - - radeon_bo_kunmap(rdev->rlc.save_restore_obj); - radeon_bo_unreserve(rdev->rlc.save_restore_obj); - /* clear state block */ - reg_list_num = 0; - dws = 0; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_list_num++; - dws += cs_data[i].section[j].reg_count; + if (cs_data) { + /* clear state block */ + reg_list_num = 0; + dws = 0; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_list_num++; + dws += cs_data[i].section[j].reg_count; + } } - } - reg_list_blk_index = (3 * reg_list_num + 2); - dws += reg_list_blk_index; + reg_list_blk_index = (3 * reg_list_num + 2); + dws += reg_list_blk_index; - if (rdev->rlc.clear_state_obj == NULL) { - r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); + if (rdev->rlc.clear_state_obj == NULL) { + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + } + r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); + if (unlikely(r != 0)) { + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->rlc.clear_state_gpu_addr); if (r) { - dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); + radeon_bo_unreserve(rdev->rlc.clear_state_obj); + dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r); sumo_rlc_fini(rdev); return r; } - } - r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - sumo_rlc_fini(rdev); - return r; - } - r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->rlc.clear_state_gpu_addr); - if (r) { - radeon_bo_unreserve(rdev->rlc.clear_state_obj); - dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r); - sumo_rlc_fini(rdev); - return r; - } - r = radeon_bo_kmap(rdev->rlc.clear_state_obj, (void **)&rdev->rlc.cs_ptr); - if (r) { - dev_warn(rdev->dev, "(%d) map RLC c bo failed\n", r); - sumo_rlc_fini(rdev); - return r; - } - /* set up the cs buffer */ - dst_ptr = rdev->rlc.cs_ptr; - reg_list_hdr_blk_index = 0; - reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); - data = upper_32_bits(reg_list_mc_addr); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_num = cs_data[i].section[j].reg_count; - data = reg_list_mc_addr & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = 0x08000000 | (reg_num * 4); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - for (k = 0; k < reg_num; k++) { - data = cs_data[i].section[j].extent[k]; - dst_ptr[reg_list_blk_index + k] = data; + r = radeon_bo_kmap(rdev->rlc.clear_state_obj, (void **)&rdev->rlc.cs_ptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC c bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + /* set up the cs buffer */ + dst_ptr = rdev->rlc.cs_ptr; + reg_list_hdr_blk_index = 0; + reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); + data = upper_32_bits(reg_list_mc_addr); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_num = cs_data[i].section[j].reg_count; + data = reg_list_mc_addr & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = 0x08000000 | (reg_num * 4); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + for (k = 0; k < reg_num; k++) { + data = cs_data[i].section[j].extent[k]; + dst_ptr[reg_list_blk_index + k] = data; + } + reg_list_mc_addr += reg_num * 4; + reg_list_blk_index += reg_num; } - reg_list_mc_addr += reg_num * 4; - reg_list_blk_index += reg_num; } - } - dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; + dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; - radeon_bo_kunmap(rdev->rlc.clear_state_obj); - radeon_bo_unreserve(rdev->rlc.clear_state_obj); + radeon_bo_kunmap(rdev->rlc.clear_state_obj); + radeon_bo_unreserve(rdev->rlc.clear_state_obj); + } return 0; } -- cgit v1.2.3 From 1fd11777c2f0e6b6b37432b984bf40e3c6072f23 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 17 Apr 2013 17:53:50 -0400 Subject: drm/radeon: convert SI,CIK to use sumo_rlc functions and remove duplicate si_rlc functions. Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 10 +- drivers/gpu/drm/radeon/clearstate_cayman.h | 2 +- drivers/gpu/drm/radeon/clearstate_evergreen.h | 2 +- drivers/gpu/drm/radeon/evergreen.c | 46 ++++--- drivers/gpu/drm/radeon/ni.c | 6 +- drivers/gpu/drm/radeon/radeon.h | 4 +- drivers/gpu/drm/radeon/si.c | 174 ++------------------------ 7 files changed, 49 insertions(+), 195 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index e92a9721ca25..8389917af9a2 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -57,9 +57,9 @@ extern void r600_ih_ring_fini(struct radeon_device *rdev); extern void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save); extern void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save); extern bool evergreen_is_display_hung(struct radeon_device *rdev); +extern void sumo_rlc_fini(struct radeon_device *rdev); +extern int sumo_rlc_init(struct radeon_device *rdev); extern void si_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc); -extern void si_rlc_fini(struct radeon_device *rdev); -extern int si_rlc_init(struct radeon_device *rdev); extern void si_rlc_reset(struct radeon_device *rdev); static void cik_rlc_stop(struct radeon_device *rdev); static void cik_pcie_gen3_enable(struct radeon_device *rdev); @@ -6019,7 +6019,7 @@ static int cik_startup(struct radeon_device *rdev) cik_gpu_init(rdev); /* allocate rlc buffers */ - r = si_rlc_init(rdev); + r = sumo_rlc_init(rdev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -6343,7 +6343,7 @@ int cik_init(struct radeon_device *rdev) cik_cp_fini(rdev); cik_sdma_fini(rdev); cik_irq_fini(rdev); - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); cik_mec_fini(rdev); radeon_wb_fini(rdev); radeon_ib_pool_fini(rdev); @@ -6379,7 +6379,7 @@ void cik_fini(struct radeon_device *rdev) cik_cp_fini(rdev); cik_sdma_fini(rdev); cik_irq_fini(rdev); - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); cik_mec_fini(rdev); radeon_wb_fini(rdev); radeon_vm_manager_fini(rdev); diff --git a/drivers/gpu/drm/radeon/clearstate_cayman.h b/drivers/gpu/drm/radeon/clearstate_cayman.h index c00339440c5e..aa908c55a513 100644 --- a/drivers/gpu/drm/radeon/clearstate_cayman.h +++ b/drivers/gpu/drm/radeon/clearstate_cayman.h @@ -1073,7 +1073,7 @@ static const struct cs_extent_def SECT_CTRLCONST_defs[] = {SECT_CTRLCONST_def_1, 0x0000f3fc, 2 }, { 0, 0, 0 } }; -struct cs_section_def cayman_cs_data[] = { +static const struct cs_section_def cayman_cs_data[] = { { SECT_CONTEXT_defs, SECT_CONTEXT }, { SECT_CLEAR_defs, SECT_CLEAR }, { SECT_CTRLCONST_defs, SECT_CTRLCONST }, diff --git a/drivers/gpu/drm/radeon/clearstate_evergreen.h b/drivers/gpu/drm/radeon/clearstate_evergreen.h index 4791d856b7fd..63a1ffbb3ced 100644 --- a/drivers/gpu/drm/radeon/clearstate_evergreen.h +++ b/drivers/gpu/drm/radeon/clearstate_evergreen.h @@ -1072,7 +1072,7 @@ static const struct cs_extent_def SECT_CTRLCONST_defs[] = {SECT_CTRLCONST_def_1, 0x0000f3fc, 2 }, { 0, 0, 0 } }; -struct cs_section_def evergreen_cs_data[] = { +static const struct cs_section_def evergreen_cs_data[] = { { SECT_CONTEXT_defs, SECT_CONTEXT }, { SECT_CLEAR_defs, SECT_CLEAR }, { SECT_CTRLCONST_defs, SECT_CTRLCONST }, diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 617840829114..6fc876a444d4 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -47,7 +47,7 @@ static const u32 crtc_offsets[6] = #include "clearstate_evergreen.h" -static u32 sumo_rlc_save_restore_register_list[] = +static const u32 sumo_rlc_save_restore_register_list[] = { 0x98fc, 0x9830, @@ -131,7 +131,6 @@ static u32 sumo_rlc_save_restore_register_list[] = 0x9150, 0x802c, }; -static u32 sumo_rlc_save_restore_register_list_size = ARRAY_SIZE(sumo_rlc_save_restore_register_list); static void evergreen_gpu_init(struct radeon_device *rdev); void evergreen_fini(struct radeon_device *rdev); @@ -3898,12 +3897,12 @@ void sumo_rlc_fini(struct radeon_device *rdev) int sumo_rlc_init(struct radeon_device *rdev) { - u32 *src_ptr; + const u32 *src_ptr; volatile u32 *dst_ptr; u32 dws, data, i, j, k, reg_num; u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index; u64 reg_list_mc_addr; - struct cs_section_def *cs_data; + const struct cs_section_def *cs_data; int r; src_ptr = rdev->rlc.reg_list; @@ -3943,22 +3942,28 @@ int sumo_rlc_init(struct radeon_device *rdev) } /* write the sr buffer */ dst_ptr = rdev->rlc.sr_ptr; - /* format: - * dw0: (reg2 << 16) | reg1 - * dw1: reg1 save space - * dw2: reg2 save space - */ - for (i = 0; i < dws; i++) { - data = src_ptr[i] >> 2; - i++; - if (i < dws) - data |= (src_ptr[i] >> 2) << 16; - j = (((i - 1) * 3) / 2); - dst_ptr[j] = data; + if (rdev->family >= CHIP_TAHITI) { + /* SI */ + for (i = 0; i < dws; i++) + dst_ptr[i] = src_ptr[i]; + } else { + /* ON/LN/TN */ + /* format: + * dw0: (reg2 << 16) | reg1 + * dw1: reg1 save space + * dw2: reg2 save space + */ + for (i = 0; i < dws; i++) { + data = src_ptr[i] >> 2; + i++; + if (i < dws) + data |= (src_ptr[i] >> 2) << 16; + j = (((i - 1) * 3) / 2); + dst_ptr[j] = data; + } + j = ((i * 3) / 2); + dst_ptr[j] = RLC_SAVE_RESTORE_LIST_END_MARKER; } - j = ((i * 3) / 2); - dst_ptr[j] = RLC_SAVE_RESTORE_LIST_END_MARKER; - radeon_bo_kunmap(rdev->rlc.save_restore_obj); radeon_bo_unreserve(rdev->rlc.save_restore_obj); } @@ -5152,7 +5157,8 @@ static int evergreen_startup(struct radeon_device *rdev) /* allocate rlc buffers */ if (rdev->flags & RADEON_IS_IGP) { rdev->rlc.reg_list = sumo_rlc_save_restore_register_list; - rdev->rlc.reg_list_size = sumo_rlc_save_restore_register_list_size; + rdev->rlc.reg_list_size = + (u32)ARRAY_SIZE(sumo_rlc_save_restore_register_list); rdev->rlc.cs_data = evergreen_cs_data; r = sumo_rlc_init(rdev); if (r) { diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index acdd6039ef14..21f2eceff2c6 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -35,7 +35,7 @@ #include "radeon_ucode.h" #include "clearstate_cayman.h" -static u32 tn_rlc_save_restore_register_list[] = +static const u32 tn_rlc_save_restore_register_list[] = { 0x98fc, 0x98f0, @@ -160,7 +160,6 @@ static u32 tn_rlc_save_restore_register_list[] = 0x9830, 0x802c, }; -static u32 tn_rlc_save_restore_register_list_size = ARRAY_SIZE(tn_rlc_save_restore_register_list); extern bool evergreen_is_display_hung(struct radeon_device *rdev); extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev); @@ -2121,7 +2120,8 @@ static int cayman_startup(struct radeon_device *rdev) /* allocate rlc buffers */ if (rdev->flags & RADEON_IS_IGP) { rdev->rlc.reg_list = tn_rlc_save_restore_register_list; - rdev->rlc.reg_list_size = tn_rlc_save_restore_register_list_size; + rdev->rlc.reg_list_size = + (u32)ARRAY_SIZE(tn_rlc_save_restore_register_list); rdev->rlc.cs_data = cayman_cs_data; r = sumo_rlc_init(rdev); if (r) { diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 034782804846..b9706e83e827 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -854,13 +854,13 @@ struct radeon_rlc { struct radeon_bo *save_restore_obj; uint64_t save_restore_gpu_addr; volatile uint32_t *sr_ptr; - u32 *reg_list; + const u32 *reg_list; u32 reg_list_size; /* for clear state */ struct radeon_bo *clear_state_obj; uint64_t clear_state_gpu_addr; volatile uint32_t *cs_ptr; - struct cs_section_def *cs_data; + const struct cs_section_def *cs_data; }; int radeon_ib_get(struct radeon_device *rdev, int ring, diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 610adfc86bea..8b8963d4a732 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -68,6 +68,8 @@ MODULE_FIRMWARE("radeon/HAINAN_smc.bin"); static void si_pcie_gen3_enable(struct radeon_device *rdev); static void si_program_aspm(struct radeon_device *rdev); +extern void sumo_rlc_fini(struct radeon_device *rdev); +extern int sumo_rlc_init(struct radeon_device *rdev); extern int r600_ih_ring_alloc(struct radeon_device *rdev); extern void r600_ih_ring_fini(struct radeon_device *rdev); extern void evergreen_fix_pci_max_read_req_size(struct radeon_device *rdev); @@ -5275,166 +5277,6 @@ static void si_fini_pg(struct radeon_device *rdev) /* * RLC */ -void si_rlc_fini(struct radeon_device *rdev) -{ - int r; - - /* save restore block */ - if (rdev->rlc.save_restore_obj) { - r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); - if (unlikely(r != 0)) - dev_warn(rdev->dev, "(%d) reserve RLC sr bo failed\n", r); - radeon_bo_unpin(rdev->rlc.save_restore_obj); - radeon_bo_unreserve(rdev->rlc.save_restore_obj); - - radeon_bo_unref(&rdev->rlc.save_restore_obj); - rdev->rlc.save_restore_obj = NULL; - } - - /* clear state block */ - if (rdev->rlc.clear_state_obj) { - r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); - if (unlikely(r != 0)) - dev_warn(rdev->dev, "(%d) reserve RLC c bo failed\n", r); - radeon_bo_unpin(rdev->rlc.clear_state_obj); - radeon_bo_unreserve(rdev->rlc.clear_state_obj); - - radeon_bo_unref(&rdev->rlc.clear_state_obj); - rdev->rlc.clear_state_obj = NULL; - } -} - -#define RLC_CLEAR_STATE_END_MARKER 0x00000001 - -int si_rlc_init(struct radeon_device *rdev) -{ - volatile u32 *dst_ptr; - u32 dws, data, i, j, k, reg_num; - u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index; - u64 reg_list_mc_addr; - const struct cs_section_def *cs_data = si_cs_data; - int r; - - /* save restore block */ - if (rdev->rlc.save_restore_obj == NULL) { - r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, NULL, - &rdev->rlc.save_restore_obj); - if (r) { - dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r); - return r; - } - } - - r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); - if (unlikely(r != 0)) { - si_rlc_fini(rdev); - return r; - } - r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->rlc.save_restore_gpu_addr); - if (r) { - radeon_bo_unreserve(rdev->rlc.save_restore_obj); - dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r); - si_rlc_fini(rdev); - return r; - } - - if (rdev->family == CHIP_VERDE) { - r = radeon_bo_kmap(rdev->rlc.save_restore_obj, (void **)&rdev->rlc.sr_ptr); - if (r) { - dev_warn(rdev->dev, "(%d) map RLC sr bo failed\n", r); - si_rlc_fini(rdev); - return r; - } - /* write the sr buffer */ - dst_ptr = rdev->rlc.sr_ptr; - for (i = 0; i < ARRAY_SIZE(verde_rlc_save_restore_register_list); i++) { - dst_ptr[i] = verde_rlc_save_restore_register_list[i]; - } - radeon_bo_kunmap(rdev->rlc.save_restore_obj); - } - radeon_bo_unreserve(rdev->rlc.save_restore_obj); - - /* clear state block */ - reg_list_num = 0; - dws = 0; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_list_num++; - dws += cs_data[i].section[j].reg_count; - } - } - reg_list_blk_index = (3 * reg_list_num + 2); - dws += reg_list_blk_index; - - if (rdev->rlc.clear_state_obj == NULL) { - r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); - if (r) { - dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); - si_rlc_fini(rdev); - return r; - } - } - r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); - if (unlikely(r != 0)) { - si_rlc_fini(rdev); - return r; - } - r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM, - &rdev->rlc.clear_state_gpu_addr); - if (r) { - - radeon_bo_unreserve(rdev->rlc.clear_state_obj); - dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r); - si_rlc_fini(rdev); - return r; - } - r = radeon_bo_kmap(rdev->rlc.clear_state_obj, (void **)&rdev->rlc.cs_ptr); - if (r) { - dev_warn(rdev->dev, "(%d) map RLC c bo failed\n", r); - si_rlc_fini(rdev); - return r; - } - /* set up the cs buffer */ - dst_ptr = rdev->rlc.cs_ptr; - reg_list_hdr_blk_index = 0; - reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); - data = upper_32_bits(reg_list_mc_addr); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_num = cs_data[i].section[j].reg_count; - data = reg_list_mc_addr & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = 0x08000000 | (reg_num * 4); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - for (k = 0; k < reg_num; k++) { - data = cs_data[i].section[j].extent[k]; - dst_ptr[reg_list_blk_index + k] = data; - } - reg_list_mc_addr += reg_num * 4; - reg_list_blk_index += reg_num; - } - } - dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; - - radeon_bo_kunmap(rdev->rlc.clear_state_obj); - radeon_bo_unreserve(rdev->rlc.clear_state_obj); - - return 0; -} - void si_rlc_reset(struct radeon_device *rdev) { u32 tmp = RREG32(GRBM_SOFT_RESET); @@ -6449,7 +6291,13 @@ static int si_startup(struct radeon_device *rdev) si_gpu_init(rdev); /* allocate rlc buffers */ - r = si_rlc_init(rdev); + if (rdev->family == CHIP_VERDE) { + rdev->rlc.reg_list = verde_rlc_save_restore_register_list; + rdev->rlc.reg_list_size = + (u32)ARRAY_SIZE(verde_rlc_save_restore_register_list); + } + rdev->rlc.cs_data = si_cs_data; + r = sumo_rlc_init(rdev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -6735,7 +6583,7 @@ int si_init(struct radeon_device *rdev) si_cp_fini(rdev); cayman_dma_fini(rdev); si_irq_fini(rdev); - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); radeon_wb_fini(rdev); radeon_ib_pool_fini(rdev); radeon_vm_manager_fini(rdev); @@ -6761,7 +6609,7 @@ void si_fini(struct radeon_device *rdev) si_cp_fini(rdev); cayman_dma_fini(rdev); si_irq_fini(rdev); - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); si_fini_cg(rdev); si_fini_pg(rdev); radeon_wb_fini(rdev); -- cgit v1.2.3 From 22c775ce80ed921fe9490f3cc2ca66dcda44f572 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 23 Jul 2013 09:41:05 -0400 Subject: drm/radeon: implement clock and power gating for CIK (v3) Only the APUs support power gating. v2: disable cgcg for now v3: workaround hw issue in mgcg Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 1475 +++++++++++++++++++++++++++++++- drivers/gpu/drm/radeon/cikd.h | 95 +- drivers/gpu/drm/radeon/clearstate_ci.h | 944 ++++++++++++++++++++ drivers/gpu/drm/radeon/evergreen.c | 57 +- drivers/gpu/drm/radeon/radeon.h | 14 + drivers/gpu/drm/radeon/radeon_asic.c | 2 + drivers/gpu/drm/radeon/si.c | 2 +- 7 files changed, 2565 insertions(+), 24 deletions(-) create mode 100644 drivers/gpu/drm/radeon/clearstate_ci.h (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 8389917af9a2..a36e98c9a875 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -31,6 +31,7 @@ #include "atom.h" #include "cik_blit_shaders.h" #include "radeon_ucode.h" +#include "clearstate_ci.h" MODULE_FIRMWARE("radeon/BONAIRE_pfp.bin"); MODULE_FIRMWARE("radeon/BONAIRE_me.bin"); @@ -61,9 +62,12 @@ extern void sumo_rlc_fini(struct radeon_device *rdev); extern int sumo_rlc_init(struct radeon_device *rdev); extern void si_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc); extern void si_rlc_reset(struct radeon_device *rdev); +extern void si_init_uvd_internal_cg(struct radeon_device *rdev); static void cik_rlc_stop(struct radeon_device *rdev); static void cik_pcie_gen3_enable(struct radeon_device *rdev); static void cik_program_aspm(struct radeon_device *rdev); +static void cik_init_pg(struct radeon_device *rdev); +static void cik_init_cg(struct radeon_device *rdev); /* * Indirect registers accessor @@ -86,6 +90,778 @@ void cik_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v) (void)RREG32(PCIE_DATA); } +static const u32 spectre_rlc_save_restore_register_list[] = +{ + (0x0e00 << 16) | (0xc12c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc140 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc150 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc15c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc168 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc170 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc178 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc204 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2b8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2bc >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2c0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8228 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x829c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x869c >> 2), + 0x00000000, + (0x0600 << 16) | (0x98f4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x98f8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9900 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc260 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x90e8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c000 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c00c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c1c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9700 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x8e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x9e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0xae00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0xbe00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89bc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8900 >> 2), + 0x00000000, + 0x3, + (0x0e00 << 16) | (0xc130 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc134 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc1fc >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc208 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc264 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc268 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc26c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc270 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc274 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc278 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc27c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc280 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc284 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc288 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc28c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc290 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc294 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc298 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc29c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2ac >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2b0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x301d0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30238 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30250 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30254 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30258 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3025c >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x8e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x9e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0xae00 << 16) | (0xc900 >> 2), + 0x00000000, + (0xbe00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x8e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x9e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0xae00 << 16) | (0xc904 >> 2), + 0x00000000, + (0xbe00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x8e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x9e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0xae00 << 16) | (0xc908 >> 2), + 0x00000000, + (0xbe00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x8e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x9e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0xae00 << 16) | (0xc90c >> 2), + 0x00000000, + (0xbe00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x8e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x9e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0xae00 << 16) | (0xc910 >> 2), + 0x00000000, + (0xbe00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc99c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9834 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f00 >> 2), + 0x00000000, + (0x0001 << 16) | (0x30f00 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f04 >> 2), + 0x00000000, + (0x0001 << 16) | (0x30f04 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f08 >> 2), + 0x00000000, + (0x0001 << 16) | (0x30f08 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f0c >> 2), + 0x00000000, + (0x0001 << 16) | (0x30f0c >> 2), + 0x00000000, + (0x0600 << 16) | (0x9b7c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8a14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8a18 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8bf0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8bcc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8b24 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30a04 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a10 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a14 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a18 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a2c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc700 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc704 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc708 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc768 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc770 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc774 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc778 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc77c >> 2), + 0x00000000, + (0x0400 << 16) | (0xc780 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc784 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc788 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc78c >> 2), + 0x00000000, + (0x0400 << 16) | (0xc798 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc79c >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7a0 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7a4 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7a8 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7ac >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7b0 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc7b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9100 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c010 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92a8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92ac >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92b8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92bc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92c0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92c4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92c8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92cc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x92d0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c04 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c20 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c38 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c3c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xae00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9604 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac08 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac0c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac10 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac58 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac68 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac6c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac70 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac74 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac78 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac7c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac80 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac84 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac88 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac8c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x970c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9714 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9718 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x971c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x4e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x5e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x6e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x7e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x8e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x9e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0xae00 << 16) | (0x31068 >> 2), + 0x00000000, + (0xbe00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd10 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88bc >> 2), + 0x00000000, + (0x0400 << 16) | (0x89c0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88c4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88c8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8980 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30938 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3093c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30940 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89a0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30900 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30904 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c210 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c214 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c218 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8904 >> 2), + 0x00000000, + 0x5, + (0x0e00 << 16) | (0x8c28 >> 2), + (0x0e00 << 16) | (0x8c2c >> 2), + (0x0e00 << 16) | (0x8c30 >> 2), + (0x0e00 << 16) | (0x8c34 >> 2), + (0x0e00 << 16) | (0x9600 >> 2), +}; + +static const u32 kalindi_rlc_save_restore_register_list[] = +{ + (0x0e00 << 16) | (0xc12c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc140 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc150 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc15c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc168 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc170 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc204 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2b8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2bc >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2c0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8228 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x829c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x869c >> 2), + 0x00000000, + (0x0600 << 16) | (0x98f4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x98f8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9900 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc260 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x90e8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c000 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c00c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c1c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9700 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xcd20 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89bc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8900 >> 2), + 0x00000000, + 0x3, + (0x0e00 << 16) | (0xc130 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc134 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc1fc >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc208 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc264 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc268 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc26c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc270 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc274 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc28c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc290 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc294 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc298 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2a8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc2ac >> 2), + 0x00000000, + (0x0e00 << 16) | (0x301d0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30238 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30250 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30254 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30258 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3025c >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc900 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc904 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc908 >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc90c >> 2), + 0x00000000, + (0x4e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x5e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x6e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x7e00 << 16) | (0xc910 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc99c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9834 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f00 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f04 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f08 >> 2), + 0x00000000, + (0x0000 << 16) | (0x30f0c >> 2), + 0x00000000, + (0x0600 << 16) | (0x9b7c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8a14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8a18 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8bf0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8bcc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8b24 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30a04 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a10 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a14 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a18 >> 2), + 0x00000000, + (0x0600 << 16) | (0x30a2c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc700 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc704 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc708 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xc768 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc770 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc774 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc798 >> 2), + 0x00000000, + (0x0400 << 16) | (0xc79c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9100 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c010 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c04 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c20 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c38 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8c3c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xae00 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9604 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac08 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac0c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac10 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac58 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac68 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac6c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac70 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac74 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac78 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac7c >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac80 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac84 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac88 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xac8c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x970c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9714 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x9718 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x971c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x4e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x5e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x6e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x7e00 << 16) | (0x31068 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd10 >> 2), + 0x00000000, + (0x0e00 << 16) | (0xcd14 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88b8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88bc >> 2), + 0x00000000, + (0x0400 << 16) | (0x89c0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88c4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88c8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x88d8 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8980 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30938 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3093c >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30940 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89a0 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30900 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x30904 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x89b4 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3e1fc >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c210 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c214 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x3c218 >> 2), + 0x00000000, + (0x0e00 << 16) | (0x8904 >> 2), + 0x00000000, + 0x5, + (0x0e00 << 16) | (0x8c28 >> 2), + (0x0e00 << 16) | (0x8c2c >> 2), + (0x0e00 << 16) | (0x8c30 >> 2), + (0x0e00 << 16) | (0x8c34 >> 2), + (0x0e00 << 16) | (0x9600 >> 2), +}; + static const u32 bonaire_golden_spm_registers[] = { 0x30800, 0xe0ffffff, 0xe0000000 @@ -4778,6 +5554,39 @@ static void cik_wait_for_rlc_serdes(struct radeon_device *rdev) } } +static void cik_update_rlc(struct radeon_device *rdev, u32 rlc) +{ + u32 tmp; + + tmp = RREG32(RLC_CNTL); + if (tmp != rlc) + WREG32(RLC_CNTL, rlc); +} + +static u32 cik_halt_rlc(struct radeon_device *rdev) +{ + u32 data, orig; + + orig = data = RREG32(RLC_CNTL); + + if (data & RLC_ENABLE) { + u32 i; + + data &= ~RLC_ENABLE; + WREG32(RLC_CNTL, data); + + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(RLC_GPM_STAT) & RLC_GPM_BUSY) == 0) + break; + udelay(1); + } + + cik_wait_for_rlc_serdes(rdev); + } + + return orig; +} + /** * cik_rlc_stop - stop the RLC ME * @@ -4787,20 +5596,10 @@ static void cik_wait_for_rlc_serdes(struct radeon_device *rdev) */ static void cik_rlc_stop(struct radeon_device *rdev) { - u32 tmp; + WREG32(RLC_CNTL, 0); cik_enable_gui_idle_interrupt(rdev, false); - RREG32(CB_CGTT_SCLK_CTRL); - RREG32(CB_CGTT_SCLK_CTRL); - RREG32(CB_CGTT_SCLK_CTRL); - RREG32(CB_CGTT_SCLK_CTRL); - - tmp = RREG32(RLC_CGCG_CGLS_CTRL) & 0xfffffffc; - WREG32(RLC_CGCG_CGLS_CTRL, tmp); - - WREG32(RLC_CNTL, 0); - cik_wait_for_rlc_serdes(rdev); } @@ -4831,8 +5630,7 @@ static void cik_rlc_start(struct radeon_device *rdev) */ static int cik_rlc_resume(struct radeon_device *rdev) { - u32 i, size; - u32 clear_state_info[3]; + u32 i, size, tmp; const __be32 *fw_data; if (!rdev->rlc_fw) @@ -4853,8 +5651,16 @@ static int cik_rlc_resume(struct radeon_device *rdev) cik_rlc_stop(rdev); + /* disable CG */ + tmp = RREG32(RLC_CGCG_CGLS_CTRL) & 0xfffffffc; + WREG32(RLC_CGCG_CGLS_CTRL, tmp); + si_rlc_reset(rdev); + cik_init_pg(rdev); + + cik_init_cg(rdev); + WREG32(RLC_LB_CNTR_INIT, 0); WREG32(RLC_LB_CNTR_MAX, 0x00008000); @@ -4875,20 +5681,634 @@ static int cik_rlc_resume(struct radeon_device *rdev) /* XXX - find out what chips support lbpw */ cik_enable_lbpw(rdev, false); - /* XXX */ - clear_state_info[0] = 0;//upper_32_bits(rdev->rlc.save_restore_gpu_addr); - clear_state_info[1] = 0;//rdev->rlc.save_restore_gpu_addr; - clear_state_info[2] = 0;//cik_default_size; - WREG32(RLC_GPM_SCRATCH_ADDR, 0x3d); - for (i = 0; i < 3; i++) - WREG32(RLC_GPM_SCRATCH_DATA, clear_state_info[i]); - WREG32(RLC_DRIVER_DMA_STATUS, 0); + if (rdev->family == CHIP_BONAIRE) + WREG32(RLC_DRIVER_DMA_STATUS, 0); cik_rlc_start(rdev); return 0; } +static void cik_enable_cgcg(struct radeon_device *rdev, bool enable) +{ + u32 data, orig, tmp, tmp2; + + orig = data = RREG32(RLC_CGCG_CGLS_CTRL); + + cik_enable_gui_idle_interrupt(rdev, enable); + + if (enable) { + tmp = cik_halt_rlc(rdev); + + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); + WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); + tmp2 = BPM_ADDR_MASK | CGCG_OVERRIDE_0 | CGLS_ENABLE; + WREG32(RLC_SERDES_WR_CTRL, tmp2); + + cik_update_rlc(rdev, tmp); + + data |= CGCG_EN | CGLS_EN; + } else { + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + + data &= ~(CGCG_EN | CGLS_EN); + } + + if (orig != data) + WREG32(RLC_CGCG_CGLS_CTRL, data); + +} + +static void cik_enable_mgcg(struct radeon_device *rdev, bool enable) +{ + u32 data, orig, tmp = 0; + + if (enable) { + orig = data = RREG32(CP_MEM_SLP_CNTL); + data |= CP_MEM_LS_EN; + if (orig != data) + WREG32(CP_MEM_SLP_CNTL, data); + + orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); + data &= 0xfffffffd; + if (orig != data) + WREG32(RLC_CGTT_MGCG_OVERRIDE, data); + + tmp = cik_halt_rlc(rdev); + + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); + WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); + data = BPM_ADDR_MASK | MGCG_OVERRIDE_0; + WREG32(RLC_SERDES_WR_CTRL, data); + + cik_update_rlc(rdev, tmp); + + orig = data = RREG32(CGTS_SM_CTRL_REG); + data &= ~SM_MODE_MASK; + data |= SM_MODE(0x2); + data |= SM_MODE_ENABLE; + data &= ~CGTS_OVERRIDE; + data &= ~CGTS_LS_OVERRIDE; + data &= ~ON_MONITOR_ADD_MASK; + data |= ON_MONITOR_ADD_EN; + data |= ON_MONITOR_ADD(0x96); + if (orig != data) + WREG32(CGTS_SM_CTRL_REG, data); + } else { + orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); + data |= 0x00000002; + if (orig != data) + WREG32(RLC_CGTT_MGCG_OVERRIDE, data); + + data = RREG32(RLC_MEM_SLP_CNTL); + if (data & RLC_MEM_LS_EN) { + data &= ~RLC_MEM_LS_EN; + WREG32(RLC_MEM_SLP_CNTL, data); + } + + data = RREG32(CP_MEM_SLP_CNTL); + if (data & CP_MEM_LS_EN) { + data &= ~CP_MEM_LS_EN; + WREG32(CP_MEM_SLP_CNTL, data); + } + + orig = data = RREG32(CGTS_SM_CTRL_REG); + data |= CGTS_OVERRIDE | CGTS_LS_OVERRIDE; + if (orig != data) + WREG32(CGTS_SM_CTRL_REG, data); + + tmp = cik_halt_rlc(rdev); + + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + WREG32(RLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff); + WREG32(RLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff); + data = BPM_ADDR_MASK | MGCG_OVERRIDE_1; + WREG32(RLC_SERDES_WR_CTRL, data); + + cik_update_rlc(rdev, tmp); + } +} + +static const u32 mc_cg_registers[] = +{ + MC_HUB_MISC_HUB_CG, + MC_HUB_MISC_SIP_CG, + MC_HUB_MISC_VM_CG, + MC_XPB_CLK_GAT, + ATC_MISC_CG, + MC_CITF_MISC_WR_CG, + MC_CITF_MISC_RD_CG, + MC_CITF_MISC_VM_CG, + VM_L2_CG, +}; + +static void cik_enable_mc_ls(struct radeon_device *rdev, + bool enable) +{ + int i; + u32 orig, data; + + for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) { + orig = data = RREG32(mc_cg_registers[i]); + if (enable) + data |= MC_LS_ENABLE; + else + data &= ~MC_LS_ENABLE; + if (data != orig) + WREG32(mc_cg_registers[i], data); + } +} + +static void cik_enable_mc_mgcg(struct radeon_device *rdev, + bool enable) +{ + int i; + u32 orig, data; + + for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) { + orig = data = RREG32(mc_cg_registers[i]); + if (enable) + data |= MC_CG_ENABLE; + else + data &= ~MC_CG_ENABLE; + if (data != orig) + WREG32(mc_cg_registers[i], data); + } +} + +static void cik_enable_sdma_mgcg(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data; + + if (enable) { + WREG32(SDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET, 0x00000100); + WREG32(SDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET, 0x00000100); + } else { + orig = data = RREG32(SDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET); + data |= 0xff000000; + if (data != orig) + WREG32(SDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET, data); + + orig = data = RREG32(SDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET); + data |= 0xff000000; + if (data != orig) + WREG32(SDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET, data); + } +} + +static void cik_enable_sdma_mgls(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data; + + if (enable) { + orig = data = RREG32(SDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET); + data |= 0x100; + if (orig != data) + WREG32(SDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET, data); + + orig = data = RREG32(SDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET); + data |= 0x100; + if (orig != data) + WREG32(SDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET, data); + } else { + orig = data = RREG32(SDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET); + data &= ~0x100; + if (orig != data) + WREG32(SDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET, data); + + orig = data = RREG32(SDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET); + data &= ~0x100; + if (orig != data) + WREG32(SDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET, data); + } +} + +static void cik_enable_uvd_mgcg(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data; + + if (enable) { + data = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL); + data = 0xfff; + WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, data); + + orig = data = RREG32(UVD_CGC_CTRL); + data |= DCM; + if (orig != data) + WREG32(UVD_CGC_CTRL, data); + } else { + data = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL); + data &= ~0xfff; + WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, data); + + orig = data = RREG32(UVD_CGC_CTRL); + data &= ~DCM; + if (orig != data) + WREG32(UVD_CGC_CTRL, data); + } +} + +static void cik_enable_hdp_mgcg(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data; + + orig = data = RREG32(HDP_HOST_PATH_CNTL); + + if (enable) + data &= ~CLOCK_GATING_DIS; + else + data |= CLOCK_GATING_DIS; + + if (orig != data) + WREG32(HDP_HOST_PATH_CNTL, data); +} + +static void cik_enable_hdp_ls(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data; + + orig = data = RREG32(HDP_MEM_POWER_LS); + + if (enable) + data |= HDP_LS_ENABLE; + else + data &= ~HDP_LS_ENABLE; + + if (orig != data) + WREG32(HDP_MEM_POWER_LS, data); +} + +void cik_update_cg(struct radeon_device *rdev, + u32 block, bool enable) +{ + if (block & RADEON_CG_BLOCK_GFX) { + /* order matters! */ + if (enable) { + cik_enable_mgcg(rdev, true); + cik_enable_cgcg(rdev, true); + } else { + cik_enable_cgcg(rdev, false); + cik_enable_mgcg(rdev, false); + } + } + + if (block & RADEON_CG_BLOCK_MC) { + if (!(rdev->flags & RADEON_IS_IGP)) { + cik_enable_mc_mgcg(rdev, enable); + cik_enable_mc_ls(rdev, enable); + } + } + + if (block & RADEON_CG_BLOCK_SDMA) { + cik_enable_sdma_mgcg(rdev, enable); + cik_enable_sdma_mgls(rdev, enable); + } + + if (block & RADEON_CG_BLOCK_UVD) { + if (rdev->has_uvd) + cik_enable_uvd_mgcg(rdev, enable); + } + + if (block & RADEON_CG_BLOCK_HDP) { + cik_enable_hdp_mgcg(rdev, enable); + cik_enable_hdp_ls(rdev, enable); + } +} + +static void cik_init_cg(struct radeon_device *rdev) +{ + + cik_update_cg(rdev, RADEON_CG_BLOCK_GFX, false); /* XXX true */ + + if (rdev->has_uvd) + si_init_uvd_internal_cg(rdev); + + cik_update_cg(rdev, (RADEON_CG_BLOCK_MC | + RADEON_CG_BLOCK_SDMA | + RADEON_CG_BLOCK_UVD | + RADEON_CG_BLOCK_HDP), true); +} + +static void cik_enable_sck_slowdown_on_pu(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data |= SMU_CLK_SLOWDOWN_ON_PU_ENABLE; + else + data &= ~SMU_CLK_SLOWDOWN_ON_PU_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +static void cik_enable_sck_slowdown_on_pd(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data |= SMU_CLK_SLOWDOWN_ON_PD_ENABLE; + else + data &= ~SMU_CLK_SLOWDOWN_ON_PD_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +static void cik_enable_cp_pg(struct radeon_device *rdev, bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data &= ~DISABLE_CP_PG; + else + data |= DISABLE_CP_PG; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +static void cik_enable_gds_pg(struct radeon_device *rdev, bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data &= ~DISABLE_GDS_PG; + else + data |= DISABLE_GDS_PG; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +#define CP_ME_TABLE_SIZE 96 +#define CP_ME_TABLE_OFFSET 2048 +#define CP_MEC_TABLE_OFFSET 4096 + +void cik_init_cp_pg_table(struct radeon_device *rdev) +{ + const __be32 *fw_data; + volatile u32 *dst_ptr; + int me, i, max_me = 4; + u32 bo_offset = 0; + u32 table_offset; + + if (rdev->family == CHIP_KAVERI) + max_me = 5; + + if (rdev->rlc.cp_table_ptr == NULL) + return; + + /* write the cp table buffer */ + dst_ptr = rdev->rlc.cp_table_ptr; + for (me = 0; me < max_me; me++) { + if (me == 0) { + fw_data = (const __be32 *)rdev->ce_fw->data; + table_offset = CP_ME_TABLE_OFFSET; + } else if (me == 1) { + fw_data = (const __be32 *)rdev->pfp_fw->data; + table_offset = CP_ME_TABLE_OFFSET; + } else if (me == 2) { + fw_data = (const __be32 *)rdev->me_fw->data; + table_offset = CP_ME_TABLE_OFFSET; + } else { + fw_data = (const __be32 *)rdev->mec_fw->data; + table_offset = CP_MEC_TABLE_OFFSET; + } + + for (i = 0; i < CP_ME_TABLE_SIZE; i ++) { + dst_ptr[bo_offset + i] = be32_to_cpu(fw_data[table_offset + i]); + } + bo_offset += CP_ME_TABLE_SIZE; + } +} + +static void cik_enable_gfx_cgpg(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig; + + if (enable) { + orig = data = RREG32(RLC_PG_CNTL); + data |= GFX_PG_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); + + orig = data = RREG32(RLC_AUTO_PG_CTRL); + data |= AUTO_PG_EN; + if (orig != data) + WREG32(RLC_AUTO_PG_CTRL, data); + } else { + orig = data = RREG32(RLC_PG_CNTL); + data &= ~GFX_PG_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); + + orig = data = RREG32(RLC_AUTO_PG_CTRL); + data &= ~AUTO_PG_EN; + if (orig != data) + WREG32(RLC_AUTO_PG_CTRL, data); + + data = RREG32(DB_RENDER_CONTROL); + } +} + +static u32 cik_get_cu_active_bitmap(struct radeon_device *rdev, u32 se, u32 sh) +{ + u32 mask = 0, tmp, tmp1; + int i; + + cik_select_se_sh(rdev, se, sh); + tmp = RREG32(CC_GC_SHADER_ARRAY_CONFIG); + tmp1 = RREG32(GC_USER_SHADER_ARRAY_CONFIG); + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + + tmp &= 0xffff0000; + + tmp |= tmp1; + tmp >>= 16; + + for (i = 0; i < rdev->config.cik.max_cu_per_sh; i ++) { + mask <<= 1; + mask |= 1; + } + + return (~tmp) & mask; +} + +static void cik_init_ao_cu_mask(struct radeon_device *rdev) +{ + u32 i, j, k, active_cu_number = 0; + u32 mask, counter, cu_bitmap; + u32 tmp = 0; + + for (i = 0; i < rdev->config.cik.max_shader_engines; i++) { + for (j = 0; j < rdev->config.cik.max_sh_per_se; j++) { + mask = 1; + cu_bitmap = 0; + counter = 0; + for (k = 0; k < rdev->config.cik.max_cu_per_sh; k ++) { + if (cik_get_cu_active_bitmap(rdev, i, j) & mask) { + if (counter < 2) + cu_bitmap |= mask; + counter ++; + } + mask <<= 1; + } + + active_cu_number += counter; + tmp |= (cu_bitmap << (i * 16 + j * 8)); + } + } + + WREG32(RLC_PG_AO_CU_MASK, tmp); + + tmp = RREG32(RLC_MAX_PG_CU); + tmp &= ~MAX_PU_CU_MASK; + tmp |= MAX_PU_CU(active_cu_number); + WREG32(RLC_MAX_PG_CU, tmp); +} + +static void cik_enable_gfx_static_mgpg(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data |= STATIC_PER_CU_PG_ENABLE; + else + data &= ~STATIC_PER_CU_PG_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +static void cik_enable_gfx_dynamic_mgpg(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig; + + orig = data = RREG32(RLC_PG_CNTL); + if (enable) + data |= DYN_PER_CU_PG_ENABLE; + else + data &= ~DYN_PER_CU_PG_ENABLE; + if (orig != data) + WREG32(RLC_PG_CNTL, data); +} + +#define RLC_SAVE_AND_RESTORE_STARTING_OFFSET 0x90 +#define RLC_CLEAR_STATE_DESCRIPTOR_OFFSET 0x3D + +static void cik_init_gfx_cgpg(struct radeon_device *rdev) +{ + u32 data, orig; + u32 i; + + if (rdev->rlc.cs_data) { + WREG32(RLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET); + WREG32(RLC_GPM_SCRATCH_DATA, upper_32_bits(rdev->rlc.clear_state_gpu_addr)); + WREG32(RLC_GPM_SCRATCH_DATA, rdev->rlc.clear_state_gpu_addr); + WREG32(RLC_GPM_SCRATCH_DATA, rdev->rlc.clear_state_size); + } else { + WREG32(RLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET); + for (i = 0; i < 3; i++) + WREG32(RLC_GPM_SCRATCH_DATA, 0); + } + if (rdev->rlc.reg_list) { + WREG32(RLC_GPM_SCRATCH_ADDR, RLC_SAVE_AND_RESTORE_STARTING_OFFSET); + for (i = 0; i < rdev->rlc.reg_list_size; i++) + WREG32(RLC_GPM_SCRATCH_DATA, rdev->rlc.reg_list[i]); + } + + orig = data = RREG32(RLC_PG_CNTL); + data |= GFX_PG_SRC; + if (orig != data) + WREG32(RLC_PG_CNTL, data); + + WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); + WREG32(RLC_CP_TABLE_RESTORE, rdev->rlc.cp_table_gpu_addr >> 8); + + data = RREG32(CP_RB_WPTR_POLL_CNTL); + data &= ~IDLE_POLL_COUNT_MASK; + data |= IDLE_POLL_COUNT(0x60); + WREG32(CP_RB_WPTR_POLL_CNTL, data); + + data = 0x10101010; + WREG32(RLC_PG_DELAY, data); + + data = RREG32(RLC_PG_DELAY_2); + data &= ~0xff; + data |= 0x3; + WREG32(RLC_PG_DELAY_2, data); + + data = RREG32(RLC_AUTO_PG_CTRL); + data &= ~GRBM_REG_SGIT_MASK; + data |= GRBM_REG_SGIT(0x700); + WREG32(RLC_AUTO_PG_CTRL, data); + +} + +static void cik_update_gfx_pg(struct radeon_device *rdev, bool enable) +{ + bool has_pg = false; + bool has_dyn_mgpg = false; + bool has_static_mgpg = false; + + /* only APUs have PG */ + if (rdev->flags & RADEON_IS_IGP) { + has_pg = true; + has_static_mgpg = true; + if (rdev->family == CHIP_KAVERI) + has_dyn_mgpg = true; + } + + if (has_pg) { + cik_enable_gfx_cgpg(rdev, enable); + if (enable) { + cik_enable_gfx_static_mgpg(rdev, has_static_mgpg); + cik_enable_gfx_dynamic_mgpg(rdev, has_dyn_mgpg); + } else { + cik_enable_gfx_static_mgpg(rdev, false); + cik_enable_gfx_dynamic_mgpg(rdev, false); + } + } + +} + +void cik_init_pg(struct radeon_device *rdev) +{ + bool has_pg = false; + + /* only APUs have PG */ + if (rdev->flags & RADEON_IS_IGP) { + /* XXX disable this for now */ + /* has_pg = true; */ + } + + if (has_pg) { + cik_enable_sck_slowdown_on_pu(rdev, true); + cik_enable_sck_slowdown_on_pd(rdev, true); + cik_init_gfx_cgpg(rdev); + cik_enable_cp_pg(rdev, true); + cik_enable_gds_pg(rdev, true); + cik_init_ao_cu_mask(rdev); + cik_update_gfx_pg(rdev, true); + } +} + /* * Interrupts * Starting with r6xx, interrupts are handled via a ring buffer. @@ -6019,6 +7439,19 @@ static int cik_startup(struct radeon_device *rdev) cik_gpu_init(rdev); /* allocate rlc buffers */ + if (rdev->flags & RADEON_IS_IGP) { + if (rdev->family == CHIP_KAVERI) { + rdev->rlc.reg_list = spectre_rlc_save_restore_register_list; + rdev->rlc.reg_list_size = + (u32)ARRAY_SIZE(spectre_rlc_save_restore_register_list); + } else { + rdev->rlc.reg_list = kalindi_rlc_save_restore_register_list; + rdev->rlc.reg_list_size = + (u32)ARRAY_SIZE(kalindi_rlc_save_restore_register_list); + } + } + rdev->rlc.cs_data = ci_cs_data; + rdev->rlc.cp_table_size = CP_ME_TABLE_SIZE * 5 * 4; r = sumo_rlc_init(rdev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 4742f3db4aa6..63955abb1e11 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -299,6 +299,10 @@ #define VM_CONTEXT0_PAGE_TABLE_END_ADDR 0x157C #define VM_CONTEXT1_PAGE_TABLE_END_ADDR 0x1580 +#define VM_L2_CG 0x15c0 +#define MC_CG_ENABLE (1 << 18) +#define MC_LS_ENABLE (1 << 19) + #define MC_SHARED_CHMAP 0x2004 #define NOOFCHAN_SHIFT 12 #define NOOFCHAN_MASK 0x0000f000 @@ -328,6 +332,17 @@ #define MC_SHARED_BLACKOUT_CNTL 0x20ac +#define MC_HUB_MISC_HUB_CG 0x20b8 +#define MC_HUB_MISC_VM_CG 0x20bc + +#define MC_HUB_MISC_SIP_CG 0x20c0 + +#define MC_XPB_CLK_GAT 0x2478 + +#define MC_CITF_MISC_RD_CG 0x2648 +#define MC_CITF_MISC_WR_CG 0x264c +#define MC_CITF_MISC_VM_CG 0x2650 + #define MC_ARB_RAMCFG 0x2760 #define NOOFBANK_SHIFT 0 #define NOOFBANK_MASK 0x00000003 @@ -357,6 +372,7 @@ #define MC_SEQ_IO_DEBUG_DATA 0x2a48 #define HDP_HOST_PATH_CNTL 0x2C00 +#define CLOCK_GATING_DIS (1 << 23) #define HDP_NONSURFACE_BASE 0x2C04 #define HDP_NONSURFACE_INFO 0x2C08 #define HDP_NONSURFACE_SIZE 0x2C0C @@ -364,6 +380,10 @@ #define HDP_ADDR_CONFIG 0x2F48 #define HDP_MISC_CNTL 0x2F4C #define HDP_FLUSH_INVALIDATE_CACHE (1 << 0) +#define HDP_MEM_POWER_LS 0x2F50 +#define HDP_LS_ENABLE (1 << 0) + +#define ATC_MISC_CG 0x3350 #define IH_RB_CNTL 0x3e00 # define IH_RB_ENABLE (1 << 0) @@ -631,6 +651,9 @@ #define CP_RB0_RPTR 0x8700 #define CP_RB_WPTR_DELAY 0x8704 +#define CP_RB_WPTR_POLL_CNTL 0x8708 +#define IDLE_POLL_COUNT(x) ((x) << 16) +#define IDLE_POLL_COUNT_MASK (0xffff << 16) #define CP_MEQ_THRESHOLDS 0x8764 #define MEQ1_START(x) ((x) << 0) @@ -857,6 +880,9 @@ # define CP_RINGID1_INT_STAT (1 << 30) # define CP_RINGID0_INT_STAT (1 << 31) +#define CP_MEM_SLP_CNTL 0xC1E4 +# define CP_MEM_LS_EN (1 << 0) + #define CP_CPF_DEBUG 0xC200 #define CP_PQ_WPTR_POLL_CNTL 0xC20C @@ -902,6 +928,9 @@ #define RLC_MC_CNTL 0xC30C +#define RLC_MEM_SLP_CNTL 0xC318 +# define RLC_MEM_LS_EN (1 << 0) + #define RLC_LB_CNTR_MAX 0xC348 #define RLC_LB_CNTL 0xC364 @@ -910,7 +939,9 @@ #define RLC_LB_CNTR_INIT 0xC36C #define RLC_SAVE_AND_RESTORE_BASE 0xC374 -#define RLC_DRIVER_DMA_STATUS 0xC378 +#define RLC_DRIVER_DMA_STATUS 0xC378 /* dGPU */ +#define RLC_CP_TABLE_RESTORE 0xC378 /* APU */ +#define RLC_PG_DELAY_2 0xC37C #define RLC_GPM_UCODE_ADDR 0xC388 #define RLC_GPM_UCODE_DATA 0xC38C @@ -919,12 +950,50 @@ #define RLC_CAPTURE_GPU_CLOCK_COUNT 0xC398 #define RLC_UCODE_CNTL 0xC39C +#define RLC_GPM_STAT 0xC400 +# define RLC_GPM_BUSY (1 << 0) + +#define RLC_PG_CNTL 0xC40C +# define GFX_PG_ENABLE (1 << 0) +# define GFX_PG_SRC (1 << 1) +# define DYN_PER_CU_PG_ENABLE (1 << 2) +# define STATIC_PER_CU_PG_ENABLE (1 << 3) +# define DISABLE_GDS_PG (1 << 13) +# define DISABLE_CP_PG (1 << 15) +# define SMU_CLK_SLOWDOWN_ON_PU_ENABLE (1 << 17) +# define SMU_CLK_SLOWDOWN_ON_PD_ENABLE (1 << 18) + +#define RLC_CGTT_MGCG_OVERRIDE 0xC420 #define RLC_CGCG_CGLS_CTRL 0xC424 +# define CGCG_EN (1 << 0) +# define CGLS_EN (1 << 1) + +#define RLC_PG_DELAY 0xC434 #define RLC_LB_INIT_CU_MASK 0xC43C #define RLC_LB_PARAMS 0xC444 +#define RLC_PG_AO_CU_MASK 0xC44C + +#define RLC_MAX_PG_CU 0xC450 +# define MAX_PU_CU(x) ((x) << 0) +# define MAX_PU_CU_MASK (0xff << 0) +#define RLC_AUTO_PG_CTRL 0xC454 +# define AUTO_PG_EN (1 << 0) +# define GRBM_REG_SGIT(x) ((x) << 3) +# define GRBM_REG_SGIT_MASK (0xffff << 3) + +#define RLC_SERDES_WR_CU_MASTER_MASK 0xC474 +#define RLC_SERDES_WR_NONCU_MASTER_MASK 0xC478 +#define RLC_SERDES_WR_CTRL 0xC47C +#define BPM_ADDR(x) ((x) << 0) +#define BPM_ADDR_MASK (0xff << 0) +#define CGLS_ENABLE (1 << 16) +#define CGCG_OVERRIDE_0 (1 << 20) +#define MGCG_OVERRIDE_0 (1 << 22) +#define MGCG_OVERRIDE_1 (1 << 23) + #define RLC_SERDES_CU_MASTER_BUSY 0xC484 #define RLC_SERDES_NONCU_MASTER_BUSY 0xC488 # define SE_MASTER_BUSY_MASK 0x0000ffff @@ -979,6 +1048,8 @@ #define MQD_VMID(x) ((x) << 0) #define MQD_VMID_MASK (0xf << 0) +#define DB_RENDER_CONTROL 0x28000 + #define PA_SC_RASTER_CONFIG 0x28350 # define RASTER_CONFIG_RB_MAP_0 0 # define RASTER_CONFIG_RB_MAP_1 1 @@ -1072,6 +1143,16 @@ #define CP_PERFMON_CNTL 0x36020 +#define CGTS_SM_CTRL_REG 0x3c000 +#define SM_MODE(x) ((x) << 17) +#define SM_MODE_MASK (0x7 << 17) +#define SM_MODE_ENABLE (1 << 20) +#define CGTS_OVERRIDE (1 << 21) +#define CGTS_LS_OVERRIDE (1 << 22) +#define ON_MONITOR_ADD_EN (1 << 23) +#define ON_MONITOR_ADD(x) ((x) << 24) +#define ON_MONITOR_ADD_MASK (0xff << 24) + #define CGTS_TCC_DISABLE 0x3c00c #define CGTS_USER_TCC_DISABLE 0x3c010 #define TCC_DISABLE_MASK 0xFFFF0000 @@ -1304,6 +1385,8 @@ #define SDMA0_UCODE_ADDR 0xD000 #define SDMA0_UCODE_DATA 0xD004 +#define SDMA0_POWER_CNTL 0xD008 +#define SDMA0_CLK_CTRL 0xD00C #define SDMA0_CNTL 0xD010 # define TRAP_ENABLE (1 << 0) @@ -1428,6 +1511,13 @@ #define UVD_RBC_RB_RPTR 0xf690 #define UVD_RBC_RB_WPTR 0xf694 +#define UVD_CGC_CTRL 0xF4B0 +# define DCM (1 << 0) +# define CG_DT(x) ((x) << 2) +# define CG_DT_MASK (0xf << 2) +# define CLK_OD(x) ((x) << 6) +# define CLK_OD_MASK (0x1f << 6) + /* UVD clocks */ #define CG_DCLK_CNTL 0xC050009C @@ -1438,4 +1528,7 @@ #define CG_VCLK_CNTL 0xC05000A4 #define CG_VCLK_STATUS 0xC05000A8 +/* UVD CTX indirect */ +#define UVD_CGC_MEM_CTRL 0xC0 + #endif diff --git a/drivers/gpu/drm/radeon/clearstate_ci.h b/drivers/gpu/drm/radeon/clearstate_ci.h new file mode 100644 index 000000000000..c3982f9475fb --- /dev/null +++ b/drivers/gpu/drm/radeon/clearstate_ci.h @@ -0,0 +1,944 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +static const unsigned int ci_SECT_CONTEXT_def_1[] = +{ + 0x00000000, // DB_RENDER_CONTROL + 0x00000000, // DB_COUNT_CONTROL + 0x00000000, // DB_DEPTH_VIEW + 0x00000000, // DB_RENDER_OVERRIDE + 0x00000000, // DB_RENDER_OVERRIDE2 + 0x00000000, // DB_HTILE_DATA_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_DEPTH_BOUNDS_MIN + 0x00000000, // DB_DEPTH_BOUNDS_MAX + 0x00000000, // DB_STENCIL_CLEAR + 0x00000000, // DB_DEPTH_CLEAR + 0x00000000, // PA_SC_SCREEN_SCISSOR_TL + 0x40004000, // PA_SC_SCREEN_SCISSOR_BR + 0, // HOLE + 0x00000000, // DB_DEPTH_INFO + 0x00000000, // DB_Z_INFO + 0x00000000, // DB_STENCIL_INFO + 0x00000000, // DB_Z_READ_BASE + 0x00000000, // DB_STENCIL_READ_BASE + 0x00000000, // DB_Z_WRITE_BASE + 0x00000000, // DB_STENCIL_WRITE_BASE + 0x00000000, // DB_DEPTH_SIZE + 0x00000000, // DB_DEPTH_SLICE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // TA_BC_BASE_ADDR + 0x00000000, // TA_BC_BASE_ADDR_HI + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // COHER_DEST_BASE_HI_0 + 0x00000000, // COHER_DEST_BASE_HI_1 + 0x00000000, // COHER_DEST_BASE_HI_2 + 0x00000000, // COHER_DEST_BASE_HI_3 + 0x00000000, // COHER_DEST_BASE_2 + 0x00000000, // COHER_DEST_BASE_3 + 0x00000000, // PA_SC_WINDOW_OFFSET + 0x80000000, // PA_SC_WINDOW_SCISSOR_TL + 0x40004000, // PA_SC_WINDOW_SCISSOR_BR + 0x0000ffff, // PA_SC_CLIPRECT_RULE + 0x00000000, // PA_SC_CLIPRECT_0_TL + 0x40004000, // PA_SC_CLIPRECT_0_BR + 0x00000000, // PA_SC_CLIPRECT_1_TL + 0x40004000, // PA_SC_CLIPRECT_1_BR + 0x00000000, // PA_SC_CLIPRECT_2_TL + 0x40004000, // PA_SC_CLIPRECT_2_BR + 0x00000000, // PA_SC_CLIPRECT_3_TL + 0x40004000, // PA_SC_CLIPRECT_3_BR + 0xaa99aaaa, // PA_SC_EDGERULE + 0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET + 0xffffffff, // CB_TARGET_MASK + 0xffffffff, // CB_SHADER_MASK + 0x80000000, // PA_SC_GENERIC_SCISSOR_TL + 0x40004000, // PA_SC_GENERIC_SCISSOR_BR + 0x00000000, // COHER_DEST_BASE_0 + 0x00000000, // COHER_DEST_BASE_1 + 0x80000000, // PA_SC_VPORT_SCISSOR_0_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_0_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_1_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_1_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_2_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_2_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_3_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_3_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_4_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_4_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_5_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_5_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_6_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_6_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_7_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_7_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_8_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_8_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_9_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_9_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_10_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_10_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_11_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_11_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_12_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_12_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_13_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_13_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_14_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_14_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_15_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_15_BR + 0x00000000, // PA_SC_VPORT_ZMIN_0 + 0x3f800000, // PA_SC_VPORT_ZMAX_0 + 0x00000000, // PA_SC_VPORT_ZMIN_1 + 0x3f800000, // PA_SC_VPORT_ZMAX_1 + 0x00000000, // PA_SC_VPORT_ZMIN_2 + 0x3f800000, // PA_SC_VPORT_ZMAX_2 + 0x00000000, // PA_SC_VPORT_ZMIN_3 + 0x3f800000, // PA_SC_VPORT_ZMAX_3 + 0x00000000, // PA_SC_VPORT_ZMIN_4 + 0x3f800000, // PA_SC_VPORT_ZMAX_4 + 0x00000000, // PA_SC_VPORT_ZMIN_5 + 0x3f800000, // PA_SC_VPORT_ZMAX_5 + 0x00000000, // PA_SC_VPORT_ZMIN_6 + 0x3f800000, // PA_SC_VPORT_ZMAX_6 + 0x00000000, // PA_SC_VPORT_ZMIN_7 + 0x3f800000, // PA_SC_VPORT_ZMAX_7 + 0x00000000, // PA_SC_VPORT_ZMIN_8 + 0x3f800000, // PA_SC_VPORT_ZMAX_8 + 0x00000000, // PA_SC_VPORT_ZMIN_9 + 0x3f800000, // PA_SC_VPORT_ZMAX_9 + 0x00000000, // PA_SC_VPORT_ZMIN_10 + 0x3f800000, // PA_SC_VPORT_ZMAX_10 + 0x00000000, // PA_SC_VPORT_ZMIN_11 + 0x3f800000, // PA_SC_VPORT_ZMAX_11 + 0x00000000, // PA_SC_VPORT_ZMIN_12 + 0x3f800000, // PA_SC_VPORT_ZMAX_12 + 0x00000000, // PA_SC_VPORT_ZMIN_13 + 0x3f800000, // PA_SC_VPORT_ZMAX_13 + 0x00000000, // PA_SC_VPORT_ZMIN_14 + 0x3f800000, // PA_SC_VPORT_ZMAX_14 + 0x00000000, // PA_SC_VPORT_ZMIN_15 + 0x3f800000, // PA_SC_VPORT_ZMAX_15 +}; +static const unsigned int ci_SECT_CONTEXT_def_2[] = +{ + 0x00000000, // PA_SC_SCREEN_EXTENT_CONTROL + 0, // HOLE + 0x00000000, // CP_PERFMON_CNTX_CNTL + 0x00000000, // CP_RINGID + 0x00000000, // CP_VMID + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0xffffffff, // VGT_MAX_VTX_INDX + 0x00000000, // VGT_MIN_VTX_INDX + 0x00000000, // VGT_INDX_OFFSET + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX + 0, // HOLE + 0x00000000, // CB_BLEND_RED + 0x00000000, // CB_BLEND_GREEN + 0x00000000, // CB_BLEND_BLUE + 0x00000000, // CB_BLEND_ALPHA + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCIL_CONTROL + 0x00000000, // DB_STENCILREFMASK + 0x00000000, // DB_STENCILREFMASK_BF + 0, // HOLE + 0x00000000, // PA_CL_VPORT_XSCALE + 0x00000000, // PA_CL_VPORT_XOFFSET + 0x00000000, // PA_CL_VPORT_YSCALE + 0x00000000, // PA_CL_VPORT_YOFFSET + 0x00000000, // PA_CL_VPORT_ZSCALE + 0x00000000, // PA_CL_VPORT_ZOFFSET + 0x00000000, // PA_CL_VPORT_XSCALE_1 + 0x00000000, // PA_CL_VPORT_XOFFSET_1 + 0x00000000, // PA_CL_VPORT_YSCALE_1 + 0x00000000, // PA_CL_VPORT_YOFFSET_1 + 0x00000000, // PA_CL_VPORT_ZSCALE_1 + 0x00000000, // PA_CL_VPORT_ZOFFSET_1 + 0x00000000, // PA_CL_VPORT_XSCALE_2 + 0x00000000, // PA_CL_VPORT_XOFFSET_2 + 0x00000000, // PA_CL_VPORT_YSCALE_2 + 0x00000000, // PA_CL_VPORT_YOFFSET_2 + 0x00000000, // PA_CL_VPORT_ZSCALE_2 + 0x00000000, // PA_CL_VPORT_ZOFFSET_2 + 0x00000000, // PA_CL_VPORT_XSCALE_3 + 0x00000000, // PA_CL_VPORT_XOFFSET_3 + 0x00000000, // PA_CL_VPORT_YSCALE_3 + 0x00000000, // PA_CL_VPORT_YOFFSET_3 + 0x00000000, // PA_CL_VPORT_ZSCALE_3 + 0x00000000, // PA_CL_VPORT_ZOFFSET_3 + 0x00000000, // PA_CL_VPORT_XSCALE_4 + 0x00000000, // PA_CL_VPORT_XOFFSET_4 + 0x00000000, // PA_CL_VPORT_YSCALE_4 + 0x00000000, // PA_CL_VPORT_YOFFSET_4 + 0x00000000, // PA_CL_VPORT_ZSCALE_4 + 0x00000000, // PA_CL_VPORT_ZOFFSET_4 + 0x00000000, // PA_CL_VPORT_XSCALE_5 + 0x00000000, // PA_CL_VPORT_XOFFSET_5 + 0x00000000, // PA_CL_VPORT_YSCALE_5 + 0x00000000, // PA_CL_VPORT_YOFFSET_5 + 0x00000000, // PA_CL_VPORT_ZSCALE_5 + 0x00000000, // PA_CL_VPORT_ZOFFSET_5 + 0x00000000, // PA_CL_VPORT_XSCALE_6 + 0x00000000, // PA_CL_VPORT_XOFFSET_6 + 0x00000000, // PA_CL_VPORT_YSCALE_6 + 0x00000000, // PA_CL_VPORT_YOFFSET_6 + 0x00000000, // PA_CL_VPORT_ZSCALE_6 + 0x00000000, // PA_CL_VPORT_ZOFFSET_6 + 0x00000000, // PA_CL_VPORT_XSCALE_7 + 0x00000000, // PA_CL_VPORT_XOFFSET_7 + 0x00000000, // PA_CL_VPORT_YSCALE_7 + 0x00000000, // PA_CL_VPORT_YOFFSET_7 + 0x00000000, // PA_CL_VPORT_ZSCALE_7 + 0x00000000, // PA_CL_VPORT_ZOFFSET_7 + 0x00000000, // PA_CL_VPORT_XSCALE_8 + 0x00000000, // PA_CL_VPORT_XOFFSET_8 + 0x00000000, // PA_CL_VPORT_YSCALE_8 + 0x00000000, // PA_CL_VPORT_YOFFSET_8 + 0x00000000, // PA_CL_VPORT_ZSCALE_8 + 0x00000000, // PA_CL_VPORT_ZOFFSET_8 + 0x00000000, // PA_CL_VPORT_XSCALE_9 + 0x00000000, // PA_CL_VPORT_XOFFSET_9 + 0x00000000, // PA_CL_VPORT_YSCALE_9 + 0x00000000, // PA_CL_VPORT_YOFFSET_9 + 0x00000000, // PA_CL_VPORT_ZSCALE_9 + 0x00000000, // PA_CL_VPORT_ZOFFSET_9 + 0x00000000, // PA_CL_VPORT_XSCALE_10 + 0x00000000, // PA_CL_VPORT_XOFFSET_10 + 0x00000000, // PA_CL_VPORT_YSCALE_10 + 0x00000000, // PA_CL_VPORT_YOFFSET_10 + 0x00000000, // PA_CL_VPORT_ZSCALE_10 + 0x00000000, // PA_CL_VPORT_ZOFFSET_10 + 0x00000000, // PA_CL_VPORT_XSCALE_11 + 0x00000000, // PA_CL_VPORT_XOFFSET_11 + 0x00000000, // PA_CL_VPORT_YSCALE_11 + 0x00000000, // PA_CL_VPORT_YOFFSET_11 + 0x00000000, // PA_CL_VPORT_ZSCALE_11 + 0x00000000, // PA_CL_VPORT_ZOFFSET_11 + 0x00000000, // PA_CL_VPORT_XSCALE_12 + 0x00000000, // PA_CL_VPORT_XOFFSET_12 + 0x00000000, // PA_CL_VPORT_YSCALE_12 + 0x00000000, // PA_CL_VPORT_YOFFSET_12 + 0x00000000, // PA_CL_VPORT_ZSCALE_12 + 0x00000000, // PA_CL_VPORT_ZOFFSET_12 + 0x00000000, // PA_CL_VPORT_XSCALE_13 + 0x00000000, // PA_CL_VPORT_XOFFSET_13 + 0x00000000, // PA_CL_VPORT_YSCALE_13 + 0x00000000, // PA_CL_VPORT_YOFFSET_13 + 0x00000000, // PA_CL_VPORT_ZSCALE_13 + 0x00000000, // PA_CL_VPORT_ZOFFSET_13 + 0x00000000, // PA_CL_VPORT_XSCALE_14 + 0x00000000, // PA_CL_VPORT_XOFFSET_14 + 0x00000000, // PA_CL_VPORT_YSCALE_14 + 0x00000000, // PA_CL_VPORT_YOFFSET_14 + 0x00000000, // PA_CL_VPORT_ZSCALE_14 + 0x00000000, // PA_CL_VPORT_ZOFFSET_14 + 0x00000000, // PA_CL_VPORT_XSCALE_15 + 0x00000000, // PA_CL_VPORT_XOFFSET_15 + 0x00000000, // PA_CL_VPORT_YSCALE_15 + 0x00000000, // PA_CL_VPORT_YOFFSET_15 + 0x00000000, // PA_CL_VPORT_ZSCALE_15 + 0x00000000, // PA_CL_VPORT_ZOFFSET_15 + 0x00000000, // PA_CL_UCP_0_X + 0x00000000, // PA_CL_UCP_0_Y + 0x00000000, // PA_CL_UCP_0_Z + 0x00000000, // PA_CL_UCP_0_W + 0x00000000, // PA_CL_UCP_1_X + 0x00000000, // PA_CL_UCP_1_Y + 0x00000000, // PA_CL_UCP_1_Z + 0x00000000, // PA_CL_UCP_1_W + 0x00000000, // PA_CL_UCP_2_X + 0x00000000, // PA_CL_UCP_2_Y + 0x00000000, // PA_CL_UCP_2_Z + 0x00000000, // PA_CL_UCP_2_W + 0x00000000, // PA_CL_UCP_3_X + 0x00000000, // PA_CL_UCP_3_Y + 0x00000000, // PA_CL_UCP_3_Z + 0x00000000, // PA_CL_UCP_3_W + 0x00000000, // PA_CL_UCP_4_X + 0x00000000, // PA_CL_UCP_4_Y + 0x00000000, // PA_CL_UCP_4_Z + 0x00000000, // PA_CL_UCP_4_W + 0x00000000, // PA_CL_UCP_5_X + 0x00000000, // PA_CL_UCP_5_Y + 0x00000000, // PA_CL_UCP_5_Z + 0x00000000, // PA_CL_UCP_5_W + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_CNTL_0 + 0x00000000, // SPI_PS_INPUT_CNTL_1 + 0x00000000, // SPI_PS_INPUT_CNTL_2 + 0x00000000, // SPI_PS_INPUT_CNTL_3 + 0x00000000, // SPI_PS_INPUT_CNTL_4 + 0x00000000, // SPI_PS_INPUT_CNTL_5 + 0x00000000, // SPI_PS_INPUT_CNTL_6 + 0x00000000, // SPI_PS_INPUT_CNTL_7 + 0x00000000, // SPI_PS_INPUT_CNTL_8 + 0x00000000, // SPI_PS_INPUT_CNTL_9 + 0x00000000, // SPI_PS_INPUT_CNTL_10 + 0x00000000, // SPI_PS_INPUT_CNTL_11 + 0x00000000, // SPI_PS_INPUT_CNTL_12 + 0x00000000, // SPI_PS_INPUT_CNTL_13 + 0x00000000, // SPI_PS_INPUT_CNTL_14 + 0x00000000, // SPI_PS_INPUT_CNTL_15 + 0x00000000, // SPI_PS_INPUT_CNTL_16 + 0x00000000, // SPI_PS_INPUT_CNTL_17 + 0x00000000, // SPI_PS_INPUT_CNTL_18 + 0x00000000, // SPI_PS_INPUT_CNTL_19 + 0x00000000, // SPI_PS_INPUT_CNTL_20 + 0x00000000, // SPI_PS_INPUT_CNTL_21 + 0x00000000, // SPI_PS_INPUT_CNTL_22 + 0x00000000, // SPI_PS_INPUT_CNTL_23 + 0x00000000, // SPI_PS_INPUT_CNTL_24 + 0x00000000, // SPI_PS_INPUT_CNTL_25 + 0x00000000, // SPI_PS_INPUT_CNTL_26 + 0x00000000, // SPI_PS_INPUT_CNTL_27 + 0x00000000, // SPI_PS_INPUT_CNTL_28 + 0x00000000, // SPI_PS_INPUT_CNTL_29 + 0x00000000, // SPI_PS_INPUT_CNTL_30 + 0x00000000, // SPI_PS_INPUT_CNTL_31 + 0x00000000, // SPI_VS_OUT_CONFIG + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_ENA + 0x00000000, // SPI_PS_INPUT_ADDR + 0x00000000, // SPI_INTERP_CONTROL_0 + 0x00000002, // SPI_PS_IN_CONTROL + 0, // HOLE + 0x00000000, // SPI_BARYC_CNTL + 0, // HOLE + 0x00000000, // SPI_TMPRING_SIZE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_SHADER_POS_FORMAT + 0x00000000, // SPI_SHADER_Z_FORMAT + 0x00000000, // SPI_SHADER_COL_FORMAT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_BLEND0_CONTROL + 0x00000000, // CB_BLEND1_CONTROL + 0x00000000, // CB_BLEND2_CONTROL + 0x00000000, // CB_BLEND3_CONTROL + 0x00000000, // CB_BLEND4_CONTROL + 0x00000000, // CB_BLEND5_CONTROL + 0x00000000, // CB_BLEND6_CONTROL + 0x00000000, // CB_BLEND7_CONTROL +}; +static const unsigned int ci_SECT_CONTEXT_def_3[] = +{ + 0x00000000, // PA_CL_POINT_X_RAD + 0x00000000, // PA_CL_POINT_Y_RAD + 0x00000000, // PA_CL_POINT_SIZE + 0x00000000, // PA_CL_POINT_CULL_RAD + 0x00000000, // VGT_DMA_BASE_HI + 0x00000000, // VGT_DMA_BASE +}; +static const unsigned int ci_SECT_CONTEXT_def_4[] = +{ + 0x00000000, // DB_DEPTH_CONTROL + 0x00000000, // DB_EQAA + 0x00000000, // CB_COLOR_CONTROL + 0x00000000, // DB_SHADER_CONTROL + 0x00090000, // PA_CL_CLIP_CNTL + 0x00000004, // PA_SU_SC_MODE_CNTL + 0x00000000, // PA_CL_VTE_CNTL + 0x00000000, // PA_CL_VS_OUT_CNTL + 0x00000000, // PA_CL_NANINF_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_SCALE + 0x00000000, // PA_SU_PRIM_FILTER_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SU_POINT_SIZE + 0x00000000, // PA_SU_POINT_MINMAX + 0x00000000, // PA_SU_LINE_CNTL + 0x00000000, // PA_SC_LINE_STIPPLE + 0x00000000, // VGT_OUTPUT_PATH_CNTL + 0x00000000, // VGT_HOS_CNTL + 0x00000000, // VGT_HOS_MAX_TESS_LEVEL + 0x00000000, // VGT_HOS_MIN_TESS_LEVEL + 0x00000000, // VGT_HOS_REUSE_DEPTH + 0x00000000, // VGT_GROUP_PRIM_TYPE + 0x00000000, // VGT_GROUP_FIRST_DECR + 0x00000000, // VGT_GROUP_DECR + 0x00000000, // VGT_GROUP_VECT_0_CNTL + 0x00000000, // VGT_GROUP_VECT_1_CNTL + 0x00000000, // VGT_GROUP_VECT_0_FMT_CNTL + 0x00000000, // VGT_GROUP_VECT_1_FMT_CNTL + 0x00000000, // VGT_GS_MODE + 0x00000000, // VGT_GS_ONCHIP_CNTL + 0x00000000, // PA_SC_MODE_CNTL_0 + 0x00000000, // PA_SC_MODE_CNTL_1 + 0x00000000, // VGT_ENHANCE + 0x00000100, // VGT_GS_PER_ES + 0x00000080, // VGT_ES_PER_GS + 0x00000002, // VGT_GS_PER_VS + 0x00000000, // VGT_GSVS_RING_OFFSET_1 + 0x00000000, // VGT_GSVS_RING_OFFSET_2 + 0x00000000, // VGT_GSVS_RING_OFFSET_3 + 0x00000000, // VGT_GS_OUT_PRIM_TYPE + 0x00000000, // IA_ENHANCE +}; +static const unsigned int ci_SECT_CONTEXT_def_5[] = +{ + 0x00000000, // WD_ENHANCE + 0x00000000, // VGT_PRIMITIVEID_EN +}; +static const unsigned int ci_SECT_CONTEXT_def_6[] = +{ + 0x00000000, // VGT_PRIMITIVEID_RESET +}; +static const unsigned int ci_SECT_CONTEXT_def_7[] = +{ + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_INSTANCE_STEP_RATE_0 + 0x00000000, // VGT_INSTANCE_STEP_RATE_1 + 0x000000ff, // IA_MULTI_VGT_PARAM + 0x00000000, // VGT_ESGS_RING_ITEMSIZE + 0x00000000, // VGT_GSVS_RING_ITEMSIZE + 0x00000000, // VGT_REUSE_OFF + 0x00000000, // VGT_VTX_CNT_EN + 0x00000000, // DB_HTILE_SURFACE + 0x00000000, // DB_SRESULTS_COMPARE_STATE0 + 0x00000000, // DB_SRESULTS_COMPARE_STATE1 + 0x00000000, // DB_PRELOAD_CONTROL + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_0 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_0 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_1 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_1 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_2 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_2 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_3 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_3 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_3 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE + 0, // HOLE + 0x00000000, // VGT_GS_MAX_VERT_OUT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_SHADER_STAGES_EN + 0x00000000, // VGT_LS_HS_CONFIG + 0x00000000, // VGT_GS_VERT_ITEMSIZE + 0x00000000, // VGT_GS_VERT_ITEMSIZE_1 + 0x00000000, // VGT_GS_VERT_ITEMSIZE_2 + 0x00000000, // VGT_GS_VERT_ITEMSIZE_3 + 0x00000000, // VGT_TF_PARAM + 0x00000000, // DB_ALPHA_TO_MASK + 0, // HOLE + 0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL + 0x00000000, // PA_SU_POLY_OFFSET_CLAMP + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET + 0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET + 0x00000000, // VGT_GS_INSTANCE_CNT + 0x00000000, // VGT_STRMOUT_CONFIG + 0x00000000, // VGT_STRMOUT_BUFFER_CONFIG + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_CENTROID_PRIORITY_0 + 0x00000000, // PA_SC_CENTROID_PRIORITY_1 + 0x00001000, // PA_SC_LINE_CNTL + 0x00000000, // PA_SC_AA_CONFIG + 0x00000005, // PA_SU_VTX_CNTL + 0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ + 0x3f800000, // PA_CL_GB_VERT_DISC_ADJ + 0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ + 0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3 + 0xffffffff, // PA_SC_AA_MASK_X0Y0_X1Y0 + 0xffffffff, // PA_SC_AA_MASK_X0Y1_X1Y1 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x0000000e, // VGT_VERTEX_REUSE_BLOCK_CNTL + 0x00000010, // VGT_OUT_DEALLOC_CNTL + 0x00000000, // CB_COLOR0_BASE + 0x00000000, // CB_COLOR0_PITCH + 0x00000000, // CB_COLOR0_SLICE + 0x00000000, // CB_COLOR0_VIEW + 0x00000000, // CB_COLOR0_INFO + 0x00000000, // CB_COLOR0_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR0_CMASK + 0x00000000, // CB_COLOR0_CMASK_SLICE + 0x00000000, // CB_COLOR0_FMASK + 0x00000000, // CB_COLOR0_FMASK_SLICE + 0x00000000, // CB_COLOR0_CLEAR_WORD0 + 0x00000000, // CB_COLOR0_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR1_BASE + 0x00000000, // CB_COLOR1_PITCH + 0x00000000, // CB_COLOR1_SLICE + 0x00000000, // CB_COLOR1_VIEW + 0x00000000, // CB_COLOR1_INFO + 0x00000000, // CB_COLOR1_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR1_CMASK + 0x00000000, // CB_COLOR1_CMASK_SLICE + 0x00000000, // CB_COLOR1_FMASK + 0x00000000, // CB_COLOR1_FMASK_SLICE + 0x00000000, // CB_COLOR1_CLEAR_WORD0 + 0x00000000, // CB_COLOR1_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR2_BASE + 0x00000000, // CB_COLOR2_PITCH + 0x00000000, // CB_COLOR2_SLICE + 0x00000000, // CB_COLOR2_VIEW + 0x00000000, // CB_COLOR2_INFO + 0x00000000, // CB_COLOR2_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR2_CMASK + 0x00000000, // CB_COLOR2_CMASK_SLICE + 0x00000000, // CB_COLOR2_FMASK + 0x00000000, // CB_COLOR2_FMASK_SLICE + 0x00000000, // CB_COLOR2_CLEAR_WORD0 + 0x00000000, // CB_COLOR2_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR3_BASE + 0x00000000, // CB_COLOR3_PITCH + 0x00000000, // CB_COLOR3_SLICE + 0x00000000, // CB_COLOR3_VIEW + 0x00000000, // CB_COLOR3_INFO + 0x00000000, // CB_COLOR3_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR3_CMASK + 0x00000000, // CB_COLOR3_CMASK_SLICE + 0x00000000, // CB_COLOR3_FMASK + 0x00000000, // CB_COLOR3_FMASK_SLICE + 0x00000000, // CB_COLOR3_CLEAR_WORD0 + 0x00000000, // CB_COLOR3_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR4_BASE + 0x00000000, // CB_COLOR4_PITCH + 0x00000000, // CB_COLOR4_SLICE + 0x00000000, // CB_COLOR4_VIEW + 0x00000000, // CB_COLOR4_INFO + 0x00000000, // CB_COLOR4_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR4_CMASK + 0x00000000, // CB_COLOR4_CMASK_SLICE + 0x00000000, // CB_COLOR4_FMASK + 0x00000000, // CB_COLOR4_FMASK_SLICE + 0x00000000, // CB_COLOR4_CLEAR_WORD0 + 0x00000000, // CB_COLOR4_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR5_BASE + 0x00000000, // CB_COLOR5_PITCH + 0x00000000, // CB_COLOR5_SLICE + 0x00000000, // CB_COLOR5_VIEW + 0x00000000, // CB_COLOR5_INFO + 0x00000000, // CB_COLOR5_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR5_CMASK + 0x00000000, // CB_COLOR5_CMASK_SLICE + 0x00000000, // CB_COLOR5_FMASK + 0x00000000, // CB_COLOR5_FMASK_SLICE + 0x00000000, // CB_COLOR5_CLEAR_WORD0 + 0x00000000, // CB_COLOR5_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR6_BASE + 0x00000000, // CB_COLOR6_PITCH + 0x00000000, // CB_COLOR6_SLICE + 0x00000000, // CB_COLOR6_VIEW + 0x00000000, // CB_COLOR6_INFO + 0x00000000, // CB_COLOR6_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR6_CMASK + 0x00000000, // CB_COLOR6_CMASK_SLICE + 0x00000000, // CB_COLOR6_FMASK + 0x00000000, // CB_COLOR6_FMASK_SLICE + 0x00000000, // CB_COLOR6_CLEAR_WORD0 + 0x00000000, // CB_COLOR6_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR7_BASE + 0x00000000, // CB_COLOR7_PITCH + 0x00000000, // CB_COLOR7_SLICE + 0x00000000, // CB_COLOR7_VIEW + 0x00000000, // CB_COLOR7_INFO + 0x00000000, // CB_COLOR7_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR7_CMASK + 0x00000000, // CB_COLOR7_CMASK_SLICE + 0x00000000, // CB_COLOR7_FMASK + 0x00000000, // CB_COLOR7_FMASK_SLICE + 0x00000000, // CB_COLOR7_CLEAR_WORD0 + 0x00000000, // CB_COLOR7_CLEAR_WORD1 +}; +static const struct cs_extent_def ci_SECT_CONTEXT_defs[] = +{ + {ci_SECT_CONTEXT_def_1, 0x0000a000, 212 }, + {ci_SECT_CONTEXT_def_2, 0x0000a0d6, 274 }, + {ci_SECT_CONTEXT_def_3, 0x0000a1f5, 6 }, + {ci_SECT_CONTEXT_def_4, 0x0000a200, 157 }, + {ci_SECT_CONTEXT_def_5, 0x0000a2a0, 2 }, + {ci_SECT_CONTEXT_def_6, 0x0000a2a3, 1 }, + {ci_SECT_CONTEXT_def_7, 0x0000a2a5, 233 }, + { 0, 0, 0 } +}; +static const struct cs_section_def ci_cs_data[] = { + { ci_SECT_CONTEXT_defs, SECT_CONTEXT }, + { 0, SECT_NONE } +}; diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 6fc876a444d4..2ce12ee3e67f 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -140,6 +140,7 @@ extern void cayman_cp_int_cntl_setup(struct radeon_device *rdev, int ring, u32 cp_int_cntl); extern void cayman_vm_decode_fault(struct radeon_device *rdev, u32 status, u32 addr); +void cik_init_cp_pg_table(struct radeon_device *rdev); static const u32 evergreen_golden_registers[] = { @@ -3893,8 +3894,22 @@ void sumo_rlc_fini(struct radeon_device *rdev) radeon_bo_unref(&rdev->rlc.clear_state_obj); rdev->rlc.clear_state_obj = NULL; } + + /* clear state block */ + if (rdev->rlc.cp_table_obj) { + r = radeon_bo_reserve(rdev->rlc.cp_table_obj, false); + if (unlikely(r != 0)) + dev_warn(rdev->dev, "(%d) reserve RLC cp table bo failed\n", r); + radeon_bo_unpin(rdev->rlc.cp_table_obj); + radeon_bo_unreserve(rdev->rlc.cp_table_obj); + + radeon_bo_unref(&rdev->rlc.cp_table_obj); + rdev->rlc.cp_table_obj = NULL; + } } +#define CP_ME_TABLE_SIZE 96 + int sumo_rlc_init(struct radeon_device *rdev) { const u32 *src_ptr; @@ -3980,9 +3995,10 @@ int sumo_rlc_init(struct radeon_device *rdev) } reg_list_blk_index = (3 * reg_list_num + 2); dws += reg_list_blk_index; + rdev->rlc.clear_state_size = dws; if (rdev->rlc.clear_state_obj == NULL) { - r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + r = radeon_bo_create(rdev, rdev->rlc.clear_state_size * 4, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); if (r) { dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); @@ -4046,6 +4062,45 @@ int sumo_rlc_init(struct radeon_device *rdev) radeon_bo_unreserve(rdev->rlc.clear_state_obj); } + if (rdev->rlc.cp_table_size) { + if (rdev->rlc.cp_table_obj == NULL) { + r = radeon_bo_create(rdev, rdev->rlc.cp_table_size, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.cp_table_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create RLC cp table bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + } + + r = radeon_bo_reserve(rdev->rlc.cp_table_obj, false); + if (unlikely(r != 0)) { + dev_warn(rdev->dev, "(%d) reserve RLC cp table bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->rlc.cp_table_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->rlc.cp_table_gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->rlc.cp_table_obj); + dev_warn(rdev->dev, "(%d) pin RLC cp_table bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_kmap(rdev->rlc.cp_table_obj, (void **)&rdev->rlc.cp_table_ptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC cp table bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + + cik_init_cp_pg_table(rdev); + + radeon_bo_kunmap(rdev->rlc.cp_table_obj); + radeon_bo_unreserve(rdev->rlc.cp_table_obj); + + } + return 0; } diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index b9706e83e827..5941ada063d1 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -152,6 +152,14 @@ extern int radeon_aspm; #define RADEON_RESET_MC (1 << 10) #define RADEON_RESET_DISPLAY (1 << 11) +/* CG block flags */ +#define RADEON_CG_BLOCK_GFX (1 << 0) +#define RADEON_CG_BLOCK_MC (1 << 1) +#define RADEON_CG_BLOCK_SDMA (1 << 2) +#define RADEON_CG_BLOCK_UVD (1 << 3) +#define RADEON_CG_BLOCK_VCE (1 << 4) +#define RADEON_CG_BLOCK_HDP (1 << 5) + /* max cursor sizes (in pixels) */ #define CURSOR_WIDTH 64 #define CURSOR_HEIGHT 64 @@ -861,6 +869,12 @@ struct radeon_rlc { uint64_t clear_state_gpu_addr; volatile uint32_t *cs_ptr; const struct cs_section_def *cs_data; + u32 clear_state_size; + /* for cp tables */ + struct radeon_bo *cp_table_obj; + uint64_t cp_table_gpu_addr; + volatile uint32_t *cp_table_ptr; + u32 cp_table_size; }; int radeon_ib_get(struct radeon_device *rdev, int ring, diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index 1926ec06a638..880551b6df61 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -2780,6 +2780,7 @@ int radeon_asic_init(struct radeon_device *rdev) case CHIP_BONAIRE: rdev->asic = &ci_asic; rdev->num_crtc = 6; + rdev->has_uvd = true; break; case CHIP_KAVERI: case CHIP_KABINI: @@ -2789,6 +2790,7 @@ int radeon_asic_init(struct radeon_device *rdev) rdev->num_crtc = 4; else rdev->num_crtc = 2; + rdev->has_uvd = true; break; default: /* FIXME: not supported yet */ diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 8b8963d4a732..4f91e1f4d814 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -4901,7 +4901,7 @@ static void si_set_uvd_dcm(struct radeon_device *rdev, WREG32_UVD_CTX(UVD_CGC_CTRL2, tmp2); } -static void si_init_uvd_internal_cg(struct radeon_device *rdev) +void si_init_uvd_internal_cg(struct radeon_device *rdev) { bool hw_mode = true; -- cgit v1.2.3 From 5e884f606cdba9c599c9c9373808f272ae794088 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 6 Aug 2013 11:39:38 -0400 Subject: drm/radeon: restructure UVD code to handle UVD PG (v2) When we PG (powergate) UVD, we need to re-initialize it before we can use it again. v2: rebase on UVD stop fixes Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 14 +++---- drivers/gpu/drm/radeon/evergreen.c | 2 +- drivers/gpu/drm/radeon/ni.c | 2 +- drivers/gpu/drm/radeon/r600.c | 74 +++++++++++++++++++----------------- drivers/gpu/drm/radeon/radeon_asic.h | 4 +- drivers/gpu/drm/radeon/rv770.c | 2 +- drivers/gpu/drm/radeon/si.c | 2 +- 7 files changed, 50 insertions(+), 50 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 736a416b51a7..59b866aa08d9 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -69,6 +69,7 @@ static void cik_pcie_gen3_enable(struct radeon_device *rdev); static void cik_program_aspm(struct radeon_device *rdev); static void cik_init_pg(struct radeon_device *rdev); static void cik_init_cg(struct radeon_device *rdev); +static void cik_uvd_resume(struct radeon_device *rdev); /* get temperature in millidegrees */ int ci_get_temp(struct radeon_device *rdev) @@ -7619,8 +7620,9 @@ static int cik_startup(struct radeon_device *rdev) return r; } - r = cik_uvd_resume(rdev); + r = radeon_uvd_resume(rdev); if (!r) { + cik_uvd_resume(rdev); r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); if (r) @@ -7708,7 +7710,7 @@ static int cik_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev); + r = r600_uvd_init(rdev, true); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } @@ -8598,15 +8600,10 @@ int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) return r; } -int cik_uvd_resume(struct radeon_device *rdev) +static void cik_uvd_resume(struct radeon_device *rdev) { uint64_t addr; uint32_t size; - int r; - - r = radeon_uvd_resume(rdev); - if (r) - return r; /* programm the VCPU memory controller bits 0-27 */ addr = rdev->uvd.gpu_addr >> 3; @@ -8632,7 +8629,6 @@ int cik_uvd_resume(struct radeon_device *rdev) addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); - return 0; } static void cik_pcie_gen3_enable(struct radeon_device *rdev) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 2ce12ee3e67f..710c1d4ae5db 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5296,7 +5296,7 @@ static int evergreen_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev); + r = r600_uvd_init(rdev, true); if (r) DRM_ERROR("radeon: error initializing UVD (%d).\n", r); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 21f2eceff2c6..bc298a3500a4 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2230,7 +2230,7 @@ static int cayman_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev); + r = r600_uvd_init(rdev, true); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 3db2e4ddb2d6..8a600153ef6c 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2623,7 +2623,7 @@ void r600_dma_fini(struct radeon_device *rdev) /* * UVD */ -int r600_uvd_rbc_start(struct radeon_device *rdev) +static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test) { struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; uint64_t rptr_addr; @@ -2664,47 +2664,47 @@ int r600_uvd_rbc_start(struct radeon_device *rdev) rb_bufsz = (0x1 << 8) | rb_bufsz; WREG32(UVD_RBC_RB_CNTL, rb_bufsz); - ring->ready = true; - r = radeon_ring_test(rdev, R600_RING_TYPE_UVD_INDEX, ring); - if (r) { - ring->ready = false; - return r; - } + if (ring_test) { + ring->ready = true; + r = radeon_ring_test(rdev, R600_RING_TYPE_UVD_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } - r = radeon_ring_lock(rdev, ring, 10); - if (r) { - DRM_ERROR("radeon: ring failed to lock UVD ring (%d).\n", r); - return r; - } + r = radeon_ring_lock(rdev, ring, 10); + if (r) { + DRM_ERROR("radeon: ring failed to lock UVD ring (%d).\n", r); + return r; + } - tmp = PACKET0(UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); + tmp = PACKET0(UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); - tmp = PACKET0(UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); + tmp = PACKET0(UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); - tmp = PACKET0(UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); + tmp = PACKET0(UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); - /* Clear timeout status bits */ - radeon_ring_write(ring, PACKET0(UVD_SEMA_TIMEOUT_STATUS, 0)); - radeon_ring_write(ring, 0x8); + /* Clear timeout status bits */ + radeon_ring_write(ring, PACKET0(UVD_SEMA_TIMEOUT_STATUS, 0)); + radeon_ring_write(ring, 0x8); - radeon_ring_write(ring, PACKET0(UVD_SEMA_CNTL, 0)); - radeon_ring_write(ring, 3); + radeon_ring_write(ring, PACKET0(UVD_SEMA_CNTL, 0)); + radeon_ring_write(ring, 3); - radeon_ring_unlock_commit(rdev, ring); + radeon_ring_unlock_commit(rdev, ring); + } return 0; } -void r600_uvd_stop(struct radeon_device *rdev) +void r600_do_uvd_stop(struct radeon_device *rdev) { - struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - /* force RBC into idle state */ WREG32(UVD_RBC_RB_CNTL, 0x11010101); @@ -2723,11 +2723,17 @@ void r600_uvd_stop(struct radeon_device *rdev) /* Unstall UMC and register bus */ WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); +} +void r600_uvd_stop(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + + r600_do_uvd_stop(rdev); ring->ready = false; } -int r600_uvd_init(struct radeon_device *rdev) +int r600_uvd_init(struct radeon_device *rdev, bool ring_test) { int i, j, r; /* disable byte swapping */ @@ -2815,17 +2821,17 @@ int r600_uvd_init(struct radeon_device *rdev) if (r) { DRM_ERROR("UVD not responding, giving up!!!\n"); - radeon_set_uvd_clocks(rdev, 0, 0); - return r; + goto done; } /* enable interupt */ WREG32_P(UVD_MASTINT_EN, 3<<1, ~(3 << 1)); - r = r600_uvd_rbc_start(rdev); + r = r600_uvd_rbc_start(rdev, ring_test); if (!r) DRM_INFO("UVD initialized successfully.\n"); +done: /* lower clocks again */ radeon_set_uvd_clocks(rdev, 0, 0); diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 1e386c48ae2d..3570817a5847 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -424,8 +424,7 @@ void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rde struct seq_file *m); /* uvd */ -int r600_uvd_init(struct radeon_device *rdev); -int r600_uvd_rbc_start(struct radeon_device *rdev); +int r600_uvd_init(struct radeon_device *rdev, bool ring_test); void r600_uvd_stop(struct radeon_device *rdev); int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); void r600_uvd_fence_emit(struct radeon_device *rdev, @@ -696,7 +695,6 @@ u32 cik_get_xclk(struct radeon_device *rdev); uint32_t cik_pciep_rreg(struct radeon_device *rdev, uint32_t reg); void cik_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); -int cik_uvd_resume(struct radeon_device *rdev); void cik_sdma_fence_ring_emit(struct radeon_device *rdev, struct radeon_fence *fence); void cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index 95590bd07afb..52253b2ab0d5 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1928,7 +1928,7 @@ static int rv770_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev); + r = r600_uvd_init(rdev, true); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 4f91e1f4d814..da23ce8f4388 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6421,7 +6421,7 @@ static int si_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev); + r = r600_uvd_init(rdev, true); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } -- cgit v1.2.3 From 02c9f7fa4e7230fc4ae8bf26f64e45aa76011f9c Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 13 Aug 2013 11:56:51 +0200 Subject: drm/radeon: rework UVD writeback & [rw]ptr handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The hardware just doesn't support this correctly. Disable it before we accidentally write anywhere we shouldn't. Signed-off-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 3 +-- drivers/gpu/drm/radeon/evergreen.c | 3 +-- drivers/gpu/drm/radeon/ni.c | 3 +-- drivers/gpu/drm/radeon/r600.c | 33 ++++++++++++++++++++------------- drivers/gpu/drm/radeon/radeon.h | 1 - drivers/gpu/drm/radeon/radeon_asic.c | 12 ++++++------ drivers/gpu/drm/radeon/radeon_asic.h | 6 ++++++ drivers/gpu/drm/radeon/radeon_ring.c | 2 +- drivers/gpu/drm/radeon/rv770.c | 3 +-- drivers/gpu/drm/radeon/si.c | 3 +-- 10 files changed, 38 insertions(+), 31 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index e661aec734b2..ce7036ae9f5a 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -7705,8 +7705,7 @@ static int cik_startup(struct radeon_device *rdev) ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; if (ring->ring_size) { - r = radeon_ring_init(rdev, ring, ring->ring_size, - R600_WB_UVD_RPTR_OFFSET, + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 710c1d4ae5db..2139f6c64341 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5291,8 +5291,7 @@ static int evergreen_startup(struct radeon_device *rdev) ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; if (ring->ring_size) { - r = radeon_ring_init(rdev, ring, ring->ring_size, - R600_WB_UVD_RPTR_OFFSET, + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index bc298a3500a4..f543f4ca4dda 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2225,8 +2225,7 @@ static int cayman_startup(struct radeon_device *rdev) ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; if (ring->ring_size) { - r = radeon_ring_init(rdev, ring, ring->ring_size, - R600_WB_UVD_RPTR_OFFSET, + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 8a600153ef6c..c1b0aba4431a 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2623,31 +2623,38 @@ void r600_dma_fini(struct radeon_device *rdev) /* * UVD */ +uint32_t r600_uvd_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return RREG32(UVD_RBC_RB_RPTR); +} + +uint32_t r600_uvd_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return RREG32(UVD_RBC_RB_WPTR); +} + +void r600_uvd_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + WREG32(UVD_RBC_RB_WPTR, ring->wptr); +} + static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test) { struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - uint64_t rptr_addr; uint32_t rb_bufsz, tmp; int r; - rptr_addr = rdev->wb.gpu_addr + R600_WB_UVD_RPTR_OFFSET; - - if (upper_32_bits(rptr_addr) != upper_32_bits(ring->gpu_addr)) { - DRM_ERROR("UVD ring and rptr not in the same 4GB segment!\n"); - return -EINVAL; - } - /* force RBC into idle state */ WREG32(UVD_RBC_RB_CNTL, 0x11010101); /* Set the write pointer delay */ WREG32(UVD_RBC_RB_WPTR_CNTL, 0); - /* set the wb address */ - WREG32(UVD_RBC_RB_RPTR_ADDR, rptr_addr >> 2); - /* programm the 4GB memory segment for rptr and ring buffer */ - WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(rptr_addr) | + WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) | (0x7 << 16) | (0x1 << 31)); /* Initialize the ring buffer's read and write pointers */ @@ -2662,7 +2669,7 @@ static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test) /* Set ring buffer size */ rb_bufsz = drm_order(ring->ring_size); rb_bufsz = (0x1 << 8) | rb_bufsz; - WREG32(UVD_RBC_RB_CNTL, rb_bufsz); + WREG32_P(UVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f); if (ring_test) { ring->ready = true; diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index b26a20fe2859..2eab174bf22e 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1027,7 +1027,6 @@ struct radeon_wb { #define R600_WB_DMA_RPTR_OFFSET 1792 #define R600_WB_IH_WPTR_OFFSET 2048 #define CAYMAN_WB_DMA1_RPTR_OFFSET 2304 -#define R600_WB_UVD_RPTR_OFFSET 2560 #define R600_WB_EVENT_OFFSET 3072 #define CIK_WB_CP1_WPTR_OFFSET 3328 #define CIK_WB_CP2_WPTR_OFFSET 3584 diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index 012fe7218c74..7432247a812a 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -1157,9 +1157,9 @@ static struct radeon_asic_ring rv770_uvd_ring = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_uvd_get_rptr, + .get_wptr = &r600_uvd_get_wptr, + .set_wptr = &r600_uvd_set_wptr, }; static struct radeon_asic rv770_asic = { @@ -1593,9 +1593,9 @@ static struct radeon_asic_ring cayman_uvd_ring = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_uvd_get_rptr, + .get_wptr = &r600_uvd_get_wptr, + .set_wptr = &r600_uvd_set_wptr, }; static struct radeon_asic cayman_asic = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 5630291c4b06..37baf9c696f0 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -424,6 +424,12 @@ void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rde struct seq_file *m); /* uvd */ +uint32_t r600_uvd_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +uint32_t r600_uvd_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void r600_uvd_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); int r600_uvd_init(struct radeon_device *rdev, bool ring_test); void r600_uvd_stop(struct radeon_device *rdev); int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index fb5ea6208970..cb4b931d8d9f 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -363,7 +363,7 @@ u32 radeon_ring_generic_get_rptr(struct radeon_device *rdev, { u32 rptr; - if (rdev->wb.enabled && ring != &rdev->ring[R600_RING_TYPE_UVD_INDEX]) + if (rdev->wb.enabled) rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]); else rptr = RREG32(ring->rptr_reg); diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index 52253b2ab0d5..1e8cf49d5871 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1923,8 +1923,7 @@ static int rv770_startup(struct radeon_device *rdev) ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; if (ring->ring_size) { - r = radeon_ring_init(rdev, ring, ring->ring_size, - R600_WB_UVD_RPTR_OFFSET, + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index da23ce8f4388..4ff59c8f508f 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6416,8 +6416,7 @@ static int si_startup(struct radeon_device *rdev) if (rdev->has_uvd) { ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; if (ring->ring_size) { - r = radeon_ring_init(rdev, ring, ring->ring_size, - R600_WB_UVD_RPTR_OFFSET, + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, 0, 0xfffff, RADEON_CP_PACKET2); if (!r) -- cgit v1.2.3 From 2e1e6dad6a6d437e4c40611fdcc4e6cd9e2f969e Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 13 Aug 2013 11:56:52 +0200 Subject: drm/radeon: remove special handling for the DMA ring MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Now that we have callbacks for [rw]ptr handling we can remove the special handling for the DMA rings and use the callbacks instead. Signed-off-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 20 ++++++--------- drivers/gpu/drm/radeon/evergreen.c | 6 ++--- drivers/gpu/drm/radeon/ni.c | 8 +++--- drivers/gpu/drm/radeon/r100.c | 2 +- drivers/gpu/drm/radeon/r600.c | 47 ++++++++++++++++++++++++++++++++++-- drivers/gpu/drm/radeon/radeon.h | 5 +--- drivers/gpu/drm/radeon/radeon_asic.c | 30 +++++++++++------------ drivers/gpu/drm/radeon/radeon_asic.h | 7 ++++++ drivers/gpu/drm/radeon/radeon_ring.c | 11 ++------- drivers/gpu/drm/radeon/rv770.c | 6 ++--- drivers/gpu/drm/radeon/si.c | 12 ++++----- 11 files changed, 95 insertions(+), 59 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index ce7036ae9f5a..34be795de173 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -3414,7 +3414,6 @@ u32 cik_compute_ring_get_rptr(struct radeon_device *rdev, cik_srbm_select(rdev, 0, 0, 0, 0); mutex_unlock(&rdev->srbm_mutex); } - rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; return rptr; } @@ -3433,7 +3432,6 @@ u32 cik_compute_ring_get_wptr(struct radeon_device *rdev, cik_srbm_select(rdev, 0, 0, 0, 0); mutex_unlock(&rdev->srbm_mutex); } - wptr = (wptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; return wptr; } @@ -3441,10 +3439,8 @@ u32 cik_compute_ring_get_wptr(struct radeon_device *rdev, void cik_compute_ring_set_wptr(struct radeon_device *rdev, struct radeon_ring *ring) { - u32 wptr = (ring->wptr << ring->ptr_reg_shift) & ring->ptr_reg_mask; - - rdev->wb.wb[ring->wptr_offs/4] = cpu_to_le32(wptr); - WDOORBELL32(ring->doorbell_offset, wptr); + rdev->wb.wb[ring->wptr_offs/4] = cpu_to_le32(ring->wptr); + WDOORBELL32(ring->doorbell_offset, ring->wptr); } /** @@ -7649,7 +7645,7 @@ static int cik_startup(struct radeon_device *rdev) ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, CP_RB0_RPTR, CP_RB0_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; @@ -7658,7 +7654,7 @@ static int cik_startup(struct radeon_device *rdev) ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP1_RPTR_OFFSET, CP_HQD_PQ_RPTR, CP_HQD_PQ_WPTR, - 0, 0xfffff, PACKET3(PACKET3_NOP, 0x3FFF)); + PACKET3(PACKET3_NOP, 0x3FFF)); if (r) return r; ring->me = 1; /* first MEC */ @@ -7670,7 +7666,7 @@ static int cik_startup(struct radeon_device *rdev) ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP2_RPTR_OFFSET, CP_HQD_PQ_RPTR, CP_HQD_PQ_WPTR, - 0, 0xffffffff, PACKET3(PACKET3_NOP, 0x3FFF)); + PACKET3(PACKET3_NOP, 0x3FFF)); if (r) return r; /* dGPU only have 1 MEC */ @@ -7683,7 +7679,7 @@ static int cik_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, SDMA0_GFX_RB_RPTR + SDMA0_REGISTER_OFFSET, SDMA0_GFX_RB_WPTR + SDMA0_REGISTER_OFFSET, - 2, 0xfffffffc, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); if (r) return r; @@ -7691,7 +7687,7 @@ static int cik_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET, SDMA0_GFX_RB_RPTR + SDMA1_REGISTER_OFFSET, SDMA0_GFX_RB_WPTR + SDMA1_REGISTER_OFFSET, - 2, 0xfffffffc, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); if (r) return r; @@ -7707,7 +7703,7 @@ static int cik_startup(struct radeon_device *rdev) if (ring->ring_size) { r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (!r) r = r600_uvd_init(rdev, true); if (r) diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 2139f6c64341..389f5a981358 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5268,14 +5268,14 @@ static int evergreen_startup(struct radeon_device *rdev) ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, R600_CP_RB_RPTR, R600_CP_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, DMA_RB_RPTR, DMA_RB_WPTR, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0)); if (r) return r; @@ -5293,7 +5293,7 @@ static int evergreen_startup(struct radeon_device *rdev) if (ring->ring_size) { r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (!r) r = r600_uvd_init(rdev, true); diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index f543f4ca4dda..e04b17338336 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -2192,7 +2192,7 @@ static int cayman_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, CP_RB0_RPTR, CP_RB0_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; @@ -2200,7 +2200,7 @@ static int cayman_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, DMA_RB_RPTR + DMA0_REGISTER_OFFSET, DMA_RB_WPTR + DMA0_REGISTER_OFFSET, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); if (r) return r; @@ -2208,7 +2208,7 @@ static int cayman_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET, DMA_RB_RPTR + DMA1_REGISTER_OFFSET, DMA_RB_WPTR + DMA1_REGISTER_OFFSET, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); if (r) return r; @@ -2227,7 +2227,7 @@ static int cayman_startup(struct radeon_device *rdev) if (ring->ring_size) { r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (!r) r = r600_uvd_init(rdev, true); if (r) diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c index 75349cdaa84b..2cbc512645d4 100644 --- a/drivers/gpu/drm/radeon/r100.c +++ b/drivers/gpu/drm/radeon/r100.c @@ -1102,7 +1102,7 @@ int r100_cp_init(struct radeon_device *rdev, unsigned ring_size) r100_cp_load_microcode(rdev); r = radeon_ring_init(rdev, ring, ring_size, RADEON_WB_CP_RPTR_OFFSET, RADEON_CP_RB_RPTR, RADEON_CP_RB_WPTR, - 0, 0x7fffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) { return r; } diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index c1b0aba4431a..30849eca6e07 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2504,6 +2504,49 @@ void r600_cp_fini(struct radeon_device *rdev) * solid fills, and a number of other things. It also * has support for tiling/detiling of buffers. */ + +/** + * r600_dma_get_rptr - get the current read pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Get the current rptr from the hardware (r6xx+). + */ +uint32_t r600_dma_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return (radeon_ring_generic_get_rptr(rdev, ring) & 0x3fffc) >> 2; +} + +/** + * r600_dma_get_wptr - get the current write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Get the current wptr from the hardware (r6xx+). + */ +uint32_t r600_dma_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return (RREG32(ring->wptr_reg) & 0x3fffc) >> 2; +} + +/** + * r600_dma_set_wptr - commit the write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Write the wptr back to the hardware (r6xx+). + */ +void r600_dma_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + WREG32(ring->wptr_reg, (ring->wptr << 2) & 0x3fffc); +} + /** * r600_dma_stop - stop the async dma engine * @@ -3386,14 +3429,14 @@ static int r600_startup(struct radeon_device *rdev) ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, R600_CP_RB_RPTR, R600_CP_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, DMA_RB_RPTR, DMA_RB_WPTR, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); if (r) return r; diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 2eab174bf22e..791cc8de6395 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -760,8 +760,6 @@ struct radeon_ring { uint32_t align_mask; uint32_t ptr_mask; bool ready; - u32 ptr_reg_shift; - u32 ptr_reg_mask; u32 nop; u32 idx; u64 last_semaphore_signal_addr; @@ -912,8 +910,7 @@ unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring int radeon_ring_restore(struct radeon_device *rdev, struct radeon_ring *ring, unsigned size, uint32_t *data); int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ring_size, - unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, - u32 ptr_reg_shift, u32 ptr_reg_mask, u32 nop); + unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, u32 nop); void radeon_ring_fini(struct radeon_device *rdev, struct radeon_ring *cp); diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index 7432247a812a..785b7a7add77 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -896,9 +896,9 @@ static struct radeon_asic_ring r600_dma_ring = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &r600_dma_is_lockup, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_dma_get_rptr, + .get_wptr = &r600_dma_get_wptr, + .set_wptr = &r600_dma_set_wptr, }; static struct radeon_asic r600_asic = { @@ -1275,9 +1275,9 @@ static struct radeon_asic_ring evergreen_dma_ring = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &evergreen_dma_is_lockup, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_dma_get_rptr, + .get_wptr = &r600_dma_get_wptr, + .set_wptr = &r600_dma_set_wptr, }; static struct radeon_asic evergreen_asic = { @@ -1580,9 +1580,9 @@ static struct radeon_asic_ring cayman_dma_ring = { .ib_test = &r600_dma_ib_test, .is_lockup = &cayman_dma_is_lockup, .vm_flush = &cayman_dma_vm_flush, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr + .get_rptr = &r600_dma_get_rptr, + .get_wptr = &r600_dma_get_wptr, + .set_wptr = &r600_dma_set_wptr }; static struct radeon_asic_ring cayman_uvd_ring = { @@ -1822,9 +1822,9 @@ static struct radeon_asic_ring si_dma_ring = { .ib_test = &r600_dma_ib_test, .is_lockup = &si_dma_is_lockup, .vm_flush = &si_dma_vm_flush, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_dma_get_rptr, + .get_wptr = &r600_dma_get_wptr, + .set_wptr = &r600_dma_set_wptr, }; static struct radeon_asic si_asic = { @@ -1966,9 +1966,9 @@ static struct radeon_asic_ring ci_dma_ring = { .ib_test = &cik_sdma_ib_test, .is_lockup = &cik_sdma_is_lockup, .vm_flush = &cik_dma_vm_flush, - .get_rptr = &radeon_ring_generic_get_rptr, - .get_wptr = &radeon_ring_generic_get_wptr, - .set_wptr = &radeon_ring_generic_set_wptr, + .get_rptr = &r600_dma_get_rptr, + .get_wptr = &r600_dma_get_wptr, + .set_wptr = &r600_dma_set_wptr, }; static struct radeon_asic ci_asic = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 37baf9c696f0..5c53eb78b22d 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -392,6 +392,13 @@ uint64_t r600_get_gpu_clock_counter(struct radeon_device *rdev); int rv6xx_get_temp(struct radeon_device *rdev); int r600_dpm_pre_set_power_state(struct radeon_device *rdev); void r600_dpm_post_set_power_state(struct radeon_device *rdev); +/* r600 dma */ +uint32_t r600_dma_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +uint32_t r600_dma_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void r600_dma_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); /* rv6xx dpm */ int rv6xx_dpm_init(struct radeon_device *rdev); int rv6xx_dpm_enable(struct radeon_device *rdev); diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index cb4b931d8d9f..46a25f037b84 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -367,7 +367,6 @@ u32 radeon_ring_generic_get_rptr(struct radeon_device *rdev, rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]); else rptr = RREG32(ring->rptr_reg); - rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; return rptr; } @@ -378,7 +377,6 @@ u32 radeon_ring_generic_get_wptr(struct radeon_device *rdev, u32 wptr; wptr = RREG32(ring->wptr_reg); - wptr = (wptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; return wptr; } @@ -386,7 +384,7 @@ u32 radeon_ring_generic_get_wptr(struct radeon_device *rdev, void radeon_ring_generic_set_wptr(struct radeon_device *rdev, struct radeon_ring *ring) { - WREG32(ring->wptr_reg, (ring->wptr << ring->ptr_reg_shift) & ring->ptr_reg_mask); + WREG32(ring->wptr_reg, ring->wptr); (void)RREG32(ring->wptr_reg); } @@ -719,16 +717,13 @@ int radeon_ring_restore(struct radeon_device *rdev, struct radeon_ring *ring, * @rptr_offs: offset of the rptr writeback location in the WB buffer * @rptr_reg: MMIO offset of the rptr register * @wptr_reg: MMIO offset of the wptr register - * @ptr_reg_shift: bit offset of the rptr/wptr values - * @ptr_reg_mask: bit mask of the rptr/wptr values * @nop: nop packet for this ring * * Initialize the driver information for the selected ring (all asics). * Returns 0 on success, error on failure. */ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsigned ring_size, - unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, - u32 ptr_reg_shift, u32 ptr_reg_mask, u32 nop) + unsigned rptr_offs, unsigned rptr_reg, unsigned wptr_reg, u32 nop) { int r; @@ -736,8 +731,6 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig ring->rptr_offs = rptr_offs; ring->rptr_reg = rptr_reg; ring->wptr_reg = wptr_reg; - ring->ptr_reg_shift = ptr_reg_shift; - ring->ptr_reg_mask = ptr_reg_mask; ring->nop = nop; /* Allocate ring buffer */ if (ring->ring_obj == NULL) { diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index 1e8cf49d5871..fd9dcb2d182b 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1899,14 +1899,14 @@ static int rv770_startup(struct radeon_device *rdev) ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, R600_CP_RB_RPTR, R600_CP_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, DMA_RB_RPTR, DMA_RB_WPTR, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); if (r) return r; @@ -1925,7 +1925,7 @@ static int rv770_startup(struct radeon_device *rdev) if (ring->ring_size) { r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (!r) r = r600_uvd_init(rdev, true); diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 4ff59c8f508f..ae232be62921 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6368,21 +6368,21 @@ static int si_startup(struct radeon_device *rdev) ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, CP_RB0_RPTR, CP_RB0_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP1_RPTR_OFFSET, CP_RB1_RPTR, CP_RB1_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP2_RPTR_OFFSET, CP_RB2_RPTR, CP_RB2_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (r) return r; @@ -6390,7 +6390,7 @@ static int si_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, DMA_RB_RPTR + DMA0_REGISTER_OFFSET, DMA_RB_WPTR + DMA0_REGISTER_OFFSET, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0)); if (r) return r; @@ -6398,7 +6398,7 @@ static int si_startup(struct radeon_device *rdev) r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET, DMA_RB_RPTR + DMA1_REGISTER_OFFSET, DMA_RB_WPTR + DMA1_REGISTER_OFFSET, - 2, 0x3fffc, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0)); + DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0)); if (r) return r; @@ -6418,7 +6418,7 @@ static int si_startup(struct radeon_device *rdev) if (ring->ring_size) { r = radeon_ring_init(rdev, ring, ring->ring_size, 0, UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, - 0, 0xfffff, RADEON_CP_PACKET2); + RADEON_CP_PACKET2); if (!r) r = r600_uvd_init(rdev, true); if (r) -- cgit v1.2.3 From e409b128625732926c112cc9b709fb7bb1aa387f Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 13 Aug 2013 11:56:53 +0200 Subject: drm/radeon: separate UVD code v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Our different hardware blocks are actually completely separated, so it doesn't make much sense any more to structure the code by pure chipset generations. Start restructuring the code by separating our the UVD block. v2: updated commit message v3: rebased and restructurized start/stop functions for kv dpm. Signed-off-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/Makefile | 8 + drivers/gpu/drm/radeon/cik.c | 41 +--- drivers/gpu/drm/radeon/evergreen.c | 8 +- drivers/gpu/drm/radeon/kv_dpm.c | 11 +- drivers/gpu/drm/radeon/ni.c | 25 +- drivers/gpu/drm/radeon/r600.c | 345 ---------------------------- drivers/gpu/drm/radeon/radeon_asic.c | 32 +-- drivers/gpu/drm/radeon/radeon_asic.h | 55 +++-- drivers/gpu/drm/radeon/rv770.c | 105 +-------- drivers/gpu/drm/radeon/rv770d.h | 16 ++ drivers/gpu/drm/radeon/si.c | 8 +- drivers/gpu/drm/radeon/uvd_v1_0.c | 434 +++++++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/uvd_v2_2.c | 165 +++++++++++++ drivers/gpu/drm/radeon/uvd_v3_1.c | 55 +++++ drivers/gpu/drm/radeon/uvd_v4_2.c | 73 ++++++ 15 files changed, 826 insertions(+), 555 deletions(-) create mode 100644 drivers/gpu/drm/radeon/uvd_v1_0.c create mode 100644 drivers/gpu/drm/radeon/uvd_v2_2.c create mode 100644 drivers/gpu/drm/radeon/uvd_v3_1.c create mode 100644 drivers/gpu/drm/radeon/uvd_v4_2.c (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index ea913cc681b4..1e23b18d549a 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -82,6 +82,14 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \ trinity_smc.o ni_dpm.o si_smc.o si_dpm.o kv_smc.o kv_dpm.o ci_smc.o \ ci_dpm.o +# add UVD block +radeon-y += \ + radeon_uvd.o \ + uvd_v1_0.o \ + uvd_v2_2.o \ + uvd_v3_1.o \ + uvd_v4_2.o + radeon-$(CONFIG_COMPAT) += radeon_ioc32.o radeon-$(CONFIG_VGA_SWITCHEROO) += radeon_atpx_handler.o radeon-$(CONFIG_ACPI) += radeon_acpi.o diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 34be795de173..1400b5203db1 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -69,7 +69,6 @@ static void cik_pcie_gen3_enable(struct radeon_device *rdev); static void cik_program_aspm(struct radeon_device *rdev); static void cik_init_pg(struct radeon_device *rdev); static void cik_init_cg(struct radeon_device *rdev); -void cik_uvd_resume(struct radeon_device *rdev); /* get temperature in millidegrees */ int ci_get_temp(struct radeon_device *rdev) @@ -7616,9 +7615,8 @@ static int cik_startup(struct radeon_device *rdev) return r; } - r = radeon_uvd_resume(rdev); + r = uvd_v4_2_resume(rdev); if (!r) { - cik_uvd_resume(rdev); r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); if (r) @@ -7705,7 +7703,7 @@ static int cik_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev, true); + r = uvd_v1_0_init(rdev); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } @@ -7770,7 +7768,7 @@ int cik_suspend(struct radeon_device *rdev) radeon_vm_manager_fini(rdev); cik_cp_enable(rdev, false); cik_sdma_enable(rdev, false); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); cik_irq_suspend(rdev); radeon_wb_disable(rdev); @@ -7934,7 +7932,7 @@ void cik_fini(struct radeon_device *rdev) radeon_vm_manager_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); cik_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); @@ -8595,37 +8593,6 @@ int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) return r; } -void cik_uvd_resume(struct radeon_device *rdev) -{ - uint64_t addr; - uint32_t size; - - /* programm the VCPU memory controller bits 0-27 */ - addr = rdev->uvd.gpu_addr >> 3; - size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 4) >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET0, addr); - WREG32(UVD_VCPU_CACHE_SIZE0, size); - - addr += size; - size = RADEON_UVD_STACK_SIZE >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET1, addr); - WREG32(UVD_VCPU_CACHE_SIZE1, size); - - addr += size; - size = RADEON_UVD_HEAP_SIZE >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET2, addr); - WREG32(UVD_VCPU_CACHE_SIZE2, size); - - /* bits 28-31 */ - addr = (rdev->uvd.gpu_addr >> 28) & 0xF; - WREG32(UVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); - - /* bits 32-39 */ - addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; - WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); - -} - static void cik_pcie_gen3_enable(struct radeon_device *rdev) { struct pci_dev *root = rdev->pdev->bus->self; diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 389f5a981358..52ed22333f0d 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5239,7 +5239,7 @@ static int evergreen_startup(struct radeon_device *rdev) return r; } - r = rv770_uvd_resume(rdev); + r = uvd_v2_2_resume(rdev); if (!r) { r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); @@ -5295,7 +5295,7 @@ static int evergreen_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev, true); + r = uvd_v1_0_init(rdev); if (r) DRM_ERROR("radeon: error initializing UVD (%d).\n", r); @@ -5350,7 +5350,7 @@ int evergreen_resume(struct radeon_device *rdev) int evergreen_suspend(struct radeon_device *rdev) { r600_audio_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); r700_cp_stop(rdev); r600_dma_stop(rdev); @@ -5487,7 +5487,7 @@ void evergreen_fini(struct radeon_device *rdev) radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); evergreen_pcie_gart_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); diff --git a/drivers/gpu/drm/radeon/kv_dpm.c b/drivers/gpu/drm/radeon/kv_dpm.c index 3e232a4d3f4c..ef6c901690da 100644 --- a/drivers/gpu/drm/radeon/kv_dpm.c +++ b/drivers/gpu/drm/radeon/kv_dpm.c @@ -26,6 +26,7 @@ #include "cikd.h" #include "r600_dpm.h" #include "kv_dpm.h" +#include "radeon_asic.h" #include #define KV_MAX_DEEPSLEEP_DIVIDER_ID 5 @@ -59,10 +60,6 @@ extern void cik_exit_rlc_safe_mode(struct radeon_device *rdev); extern void cik_update_cg(struct radeon_device *rdev, u32 block, bool enable); -extern void cik_uvd_resume(struct radeon_device *rdev); -extern int r600_uvd_init(struct radeon_device *rdev, bool ring_test); -extern void r600_do_uvd_stop(struct radeon_device *rdev); - static const struct kv_lcac_config_values sx_local_cac_cfg_kv[] = { { 0, 4, 1 }, @@ -1473,7 +1470,7 @@ void kv_dpm_powergate_uvd(struct radeon_device *rdev, bool gate) pi->uvd_power_gated = gate; if (gate) { - r600_do_uvd_stop(rdev); + uvd_v1_0_stop(rdev); cik_update_cg(rdev, RADEON_CG_BLOCK_UVD, false); kv_update_uvd_dpm(rdev, gate); if (pi->caps_uvd_pg) @@ -1481,8 +1478,8 @@ void kv_dpm_powergate_uvd(struct radeon_device *rdev, bool gate) } else { if (pi->caps_uvd_pg) kv_notify_message_to_smu(rdev, PPSMC_MSG_UVDPowerON); - cik_uvd_resume(rdev); - r600_uvd_init(rdev, false); + uvd_v4_2_resume(rdev); + uvd_v1_0_start(rdev); cik_update_cg(rdev, RADEON_CG_BLOCK_UVD, true); kv_update_uvd_dpm(rdev, gate); } diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index e04b17338336..0205fa1594fa 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1373,23 +1373,6 @@ void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) radeon_ring_write(ring, 10); /* poll interval */ } -void cayman_uvd_semaphore_emit(struct radeon_device *rdev, - struct radeon_ring *ring, - struct radeon_semaphore *semaphore, - bool emit_wait) -{ - uint64_t addr = semaphore->gpu_addr; - - radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_LOW, 0)); - radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); - - radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_HIGH, 0)); - radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); - - radeon_ring_write(ring, PACKET0(UVD_SEMA_CMD, 0)); - radeon_ring_write(ring, 0x80 | (emit_wait ? 1 : 0)); -} - static void cayman_cp_enable(struct radeon_device *rdev, bool enable) { if (enable) @@ -2141,7 +2124,7 @@ static int cayman_startup(struct radeon_device *rdev) return r; } - r = rv770_uvd_resume(rdev); + r = uvd_v2_2_resume(rdev); if (!r) { r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); @@ -2229,7 +2212,7 @@ static int cayman_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev, true); + r = uvd_v1_0_init(rdev); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } @@ -2283,7 +2266,7 @@ int cayman_suspend(struct radeon_device *rdev) radeon_vm_manager_fini(rdev); cayman_cp_enable(rdev, false); cayman_dma_stop(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); evergreen_irq_suspend(rdev); radeon_wb_disable(rdev); @@ -2414,7 +2397,7 @@ void cayman_fini(struct radeon_device *rdev) radeon_vm_manager_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); cayman_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 30849eca6e07..3a08ef92d33f 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2663,231 +2663,6 @@ void r600_dma_fini(struct radeon_device *rdev) radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); } -/* - * UVD - */ -uint32_t r600_uvd_get_rptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - return RREG32(UVD_RBC_RB_RPTR); -} - -uint32_t r600_uvd_get_wptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - return RREG32(UVD_RBC_RB_WPTR); -} - -void r600_uvd_set_wptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - WREG32(UVD_RBC_RB_WPTR, ring->wptr); -} - -static int r600_uvd_rbc_start(struct radeon_device *rdev, bool ring_test) -{ - struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - uint32_t rb_bufsz, tmp; - int r; - - /* force RBC into idle state */ - WREG32(UVD_RBC_RB_CNTL, 0x11010101); - - /* Set the write pointer delay */ - WREG32(UVD_RBC_RB_WPTR_CNTL, 0); - - /* programm the 4GB memory segment for rptr and ring buffer */ - WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) | - (0x7 << 16) | (0x1 << 31)); - - /* Initialize the ring buffer's read and write pointers */ - WREG32(UVD_RBC_RB_RPTR, 0x0); - - ring->wptr = ring->rptr = RREG32(UVD_RBC_RB_RPTR); - WREG32(UVD_RBC_RB_WPTR, ring->wptr); - - /* set the ring address */ - WREG32(UVD_RBC_RB_BASE, ring->gpu_addr); - - /* Set ring buffer size */ - rb_bufsz = drm_order(ring->ring_size); - rb_bufsz = (0x1 << 8) | rb_bufsz; - WREG32_P(UVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f); - - if (ring_test) { - ring->ready = true; - r = radeon_ring_test(rdev, R600_RING_TYPE_UVD_INDEX, ring); - if (r) { - ring->ready = false; - return r; - } - - r = radeon_ring_lock(rdev, ring, 10); - if (r) { - DRM_ERROR("radeon: ring failed to lock UVD ring (%d).\n", r); - return r; - } - - tmp = PACKET0(UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); - - tmp = PACKET0(UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); - - tmp = PACKET0(UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0); - radeon_ring_write(ring, tmp); - radeon_ring_write(ring, 0xFFFFF); - - /* Clear timeout status bits */ - radeon_ring_write(ring, PACKET0(UVD_SEMA_TIMEOUT_STATUS, 0)); - radeon_ring_write(ring, 0x8); - - radeon_ring_write(ring, PACKET0(UVD_SEMA_CNTL, 0)); - radeon_ring_write(ring, 3); - - radeon_ring_unlock_commit(rdev, ring); - } - - return 0; -} - -void r600_do_uvd_stop(struct radeon_device *rdev) -{ - /* force RBC into idle state */ - WREG32(UVD_RBC_RB_CNTL, 0x11010101); - - /* Stall UMC and register bus before resetting VCPU */ - WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); - WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); - mdelay(1); - - /* put VCPU into reset */ - WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); - mdelay(5); - - /* disable VCPU clock */ - WREG32(UVD_VCPU_CNTL, 0x0); - - /* Unstall UMC and register bus */ - WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); - WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); -} - -void r600_uvd_stop(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; - - r600_do_uvd_stop(rdev); - ring->ready = false; -} - -int r600_uvd_init(struct radeon_device *rdev, bool ring_test) -{ - int i, j, r; - /* disable byte swapping */ - u32 lmi_swap_cntl = 0; - u32 mp_swap_cntl = 0; - - /* raise clocks while booting up the VCPU */ - radeon_set_uvd_clocks(rdev, 53300, 40000); - - /* disable clock gating */ - WREG32(UVD_CGC_GATE, 0); - - /* disable interupt */ - WREG32_P(UVD_MASTINT_EN, 0, ~(1 << 1)); - - /* Stall UMC and register bus before resetting VCPU */ - WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); - WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); - mdelay(1); - - /* put LMI, VCPU, RBC etc... into reset */ - WREG32(UVD_SOFT_RESET, LMI_SOFT_RESET | VCPU_SOFT_RESET | - LBSI_SOFT_RESET | RBC_SOFT_RESET | CSM_SOFT_RESET | - CXW_SOFT_RESET | TAP_SOFT_RESET | LMI_UMC_SOFT_RESET); - mdelay(5); - - /* take UVD block out of reset */ - WREG32_P(SRBM_SOFT_RESET, 0, ~SOFT_RESET_UVD); - mdelay(5); - - /* initialize UVD memory controller */ - WREG32(UVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) | - (1 << 21) | (1 << 9) | (1 << 20)); - -#ifdef __BIG_ENDIAN - /* swap (8 in 32) RB and IB */ - lmi_swap_cntl = 0xa; - mp_swap_cntl = 0; -#endif - WREG32(UVD_LMI_SWAP_CNTL, lmi_swap_cntl); - WREG32(UVD_MP_SWAP_CNTL, mp_swap_cntl); - - WREG32(UVD_MPC_SET_MUXA0, 0x40c2040); - WREG32(UVD_MPC_SET_MUXA1, 0x0); - WREG32(UVD_MPC_SET_MUXB0, 0x40c2040); - WREG32(UVD_MPC_SET_MUXB1, 0x0); - WREG32(UVD_MPC_SET_ALU, 0); - WREG32(UVD_MPC_SET_MUX, 0x88); - - /* take all subblocks out of reset, except VCPU */ - WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); - mdelay(5); - - /* enable VCPU clock */ - WREG32(UVD_VCPU_CNTL, 1 << 9); - - /* enable UMC */ - WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); - - /* boot up the VCPU */ - WREG32(UVD_SOFT_RESET, 0); - mdelay(10); - - WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); - - for (i = 0; i < 10; ++i) { - uint32_t status; - for (j = 0; j < 100; ++j) { - status = RREG32(UVD_STATUS); - if (status & 2) - break; - mdelay(10); - } - r = 0; - if (status & 2) - break; - - DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); - WREG32_P(UVD_SOFT_RESET, VCPU_SOFT_RESET, ~VCPU_SOFT_RESET); - mdelay(10); - WREG32_P(UVD_SOFT_RESET, 0, ~VCPU_SOFT_RESET); - mdelay(10); - r = -1; - } - - if (r) { - DRM_ERROR("UVD not responding, giving up!!!\n"); - goto done; - } - - /* enable interupt */ - WREG32_P(UVD_MASTINT_EN, 3<<1, ~(3 << 1)); - - r = r600_uvd_rbc_start(rdev, ring_test); - if (!r) - DRM_INFO("UVD initialized successfully.\n"); - -done: - /* lower clocks again */ - radeon_set_uvd_clocks(rdev, 0, 0); - - return r; -} - /* * GPU scratch registers helpers function. */ @@ -2997,40 +2772,6 @@ int r600_dma_ring_test(struct radeon_device *rdev, return r; } -int r600_uvd_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) -{ - uint32_t tmp = 0; - unsigned i; - int r; - - WREG32(UVD_CONTEXT_ID, 0xCAFEDEAD); - r = radeon_ring_lock(rdev, ring, 3); - if (r) { - DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", - ring->idx, r); - return r; - } - radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); - radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); - for (i = 0; i < rdev->usec_timeout; i++) { - tmp = RREG32(UVD_CONTEXT_ID); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - - if (i < rdev->usec_timeout) { - DRM_INFO("ring test on %d succeeded in %d usecs\n", - ring->idx, i); - } else { - DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - return r; -} - /* * CP fences/semaphores */ @@ -3082,30 +2823,6 @@ void r600_fence_ring_emit(struct radeon_device *rdev, } } -void r600_uvd_fence_emit(struct radeon_device *rdev, - struct radeon_fence *fence) -{ - struct radeon_ring *ring = &rdev->ring[fence->ring]; - uint64_t addr = rdev->fence_drv[fence->ring].gpu_addr; - - radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); - radeon_ring_write(ring, fence->seq); - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA0, 0)); - radeon_ring_write(ring, addr & 0xffffffff); - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA1, 0)); - radeon_ring_write(ring, upper_32_bits(addr) & 0xff); - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_CMD, 0)); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA0, 0)); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA1, 0)); - radeon_ring_write(ring, 0); - radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_CMD, 0)); - radeon_ring_write(ring, 2); - return; -} - void r600_semaphore_ring_emit(struct radeon_device *rdev, struct radeon_ring *ring, struct radeon_semaphore *semaphore, @@ -3175,23 +2892,6 @@ void r600_dma_semaphore_ring_emit(struct radeon_device *rdev, radeon_ring_write(ring, upper_32_bits(addr) & 0xff); } -void r600_uvd_semaphore_emit(struct radeon_device *rdev, - struct radeon_ring *ring, - struct radeon_semaphore *semaphore, - bool emit_wait) -{ - uint64_t addr = semaphore->gpu_addr; - - radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_LOW, 0)); - radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); - - radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_HIGH, 0)); - radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); - - radeon_ring_write(ring, PACKET0(UVD_SEMA_CMD, 0)); - radeon_ring_write(ring, emit_wait ? 1 : 0); -} - /** * r600_copy_cpdma - copy pages using the CP DMA engine * @@ -3656,16 +3356,6 @@ void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) radeon_ring_write(ring, ib->length_dw); } -void r600_uvd_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) -{ - struct radeon_ring *ring = &rdev->ring[ib->ring]; - - radeon_ring_write(ring, PACKET0(UVD_RBC_IB_BASE, 0)); - radeon_ring_write(ring, ib->gpu_addr); - radeon_ring_write(ring, PACKET0(UVD_RBC_IB_SIZE, 0)); - radeon_ring_write(ring, ib->length_dw); -} - int r600_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) { struct radeon_ib ib; @@ -3783,41 +3473,6 @@ int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) return r; } -int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) -{ - struct radeon_fence *fence = NULL; - int r; - - r = radeon_set_uvd_clocks(rdev, 53300, 40000); - if (r) { - DRM_ERROR("radeon: failed to raise UVD clocks (%d).\n", r); - return r; - } - - r = radeon_uvd_get_create_msg(rdev, ring->idx, 1, NULL); - if (r) { - DRM_ERROR("radeon: failed to get create msg (%d).\n", r); - goto error; - } - - r = radeon_uvd_get_destroy_msg(rdev, ring->idx, 1, &fence); - if (r) { - DRM_ERROR("radeon: failed to get destroy ib (%d).\n", r); - goto error; - } - - r = radeon_fence_wait(fence, false); - if (r) { - DRM_ERROR("radeon: fence wait failed (%d).\n", r); - goto error; - } - DRM_INFO("ib test on ring %d succeeded\n", ring->idx); -error: - radeon_fence_unref(&fence); - radeon_set_uvd_clocks(rdev, 0, 0); - return r; -} - /** * r600_dma_ring_ib_execute - Schedule an IB on the DMA engine * diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index 785b7a7add77..da755bf37421 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -1150,16 +1150,16 @@ static struct radeon_asic rs780_asic = { }; static struct radeon_asic_ring rv770_uvd_ring = { - .ib_execute = &r600_uvd_ib_execute, - .emit_fence = &r600_uvd_fence_emit, - .emit_semaphore = &r600_uvd_semaphore_emit, + .ib_execute = &uvd_v1_0_ib_execute, + .emit_fence = &uvd_v2_2_fence_emit, + .emit_semaphore = &uvd_v1_0_semaphore_emit, .cs_parse = &radeon_uvd_cs_parse, - .ring_test = &r600_uvd_ring_test, - .ib_test = &r600_uvd_ib_test, + .ring_test = &uvd_v1_0_ring_test, + .ib_test = &uvd_v1_0_ib_test, .is_lockup = &radeon_ring_test_lockup, - .get_rptr = &r600_uvd_get_rptr, - .get_wptr = &r600_uvd_get_wptr, - .set_wptr = &r600_uvd_set_wptr, + .get_rptr = &uvd_v1_0_get_rptr, + .get_wptr = &uvd_v1_0_get_wptr, + .set_wptr = &uvd_v1_0_set_wptr, }; static struct radeon_asic rv770_asic = { @@ -1586,16 +1586,16 @@ static struct radeon_asic_ring cayman_dma_ring = { }; static struct radeon_asic_ring cayman_uvd_ring = { - .ib_execute = &r600_uvd_ib_execute, - .emit_fence = &r600_uvd_fence_emit, - .emit_semaphore = &cayman_uvd_semaphore_emit, + .ib_execute = &uvd_v1_0_ib_execute, + .emit_fence = &uvd_v2_2_fence_emit, + .emit_semaphore = &uvd_v3_1_semaphore_emit, .cs_parse = &radeon_uvd_cs_parse, - .ring_test = &r600_uvd_ring_test, - .ib_test = &r600_uvd_ib_test, + .ring_test = &uvd_v1_0_ring_test, + .ib_test = &uvd_v1_0_ib_test, .is_lockup = &radeon_ring_test_lockup, - .get_rptr = &r600_uvd_get_rptr, - .get_wptr = &r600_uvd_get_wptr, - .set_wptr = &r600_uvd_set_wptr, + .get_rptr = &uvd_v1_0_get_rptr, + .get_wptr = &uvd_v1_0_get_wptr, + .set_wptr = &uvd_v1_0_set_wptr, }; static struct radeon_asic cayman_asic = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 5c53eb78b22d..e69f00a7f153 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -336,7 +336,6 @@ int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); int r600_dma_ring_test(struct radeon_device *rdev, struct radeon_ring *cp); -int r600_uvd_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); int r600_copy_cpdma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, struct radeon_fence **fence); @@ -430,24 +429,6 @@ void rs780_dpm_print_power_state(struct radeon_device *rdev, void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, struct seq_file *m); -/* uvd */ -uint32_t r600_uvd_get_rptr(struct radeon_device *rdev, - struct radeon_ring *ring); -uint32_t r600_uvd_get_wptr(struct radeon_device *rdev, - struct radeon_ring *ring); -void r600_uvd_set_wptr(struct radeon_device *rdev, - struct radeon_ring *ring); -int r600_uvd_init(struct radeon_device *rdev, bool ring_test); -void r600_uvd_stop(struct radeon_device *rdev); -int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); -void r600_uvd_fence_emit(struct radeon_device *rdev, - struct radeon_fence *fence); -void r600_uvd_semaphore_emit(struct radeon_device *rdev, - struct radeon_ring *ring, - struct radeon_semaphore *semaphore, - bool emit_wait); -void r600_uvd_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); - /* * rv770,rv730,rv710,rv740 */ @@ -465,7 +446,6 @@ int rv770_copy_dma(struct radeon_device *rdev, unsigned num_gpu_pages, struct radeon_fence **fence); u32 rv770_get_xclk(struct radeon_device *rdev); -int rv770_uvd_resume(struct radeon_device *rdev); int rv770_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); int rv770_get_temp(struct radeon_device *rdev); /* rv7xx pm */ @@ -800,4 +780,39 @@ int kv_dpm_force_performance_level(struct radeon_device *rdev, enum radeon_dpm_forced_level level); void kv_dpm_powergate_uvd(struct radeon_device *rdev, bool gate); +/* uvd v1.0 */ +uint32_t uvd_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +uint32_t uvd_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void uvd_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); + +int uvd_v1_0_init(struct radeon_device *rdev); +void uvd_v1_0_fini(struct radeon_device *rdev); +int uvd_v1_0_start(struct radeon_device *rdev); +void uvd_v1_0_stop(struct radeon_device *rdev); + +int uvd_v1_0_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); +int uvd_v1_0_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); +void uvd_v1_0_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait); +void uvd_v1_0_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); + +/* uvd v2.2 */ +int uvd_v2_2_resume(struct radeon_device *rdev); +void uvd_v2_2_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence); + +/* uvd v3.1 */ +void uvd_v3_1_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait); + +/* uvd v4.2 */ +int uvd_v4_2_resume(struct radeon_device *rdev); + #endif diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index fd9dcb2d182b..aaab7b1bba27 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -801,103 +801,6 @@ u32 rv770_get_xclk(struct radeon_device *rdev) return reference_clock; } -int rv770_uvd_resume(struct radeon_device *rdev) -{ - uint64_t addr; - uint32_t chip_id, size; - int r; - - r = radeon_uvd_resume(rdev); - if (r) - return r; - - /* programm the VCPU memory controller bits 0-27 */ - addr = rdev->uvd.gpu_addr >> 3; - size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 4) >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET0, addr); - WREG32(UVD_VCPU_CACHE_SIZE0, size); - - addr += size; - size = RADEON_UVD_STACK_SIZE >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET1, addr); - WREG32(UVD_VCPU_CACHE_SIZE1, size); - - addr += size; - size = RADEON_UVD_HEAP_SIZE >> 3; - WREG32(UVD_VCPU_CACHE_OFFSET2, addr); - WREG32(UVD_VCPU_CACHE_SIZE2, size); - - /* bits 28-31 */ - addr = (rdev->uvd.gpu_addr >> 28) & 0xF; - WREG32(UVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); - - /* bits 32-39 */ - addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; - WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); - - /* tell firmware which hardware it is running on */ - switch (rdev->family) { - default: - return -EINVAL; - case CHIP_RV710: - chip_id = 0x01000005; - break; - case CHIP_RV730: - chip_id = 0x01000006; - break; - case CHIP_RV740: - chip_id = 0x01000007; - break; - case CHIP_CYPRESS: - case CHIP_HEMLOCK: - chip_id = 0x01000008; - break; - case CHIP_JUNIPER: - chip_id = 0x01000009; - break; - case CHIP_REDWOOD: - chip_id = 0x0100000a; - break; - case CHIP_CEDAR: - chip_id = 0x0100000b; - break; - case CHIP_SUMO: - case CHIP_SUMO2: - chip_id = 0x0100000c; - break; - case CHIP_PALM: - chip_id = 0x0100000e; - break; - case CHIP_CAYMAN: - chip_id = 0x0100000f; - break; - case CHIP_BARTS: - chip_id = 0x01000010; - break; - case CHIP_TURKS: - chip_id = 0x01000011; - break; - case CHIP_CAICOS: - chip_id = 0x01000012; - break; - case CHIP_TAHITI: - chip_id = 0x01000014; - break; - case CHIP_VERDE: - chip_id = 0x01000015; - break; - case CHIP_PITCAIRN: - chip_id = 0x01000016; - break; - case CHIP_ARUBA: - chip_id = 0x01000017; - break; - } - WREG32(UVD_VCPU_CHIP_ID, chip_id); - - return 0; -} - u32 rv770_page_flip(struct radeon_device *rdev, int crtc_id, u64 crtc_base) { struct radeon_crtc *radeon_crtc = rdev->mode_info.crtcs[crtc_id]; @@ -1870,7 +1773,7 @@ static int rv770_startup(struct radeon_device *rdev) return r; } - r = rv770_uvd_resume(rdev); + r = uvd_v2_2_resume(rdev); if (!r) { r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); @@ -1927,7 +1830,7 @@ static int rv770_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev, true); + r = uvd_v1_0_init(rdev); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); @@ -1977,7 +1880,7 @@ int rv770_resume(struct radeon_device *rdev) int rv770_suspend(struct radeon_device *rdev) { r600_audio_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); r700_cp_stop(rdev); r600_dma_stop(rdev); @@ -2092,7 +1995,7 @@ void rv770_fini(struct radeon_device *rdev) radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); rv770_pcie_gart_fini(rdev); - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); diff --git a/drivers/gpu/drm/radeon/rv770d.h b/drivers/gpu/drm/radeon/rv770d.h index 6bef2b7d601b..9fe60e542922 100644 --- a/drivers/gpu/drm/radeon/rv770d.h +++ b/drivers/gpu/drm/radeon/rv770d.h @@ -971,7 +971,21 @@ # define TARGET_LINK_SPEED_MASK (0xf << 0) # define SELECTABLE_DEEMPHASIS (1 << 6) +/* + * PM4 + */ +#define PACKET0(reg, n) ((RADEON_PACKET_TYPE0 << 30) | \ + (((reg) >> 2) & 0xFFFF) | \ + ((n) & 0x3FFF) << 16) +#define PACKET3(op, n) ((RADEON_PACKET_TYPE3 << 30) | \ + (((op) & 0xFF) << 8) | \ + ((n) & 0x3FFF) << 16) + /* UVD */ +#define UVD_GPCOM_VCPU_CMD 0xef0c +#define UVD_GPCOM_VCPU_DATA0 0xef10 +#define UVD_GPCOM_VCPU_DATA1 0xef14 + #define UVD_LMI_EXT40_ADDR 0xf498 #define UVD_VCPU_CHIP_ID 0xf4d4 #define UVD_VCPU_CACHE_OFFSET0 0xf4d8 @@ -985,4 +999,6 @@ #define UVD_RBC_RB_RPTR 0xf690 #define UVD_RBC_RB_WPTR 0xf694 +#define UVD_CONTEXT_ID 0xf6f4 + #endif diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index ae232be62921..f3f79089405e 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6339,7 +6339,7 @@ static int si_startup(struct radeon_device *rdev) } if (rdev->has_uvd) { - r = rv770_uvd_resume(rdev); + r = uvd_v2_2_resume(rdev); if (!r) { r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); @@ -6420,7 +6420,7 @@ static int si_startup(struct radeon_device *rdev) UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, RADEON_CP_PACKET2); if (!r) - r = r600_uvd_init(rdev, true); + r = uvd_v1_0_init(rdev); if (r) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); } @@ -6473,7 +6473,7 @@ int si_suspend(struct radeon_device *rdev) si_cp_enable(rdev, false); cayman_dma_stop(rdev); if (rdev->has_uvd) { - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); } si_irq_suspend(rdev); @@ -6616,7 +6616,7 @@ void si_fini(struct radeon_device *rdev) radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); if (rdev->has_uvd) { - r600_uvd_stop(rdev); + uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); } si_pcie_gart_fini(rdev); diff --git a/drivers/gpu/drm/radeon/uvd_v1_0.c b/drivers/gpu/drm/radeon/uvd_v1_0.c new file mode 100644 index 000000000000..76ca669f0c8e --- /dev/null +++ b/drivers/gpu/drm/radeon/uvd_v1_0.c @@ -0,0 +1,434 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "r600d.h" + +/** + * uvd_v1_0_get_rptr - get read pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware read pointer + */ +uint32_t uvd_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return RREG32(UVD_RBC_RB_RPTR); +} + +/** + * uvd_v1_0_get_wptr - get write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware write pointer + */ +uint32_t uvd_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return RREG32(UVD_RBC_RB_WPTR); +} + +/** + * uvd_v1_0_set_wptr - set write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Commits the write pointer to the hardware + */ +void uvd_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + WREG32(UVD_RBC_RB_WPTR, ring->wptr); +} + +/** + * uvd_v1_0_init - start and test UVD block + * + * @rdev: radeon_device pointer + * + * Initialize the hardware, boot up the VCPU and do some testing + */ +int uvd_v1_0_init(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + uint32_t tmp; + int r; + + /* raise clocks while booting up the VCPU */ + radeon_set_uvd_clocks(rdev, 53300, 40000); + + uvd_v1_0_start(rdev); + + ring->ready = true; + r = radeon_ring_test(rdev, R600_RING_TYPE_UVD_INDEX, ring); + if (r) { + ring->ready = false; + goto done; + } + + r = radeon_ring_lock(rdev, ring, 10); + if (r) { + DRM_ERROR("radeon: ring failed to lock UVD ring (%d).\n", r); + goto done; + } + + tmp = PACKET0(UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); + + tmp = PACKET0(UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); + + tmp = PACKET0(UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, 0xFFFFF); + + /* Clear timeout status bits */ + radeon_ring_write(ring, PACKET0(UVD_SEMA_TIMEOUT_STATUS, 0)); + radeon_ring_write(ring, 0x8); + + radeon_ring_write(ring, PACKET0(UVD_SEMA_CNTL, 0)); + radeon_ring_write(ring, 3); + + radeon_ring_unlock_commit(rdev, ring); + +done: + /* lower clocks again */ + radeon_set_uvd_clocks(rdev, 0, 0); + + if (!r) + DRM_INFO("UVD initialized successfully.\n"); + + return r; +} + +/** + * uvd_v1_0_fini - stop the hardware block + * + * @rdev: radeon_device pointer + * + * Stop the UVD block, mark ring as not ready any more + */ +void uvd_v1_0_fini(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + + uvd_v1_0_stop(rdev); + ring->ready = false; +} + +/** + * uvd_v1_0_start - start UVD block + * + * @rdev: radeon_device pointer + * + * Setup and start the UVD block + */ +int uvd_v1_0_start(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + uint32_t rb_bufsz; + int i, j, r; + + /* disable byte swapping */ + u32 lmi_swap_cntl = 0; + u32 mp_swap_cntl = 0; + + /* disable clock gating */ + WREG32(UVD_CGC_GATE, 0); + + /* disable interupt */ + WREG32_P(UVD_MASTINT_EN, 0, ~(1 << 1)); + + /* Stall UMC and register bus before resetting VCPU */ + WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); + mdelay(1); + + /* put LMI, VCPU, RBC etc... into reset */ + WREG32(UVD_SOFT_RESET, LMI_SOFT_RESET | VCPU_SOFT_RESET | + LBSI_SOFT_RESET | RBC_SOFT_RESET | CSM_SOFT_RESET | + CXW_SOFT_RESET | TAP_SOFT_RESET | LMI_UMC_SOFT_RESET); + mdelay(5); + + /* take UVD block out of reset */ + WREG32_P(SRBM_SOFT_RESET, 0, ~SOFT_RESET_UVD); + mdelay(5); + + /* initialize UVD memory controller */ + WREG32(UVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) | + (1 << 21) | (1 << 9) | (1 << 20)); + +#ifdef __BIG_ENDIAN + /* swap (8 in 32) RB and IB */ + lmi_swap_cntl = 0xa; + mp_swap_cntl = 0; +#endif + WREG32(UVD_LMI_SWAP_CNTL, lmi_swap_cntl); + WREG32(UVD_MP_SWAP_CNTL, mp_swap_cntl); + + WREG32(UVD_MPC_SET_MUXA0, 0x40c2040); + WREG32(UVD_MPC_SET_MUXA1, 0x0); + WREG32(UVD_MPC_SET_MUXB0, 0x40c2040); + WREG32(UVD_MPC_SET_MUXB1, 0x0); + WREG32(UVD_MPC_SET_ALU, 0); + WREG32(UVD_MPC_SET_MUX, 0x88); + + /* take all subblocks out of reset, except VCPU */ + WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); + mdelay(5); + + /* enable VCPU clock */ + WREG32(UVD_VCPU_CNTL, 1 << 9); + + /* enable UMC */ + WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); + + /* boot up the VCPU */ + WREG32(UVD_SOFT_RESET, 0); + mdelay(10); + + WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); + + for (i = 0; i < 10; ++i) { + uint32_t status; + for (j = 0; j < 100; ++j) { + status = RREG32(UVD_STATUS); + if (status & 2) + break; + mdelay(10); + } + r = 0; + if (status & 2) + break; + + DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); + WREG32_P(UVD_SOFT_RESET, VCPU_SOFT_RESET, ~VCPU_SOFT_RESET); + mdelay(10); + WREG32_P(UVD_SOFT_RESET, 0, ~VCPU_SOFT_RESET); + mdelay(10); + r = -1; + } + + if (r) { + DRM_ERROR("UVD not responding, giving up!!!\n"); + return r; + } + + /* enable interupt */ + WREG32_P(UVD_MASTINT_EN, 3<<1, ~(3 << 1)); + + /* force RBC into idle state */ + WREG32(UVD_RBC_RB_CNTL, 0x11010101); + + /* Set the write pointer delay */ + WREG32(UVD_RBC_RB_WPTR_CNTL, 0); + + /* programm the 4GB memory segment for rptr and ring buffer */ + WREG32(UVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) | + (0x7 << 16) | (0x1 << 31)); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(UVD_RBC_RB_RPTR, 0x0); + + ring->wptr = ring->rptr = RREG32(UVD_RBC_RB_RPTR); + WREG32(UVD_RBC_RB_WPTR, ring->wptr); + + /* set the ring address */ + WREG32(UVD_RBC_RB_BASE, ring->gpu_addr); + + /* Set ring buffer size */ + rb_bufsz = drm_order(ring->ring_size); + rb_bufsz = (0x1 << 8) | rb_bufsz; + WREG32_P(UVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f); + + return 0; +} + +/** + * uvd_v1_0_stop - stop UVD block + * + * @rdev: radeon_device pointer + * + * stop the UVD block + */ +void uvd_v1_0_stop(struct radeon_device *rdev) +{ + /* force RBC into idle state */ + WREG32(UVD_RBC_RB_CNTL, 0x11010101); + + /* Stall UMC and register bus before resetting VCPU */ + WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); + mdelay(1); + + /* put VCPU into reset */ + WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); + mdelay(5); + + /* disable VCPU clock */ + WREG32(UVD_VCPU_CNTL, 0x0); + + /* Unstall UMC and register bus */ + WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); +} + +/** + * uvd_v1_0_ring_test - register write test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Test if we can successfully write to the context register + */ +int uvd_v1_0_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + uint32_t tmp = 0; + unsigned i; + int r; + + WREG32(UVD_CONTEXT_ID, 0xCAFEDEAD); + r = radeon_ring_lock(rdev, ring, 3); + if (r) { + DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); + radeon_ring_write(ring, 0xDEADBEEF); + radeon_ring_unlock_commit(rdev, ring); + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(UVD_CONTEXT_ID); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", + ring->idx, tmp); + r = -EINVAL; + } + return r; +} + +/** + * uvd_v1_0_semaphore_emit - emit semaphore command + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * @semaphore: semaphore to emit commands for + * @emit_wait: true if we should emit a wait command + * + * Emit a semaphore command (either wait or signal) to the UVD ring. + */ +void uvd_v1_0_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + uint64_t addr = semaphore->gpu_addr; + + radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_LOW, 0)); + radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); + + radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_HIGH, 0)); + radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); + + radeon_ring_write(ring, PACKET0(UVD_SEMA_CMD, 0)); + radeon_ring_write(ring, emit_wait ? 1 : 0); +} + +/** + * uvd_v1_0_ib_execute - execute indirect buffer + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to execute + * + * Write ring commands to execute the indirect buffer + */ +void uvd_v1_0_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + + radeon_ring_write(ring, PACKET0(UVD_RBC_IB_BASE, 0)); + radeon_ring_write(ring, ib->gpu_addr); + radeon_ring_write(ring, PACKET0(UVD_RBC_IB_SIZE, 0)); + radeon_ring_write(ring, ib->length_dw); +} + +/** + * uvd_v1_0_ib_test - test ib execution + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Test if we can successfully execute an IB + */ +int uvd_v1_0_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_fence *fence = NULL; + int r; + + r = radeon_set_uvd_clocks(rdev, 53300, 40000); + if (r) { + DRM_ERROR("radeon: failed to raise UVD clocks (%d).\n", r); + return r; + } + + r = radeon_uvd_get_create_msg(rdev, ring->idx, 1, NULL); + if (r) { + DRM_ERROR("radeon: failed to get create msg (%d).\n", r); + goto error; + } + + r = radeon_uvd_get_destroy_msg(rdev, ring->idx, 1, &fence); + if (r) { + DRM_ERROR("radeon: failed to get destroy ib (%d).\n", r); + goto error; + } + + r = radeon_fence_wait(fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + goto error; + } + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); +error: + radeon_fence_unref(&fence); + radeon_set_uvd_clocks(rdev, 0, 0); + return r; +} diff --git a/drivers/gpu/drm/radeon/uvd_v2_2.c b/drivers/gpu/drm/radeon/uvd_v2_2.c new file mode 100644 index 000000000000..b19ef4951085 --- /dev/null +++ b/drivers/gpu/drm/radeon/uvd_v2_2.c @@ -0,0 +1,165 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ + +#include +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "rv770d.h" + +/** + * uvd_v2_2_fence_emit - emit an fence & trap command + * + * @rdev: radeon_device pointer + * @fence: fence to emit + * + * Write a fence and a trap command to the ring. + */ +void uvd_v2_2_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + uint64_t addr = rdev->fence_drv[fence->ring].gpu_addr; + + radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); + radeon_ring_write(ring, fence->seq); + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA0, 0)); + radeon_ring_write(ring, addr & 0xffffffff); + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA1, 0)); + radeon_ring_write(ring, upper_32_bits(addr) & 0xff); + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_CMD, 0)); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA0, 0)); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_DATA1, 0)); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_CMD, 0)); + radeon_ring_write(ring, 2); + return; +} + +/** + * uvd_v2_2_resume - memory controller programming + * + * @rdev: radeon_device pointer + * + * Let the UVD memory controller know it's offsets + */ +int uvd_v2_2_resume(struct radeon_device *rdev) +{ + uint64_t addr; + uint32_t chip_id, size; + int r; + + r = radeon_uvd_resume(rdev); + if (r) + return r; + + /* programm the VCPU memory controller bits 0-27 */ + addr = rdev->uvd.gpu_addr >> 3; + size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 4) >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET0, addr); + WREG32(UVD_VCPU_CACHE_SIZE0, size); + + addr += size; + size = RADEON_UVD_STACK_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET1, addr); + WREG32(UVD_VCPU_CACHE_SIZE1, size); + + addr += size; + size = RADEON_UVD_HEAP_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET2, addr); + WREG32(UVD_VCPU_CACHE_SIZE2, size); + + /* bits 28-31 */ + addr = (rdev->uvd.gpu_addr >> 28) & 0xF; + WREG32(UVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); + + /* bits 32-39 */ + addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; + WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); + + /* tell firmware which hardware it is running on */ + switch (rdev->family) { + default: + return -EINVAL; + case CHIP_RV710: + chip_id = 0x01000005; + break; + case CHIP_RV730: + chip_id = 0x01000006; + break; + case CHIP_RV740: + chip_id = 0x01000007; + break; + case CHIP_CYPRESS: + case CHIP_HEMLOCK: + chip_id = 0x01000008; + break; + case CHIP_JUNIPER: + chip_id = 0x01000009; + break; + case CHIP_REDWOOD: + chip_id = 0x0100000a; + break; + case CHIP_CEDAR: + chip_id = 0x0100000b; + break; + case CHIP_SUMO: + case CHIP_SUMO2: + chip_id = 0x0100000c; + break; + case CHIP_PALM: + chip_id = 0x0100000e; + break; + case CHIP_CAYMAN: + chip_id = 0x0100000f; + break; + case CHIP_BARTS: + chip_id = 0x01000010; + break; + case CHIP_TURKS: + chip_id = 0x01000011; + break; + case CHIP_CAICOS: + chip_id = 0x01000012; + break; + case CHIP_TAHITI: + chip_id = 0x01000014; + break; + case CHIP_VERDE: + chip_id = 0x01000015; + break; + case CHIP_PITCAIRN: + chip_id = 0x01000016; + break; + case CHIP_ARUBA: + chip_id = 0x01000017; + break; + } + WREG32(UVD_VCPU_CHIP_ID, chip_id); + + return 0; +} diff --git a/drivers/gpu/drm/radeon/uvd_v3_1.c b/drivers/gpu/drm/radeon/uvd_v3_1.c new file mode 100644 index 000000000000..5b6fa1f62d4e --- /dev/null +++ b/drivers/gpu/drm/radeon/uvd_v3_1.c @@ -0,0 +1,55 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "nid.h" + +/** + * uvd_v3_1_semaphore_emit - emit semaphore command + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * @semaphore: semaphore to emit commands for + * @emit_wait: true if we should emit a wait command + * + * Emit a semaphore command (either wait or signal) to the UVD ring. + */ +void uvd_v3_1_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + uint64_t addr = semaphore->gpu_addr; + + radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_LOW, 0)); + radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); + + radeon_ring_write(ring, PACKET0(UVD_SEMA_ADDR_HIGH, 0)); + radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); + + radeon_ring_write(ring, PACKET0(UVD_SEMA_CMD, 0)); + radeon_ring_write(ring, 0x80 | (emit_wait ? 1 : 0)); +} diff --git a/drivers/gpu/drm/radeon/uvd_v4_2.c b/drivers/gpu/drm/radeon/uvd_v4_2.c new file mode 100644 index 000000000000..d7e480786098 --- /dev/null +++ b/drivers/gpu/drm/radeon/uvd_v4_2.c @@ -0,0 +1,73 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ + +#include +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "cikd.h" + +/** + * uvd_v4_2_resume - memory controller programming + * + * @rdev: radeon_device pointer + * + * Let the UVD memory controller know it's offsets + */ +int uvd_v4_2_resume(struct radeon_device *rdev) +{ + uint64_t addr; + uint32_t size; + int r; + + r = radeon_uvd_resume(rdev); + if (r) + return r; + + /* programm the VCPU memory controller bits 0-27 */ + addr = rdev->uvd.gpu_addr >> 3; + size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->size + 4) >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET0, addr); + WREG32(UVD_VCPU_CACHE_SIZE0, size); + + addr += size; + size = RADEON_UVD_STACK_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET1, addr); + WREG32(UVD_VCPU_CACHE_SIZE1, size); + + addr += size; + size = RADEON_UVD_HEAP_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET2, addr); + WREG32(UVD_VCPU_CACHE_SIZE2, size); + + /* bits 28-31 */ + addr = (rdev->uvd.gpu_addr >> 28) & 0xF; + WREG32(UVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); + + /* bits 32-39 */ + addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; + WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); + + return 0; +} -- cgit v1.2.3 From 2483b4ea982efe8a544697d3f9642932e9af4dc1 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 13 Aug 2013 11:56:54 +0200 Subject: drm/radeon: separate DMA code MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Similar to separating the UVD code, just put the DMA functions into separate files. Signed-off-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/Makefile | 9 + drivers/gpu/drm/radeon/cik.c | 736 +------------------------------ drivers/gpu/drm/radeon/cik_sdma.c | 785 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/evergreen.c | 161 +------ drivers/gpu/drm/radeon/evergreen_dma.c | 190 ++++++++ drivers/gpu/drm/radeon/ni.c | 293 +----------- drivers/gpu/drm/radeon/ni_dma.c | 338 ++++++++++++++ drivers/gpu/drm/radeon/r600.c | 473 +------------------- drivers/gpu/drm/radeon/r600_dma.c | 497 +++++++++++++++++++++ drivers/gpu/drm/radeon/rv770.c | 74 ---- drivers/gpu/drm/radeon/rv770_dma.c | 101 +++++ drivers/gpu/drm/radeon/si.c | 188 +------- drivers/gpu/drm/radeon/si_dma.c | 235 ++++++++++ 13 files changed, 2181 insertions(+), 1899 deletions(-) create mode 100644 drivers/gpu/drm/radeon/cik_sdma.c create mode 100644 drivers/gpu/drm/radeon/evergreen_dma.c create mode 100644 drivers/gpu/drm/radeon/ni_dma.c create mode 100644 drivers/gpu/drm/radeon/r600_dma.c create mode 100644 drivers/gpu/drm/radeon/rv770_dma.c create mode 100644 drivers/gpu/drm/radeon/si_dma.c (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index 1e23b18d549a..da2a8e9e9308 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -82,6 +82,15 @@ radeon-y += radeon_device.o radeon_asic.o radeon_kms.o \ trinity_smc.o ni_dpm.o si_smc.o si_dpm.o kv_smc.o kv_dpm.o ci_smc.o \ ci_dpm.o +# add async DMA block +radeon-y += \ + r600_dma.o \ + rv770_dma.o \ + evergreen_dma.o \ + ni_dma.o \ + si_dma.o \ + cik_sdma.o \ + # add UVD block radeon-y += \ radeon_uvd.o \ diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 1400b5203db1..692e31b95d34 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -64,6 +64,14 @@ extern int sumo_rlc_init(struct radeon_device *rdev); extern void si_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc); extern void si_rlc_reset(struct radeon_device *rdev); extern void si_init_uvd_internal_cg(struct radeon_device *rdev); +extern int cik_sdma_resume(struct radeon_device *rdev); +extern void cik_sdma_enable(struct radeon_device *rdev, bool enable); +extern void cik_sdma_fini(struct radeon_device *rdev); +extern void cik_sdma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags); static void cik_rlc_stop(struct radeon_device *rdev); static void cik_pcie_gen3_enable(struct radeon_device *rdev); static void cik_program_aspm(struct radeon_device *rdev); @@ -3987,579 +3995,6 @@ static int cik_cp_resume(struct radeon_device *rdev) return 0; } -/* - * sDMA - System DMA - * Starting with CIK, the GPU has new asynchronous - * DMA engines. These engines are used for compute - * and gfx. There are two DMA engines (SDMA0, SDMA1) - * and each one supports 1 ring buffer used for gfx - * and 2 queues used for compute. - * - * The programming model is very similar to the CP - * (ring buffer, IBs, etc.), but sDMA has it's own - * packet format that is different from the PM4 format - * used by the CP. sDMA supports copying data, writing - * embedded data, solid fills, and a number of other - * things. It also has support for tiling/detiling of - * buffers. - */ -/** - * cik_sdma_ring_ib_execute - Schedule an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ib: IB object to schedule - * - * Schedule an IB in the DMA ring (CIK). - */ -void cik_sdma_ring_ib_execute(struct radeon_device *rdev, - struct radeon_ib *ib) -{ - struct radeon_ring *ring = &rdev->ring[ib->ring]; - u32 extra_bits = (ib->vm ? ib->vm->id : 0) & 0xf; - - if (rdev->wb.enabled) { - u32 next_rptr = ring->wptr + 5; - while ((next_rptr & 7) != 4) - next_rptr++; - next_rptr += 4; - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); - radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); - radeon_ring_write(ring, 1); /* number of DWs to follow */ - radeon_ring_write(ring, next_rptr); - } - - /* IB packet must end on a 8 DW boundary */ - while ((ring->wptr & 7) != 4) - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits)); - radeon_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */ - radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xffffffff); - radeon_ring_write(ring, ib->length_dw); - -} - -/** - * cik_sdma_fence_ring_emit - emit a fence on the DMA ring - * - * @rdev: radeon_device pointer - * @fence: radeon fence object - * - * Add a DMA fence packet to the ring to write - * the fence seq number and DMA trap packet to generate - * an interrupt if needed (CIK). - */ -void cik_sdma_fence_ring_emit(struct radeon_device *rdev, - struct radeon_fence *fence) -{ - struct radeon_ring *ring = &rdev->ring[fence->ring]; - u64 addr = rdev->fence_drv[fence->ring].gpu_addr; - u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | - SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ - u32 ref_and_mask; - - if (fence->ring == R600_RING_TYPE_DMA_INDEX) - ref_and_mask = SDMA0; - else - ref_and_mask = SDMA1; - - /* write the fence */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_FENCE, 0, 0)); - radeon_ring_write(ring, addr & 0xffffffff); - radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); - radeon_ring_write(ring, fence->seq); - /* generate an interrupt */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_TRAP, 0, 0)); - /* flush HDP */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); - radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); - radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); - radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ - radeon_ring_write(ring, ref_and_mask); /* MASK */ - radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ -} - -/** - * cik_sdma_semaphore_ring_emit - emit a semaphore on the dma ring - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * @semaphore: radeon semaphore object - * @emit_wait: wait or signal semaphore - * - * Add a DMA semaphore packet to the ring wait on or signal - * other rings (CIK). - */ -void cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, - struct radeon_ring *ring, - struct radeon_semaphore *semaphore, - bool emit_wait) -{ - u64 addr = semaphore->gpu_addr; - u32 extra_bits = emit_wait ? 0 : SDMA_SEMAPHORE_EXTRA_S; - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SEMAPHORE, 0, extra_bits)); - radeon_ring_write(ring, addr & 0xfffffff8); - radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); -} - -/** - * cik_sdma_gfx_stop - stop the gfx async dma engines - * - * @rdev: radeon_device pointer - * - * Stop the gfx async dma ring buffers (CIK). - */ -static void cik_sdma_gfx_stop(struct radeon_device *rdev) -{ - u32 rb_cntl, reg_offset; - int i; - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); - - for (i = 0; i < 2; i++) { - if (i == 0) - reg_offset = SDMA0_REGISTER_OFFSET; - else - reg_offset = SDMA1_REGISTER_OFFSET; - rb_cntl = RREG32(SDMA0_GFX_RB_CNTL + reg_offset); - rb_cntl &= ~SDMA_RB_ENABLE; - WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); - WREG32(SDMA0_GFX_IB_CNTL + reg_offset, 0); - } -} - -/** - * cik_sdma_rlc_stop - stop the compute async dma engines - * - * @rdev: radeon_device pointer - * - * Stop the compute async dma queues (CIK). - */ -static void cik_sdma_rlc_stop(struct radeon_device *rdev) -{ - /* XXX todo */ -} - -/** - * cik_sdma_enable - stop the async dma engines - * - * @rdev: radeon_device pointer - * @enable: enable/disable the DMA MEs. - * - * Halt or unhalt the async dma engines (CIK). - */ -static void cik_sdma_enable(struct radeon_device *rdev, bool enable) -{ - u32 me_cntl, reg_offset; - int i; - - for (i = 0; i < 2; i++) { - if (i == 0) - reg_offset = SDMA0_REGISTER_OFFSET; - else - reg_offset = SDMA1_REGISTER_OFFSET; - me_cntl = RREG32(SDMA0_ME_CNTL + reg_offset); - if (enable) - me_cntl &= ~SDMA_HALT; - else - me_cntl |= SDMA_HALT; - WREG32(SDMA0_ME_CNTL + reg_offset, me_cntl); - } -} - -/** - * cik_sdma_gfx_resume - setup and start the async dma engines - * - * @rdev: radeon_device pointer - * - * Set up the gfx DMA ring buffers and enable them (CIK). - * Returns 0 for success, error for failure. - */ -static int cik_sdma_gfx_resume(struct radeon_device *rdev) -{ - struct radeon_ring *ring; - u32 rb_cntl, ib_cntl; - u32 rb_bufsz; - u32 reg_offset, wb_offset; - int i, r; - - for (i = 0; i < 2; i++) { - if (i == 0) { - ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; - reg_offset = SDMA0_REGISTER_OFFSET; - wb_offset = R600_WB_DMA_RPTR_OFFSET; - } else { - ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; - reg_offset = SDMA1_REGISTER_OFFSET; - wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET; - } - - WREG32(SDMA0_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0); - WREG32(SDMA0_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0); - - /* Set ring buffer size in dwords */ - rb_bufsz = drm_order(ring->ring_size / 4); - rb_cntl = rb_bufsz << 1; -#ifdef __BIG_ENDIAN - rb_cntl |= SDMA_RB_SWAP_ENABLE | SDMA_RPTR_WRITEBACK_SWAP_ENABLE; -#endif - WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); - - /* Initialize the ring buffer's read and write pointers */ - WREG32(SDMA0_GFX_RB_RPTR + reg_offset, 0); - WREG32(SDMA0_GFX_RB_WPTR + reg_offset, 0); - - /* set the wb address whether it's enabled or not */ - WREG32(SDMA0_GFX_RB_RPTR_ADDR_HI + reg_offset, - upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); - WREG32(SDMA0_GFX_RB_RPTR_ADDR_LO + reg_offset, - ((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); - - if (rdev->wb.enabled) - rb_cntl |= SDMA_RPTR_WRITEBACK_ENABLE; - - WREG32(SDMA0_GFX_RB_BASE + reg_offset, ring->gpu_addr >> 8); - WREG32(SDMA0_GFX_RB_BASE_HI + reg_offset, ring->gpu_addr >> 40); - - ring->wptr = 0; - WREG32(SDMA0_GFX_RB_WPTR + reg_offset, ring->wptr << 2); - - ring->rptr = RREG32(SDMA0_GFX_RB_RPTR + reg_offset) >> 2; - - /* enable DMA RB */ - WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl | SDMA_RB_ENABLE); - - ib_cntl = SDMA_IB_ENABLE; -#ifdef __BIG_ENDIAN - ib_cntl |= SDMA_IB_SWAP_ENABLE; -#endif - /* enable DMA IBs */ - WREG32(SDMA0_GFX_IB_CNTL + reg_offset, ib_cntl); - - ring->ready = true; - - r = radeon_ring_test(rdev, ring->idx, ring); - if (r) { - ring->ready = false; - return r; - } - } - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); - - return 0; -} - -/** - * cik_sdma_rlc_resume - setup and start the async dma engines - * - * @rdev: radeon_device pointer - * - * Set up the compute DMA queues and enable them (CIK). - * Returns 0 for success, error for failure. - */ -static int cik_sdma_rlc_resume(struct radeon_device *rdev) -{ - /* XXX todo */ - return 0; -} - -/** - * cik_sdma_load_microcode - load the sDMA ME ucode - * - * @rdev: radeon_device pointer - * - * Loads the sDMA0/1 ucode. - * Returns 0 for success, -EINVAL if the ucode is not available. - */ -static int cik_sdma_load_microcode(struct radeon_device *rdev) -{ - const __be32 *fw_data; - int i; - - if (!rdev->sdma_fw) - return -EINVAL; - - /* stop the gfx rings and rlc compute queues */ - cik_sdma_gfx_stop(rdev); - cik_sdma_rlc_stop(rdev); - - /* halt the MEs */ - cik_sdma_enable(rdev, false); - - /* sdma0 */ - fw_data = (const __be32 *)rdev->sdma_fw->data; - WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); - for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) - WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, be32_to_cpup(fw_data++)); - WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); - - /* sdma1 */ - fw_data = (const __be32 *)rdev->sdma_fw->data; - WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); - for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) - WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, be32_to_cpup(fw_data++)); - WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); - - WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); - WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); - return 0; -} - -/** - * cik_sdma_resume - setup and start the async dma engines - * - * @rdev: radeon_device pointer - * - * Set up the DMA engines and enable them (CIK). - * Returns 0 for success, error for failure. - */ -static int cik_sdma_resume(struct radeon_device *rdev) -{ - int r; - - /* Reset dma */ - WREG32(SRBM_SOFT_RESET, SOFT_RESET_SDMA | SOFT_RESET_SDMA1); - RREG32(SRBM_SOFT_RESET); - udelay(50); - WREG32(SRBM_SOFT_RESET, 0); - RREG32(SRBM_SOFT_RESET); - - r = cik_sdma_load_microcode(rdev); - if (r) - return r; - - /* unhalt the MEs */ - cik_sdma_enable(rdev, true); - - /* start the gfx rings and rlc compute queues */ - r = cik_sdma_gfx_resume(rdev); - if (r) - return r; - r = cik_sdma_rlc_resume(rdev); - if (r) - return r; - - return 0; -} - -/** - * cik_sdma_fini - tear down the async dma engines - * - * @rdev: radeon_device pointer - * - * Stop the async dma engines and free the rings (CIK). - */ -static void cik_sdma_fini(struct radeon_device *rdev) -{ - /* stop the gfx rings and rlc compute queues */ - cik_sdma_gfx_stop(rdev); - cik_sdma_rlc_stop(rdev); - /* halt the MEs */ - cik_sdma_enable(rdev, false); - radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); - radeon_ring_fini(rdev, &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]); - /* XXX - compute dma queue tear down */ -} - -/** - * cik_copy_dma - copy pages using the DMA engine - * - * @rdev: radeon_device pointer - * @src_offset: src GPU address - * @dst_offset: dst GPU address - * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object - * - * Copy GPU paging using the DMA engine (CIK). - * Used by the radeon ttm implementation to move pages if - * registered as the asic copy callback. - */ -int cik_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - int ring_index = rdev->asic->copy.dma_ring_index; - struct radeon_ring *ring = &rdev->ring[ring_index]; - u32 size_in_bytes, cur_size_in_bytes; - int i, num_loops; - int r = 0; - - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; - } - - size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); - num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); - r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, ring->idx)) { - radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, - ring->idx); - radeon_fence_note_sync(*fence, ring->idx); - } else { - radeon_semaphore_free(rdev, &sem, NULL); - } - - for (i = 0; i < num_loops; i++) { - cur_size_in_bytes = size_in_bytes; - if (cur_size_in_bytes > 0x1fffff) - cur_size_in_bytes = 0x1fffff; - size_in_bytes -= cur_size_in_bytes; - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0)); - radeon_ring_write(ring, cur_size_in_bytes); - radeon_ring_write(ring, 0); /* src/dst endian swap */ - radeon_ring_write(ring, src_offset & 0xffffffff); - radeon_ring_write(ring, upper_32_bits(src_offset) & 0xffffffff); - radeon_ring_write(ring, dst_offset & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xffffffff); - src_offset += cur_size_in_bytes; - dst_offset += cur_size_in_bytes; - } - - r = radeon_fence_emit(rdev, fence, ring->idx); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return r; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); - - return r; -} - -/** - * cik_sdma_ring_test - simple async dma engine test - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Test the DMA engine by writing using it to write an - * value to memory. (CIK). - * Returns 0 for success, error for failure. - */ -int cik_sdma_ring_test(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - unsigned i; - int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; - u32 tmp; - - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } - - tmp = 0xCAFEDEAD; - writel(tmp, ptr); - - r = radeon_ring_lock(rdev, ring, 4); - if (r) { - DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r); - return r; - } - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); - radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff); - radeon_ring_write(ring, 1); /* number of DWs to follow */ - radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); - - for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - - if (i < rdev->usec_timeout) { - DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - return r; -} - -/** - * cik_sdma_ib_test - test an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Test a simple IB in the DMA ring (CIK). - * Returns 0 on success, error on failure. - */ -int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) -{ - struct radeon_ib ib; - unsigned i; - int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; - u32 tmp = 0; - - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } - - tmp = 0xCAFEDEAD; - writel(tmp, ptr); - - r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); - if (r) { - DRM_ERROR("radeon: failed to get ib (%d).\n", r); - return r; - } - - ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; - ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff; - ib.ptr[3] = 1; - ib.ptr[4] = 0xDEADBEEF; - ib.length_dw = 5; - - r = radeon_ib_schedule(rdev, &ib, NULL); - if (r) { - radeon_ib_free(rdev, &ib); - DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); - return r; - } - r = radeon_fence_wait(ib.fence, false); - if (r) { - DRM_ERROR("radeon: fence wait failed (%d).\n", r); - return r; - } - for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < rdev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); - } else { - DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp); - r = -EINVAL; - } - radeon_ib_free(rdev, &ib); - return r; -} - - static void cik_print_gpu_status_regs(struct radeon_device *rdev) { dev_info(rdev->dev, " GRBM_STATUS=0x%08X\n", @@ -4609,7 +4044,7 @@ static void cik_print_gpu_status_regs(struct radeon_device *rdev) * mask to be used by cik_gpu_soft_reset(). * Returns a mask of the blocks to be reset. */ -static u32 cik_gpu_check_soft_reset(struct radeon_device *rdev) +u32 cik_gpu_check_soft_reset(struct radeon_device *rdev) { u32 reset_mask = 0; u32 tmp; @@ -4860,34 +4295,6 @@ bool cik_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) return radeon_ring_test_lockup(rdev, ring); } -/** - * cik_sdma_is_lockup - Check if the DMA engine is locked up - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Check if the async DMA engine is locked up (CIK). - * Returns true if the engine appears to be locked up, false if not. - */ -bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 reset_mask = cik_gpu_check_soft_reset(rdev); - u32 mask; - - if (ring->idx == R600_RING_TYPE_DMA_INDEX) - mask = RADEON_RESET_DMA; - else - mask = RADEON_RESET_DMA1; - - if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - /* MC */ /** * cik_mc_program - program the GPU memory controller @@ -5424,131 +4831,8 @@ void cik_vm_set_page(struct radeon_device *rdev, } } else { /* DMA */ - if (flags & RADEON_VM_PAGE_SYSTEM) { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = ndw; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - if (flags & RADEON_VM_PAGE_SYSTEM) { - value = radeon_vm_map_gart(rdev, addr); - value &= 0xFFFFFFFFFFFFF000ULL; - } else if (flags & RADEON_VM_PAGE_VALID) { - value = addr; - } else { - value = 0; - } - addr += incr; - value |= r600_flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } - } - } else { - while (count) { - ndw = count; - if (ndw > 0x7FFFF) - ndw = 0x7FFFF; - - if (flags & RADEON_VM_PAGE_VALID) - value = addr; - else - value = 0; - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe); - ib->ptr[ib->length_dw++] = r600_flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = ndw; /* number of entries */ - pe += ndw * 8; - addr += ndw * incr; - count -= ndw; - } - } - while (ib->length_dw & 0x7) - ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0); - } -} - -/** - * cik_dma_vm_flush - cik vm flush using sDMA - * - * @rdev: radeon_device pointer - * - * Update the page table base and flush the VM TLB - * using sDMA (CIK). - */ -void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) -{ - struct radeon_ring *ring = &rdev->ring[ridx]; - u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | - SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ - u32 ref_and_mask; - - if (vm == NULL) - return; - - if (ridx == R600_RING_TYPE_DMA_INDEX) - ref_and_mask = SDMA0; - else - ref_and_mask = SDMA1; - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - if (vm->id < 8) { - radeon_ring_write(ring, (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); - } else { - radeon_ring_write(ring, (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + cik_sdma_vm_set_page(rdev, ib, pe, addr, count, incr, flags); } - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); - - /* update SH_MEM_* regs */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); - radeon_ring_write(ring, VMID(vm->id)); - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SH_MEM_BASES >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SH_MEM_CONFIG >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SH_MEM_APE1_BASE >> 2); - radeon_ring_write(ring, 1); - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SH_MEM_APE1_LIMIT >> 2); - radeon_ring_write(ring, 0); - - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); - radeon_ring_write(ring, VMID(0)); - - /* flush HDP */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); - radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); - radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); - radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ - radeon_ring_write(ring, ref_and_mask); /* MASK */ - radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ - - /* flush TLB */ - radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); - radeon_ring_write(ring, 1 << vm->id); } /* diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c new file mode 100644 index 000000000000..8925185a0049 --- /dev/null +++ b/drivers/gpu/drm/radeon/cik_sdma.c @@ -0,0 +1,785 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "cikd.h" + +/* sdma */ +#define CIK_SDMA_UCODE_SIZE 1050 +#define CIK_SDMA_UCODE_VERSION 64 + +u32 cik_gpu_check_soft_reset(struct radeon_device *rdev); + +/* + * sDMA - System DMA + * Starting with CIK, the GPU has new asynchronous + * DMA engines. These engines are used for compute + * and gfx. There are two DMA engines (SDMA0, SDMA1) + * and each one supports 1 ring buffer used for gfx + * and 2 queues used for compute. + * + * The programming model is very similar to the CP + * (ring buffer, IBs, etc.), but sDMA has it's own + * packet format that is different from the PM4 format + * used by the CP. sDMA supports copying data, writing + * embedded data, solid fills, and a number of other + * things. It also has support for tiling/detiling of + * buffers. + */ + +/** + * cik_sdma_ring_ib_execute - Schedule an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring (CIK). + */ +void cik_sdma_ring_ib_execute(struct radeon_device *rdev, + struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + u32 extra_bits = (ib->vm ? ib->vm->id : 0) & 0xf; + + if (rdev->wb.enabled) { + u32 next_rptr = ring->wptr + 5; + while ((next_rptr & 7) != 4) + next_rptr++; + next_rptr += 4; + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + radeon_ring_write(ring, 1); /* number of DWs to follow */ + radeon_ring_write(ring, next_rptr); + } + + /* IB packet must end on a 8 DW boundary */ + while ((ring->wptr & 7) != 4) + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits)); + radeon_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */ + radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xffffffff); + radeon_ring_write(ring, ib->length_dw); + +} + +/** + * cik_sdma_fence_ring_emit - emit a fence on the DMA ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Add a DMA fence packet to the ring to write + * the fence seq number and DMA trap packet to generate + * an interrupt if needed (CIK). + */ +void cik_sdma_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | + SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ + u32 ref_and_mask; + + if (fence->ring == R600_RING_TYPE_DMA_INDEX) + ref_and_mask = SDMA0; + else + ref_and_mask = SDMA1; + + /* write the fence */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_FENCE, 0, 0)); + radeon_ring_write(ring, addr & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); + radeon_ring_write(ring, fence->seq); + /* generate an interrupt */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_TRAP, 0, 0)); + /* flush HDP */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); + radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); + radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); + radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ + radeon_ring_write(ring, ref_and_mask); /* MASK */ + radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ +} + +/** + * cik_sdma_semaphore_ring_emit - emit a semaphore on the dma ring + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * @semaphore: radeon semaphore object + * @emit_wait: wait or signal semaphore + * + * Add a DMA semaphore packet to the ring wait on or signal + * other rings (CIK). + */ +void cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + u64 addr = semaphore->gpu_addr; + u32 extra_bits = emit_wait ? 0 : SDMA_SEMAPHORE_EXTRA_S; + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SEMAPHORE, 0, extra_bits)); + radeon_ring_write(ring, addr & 0xfffffff8); + radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); +} + +/** + * cik_sdma_gfx_stop - stop the gfx async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the gfx async dma ring buffers (CIK). + */ +static void cik_sdma_gfx_stop(struct radeon_device *rdev) +{ + u32 rb_cntl, reg_offset; + int i; + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); + + for (i = 0; i < 2; i++) { + if (i == 0) + reg_offset = SDMA0_REGISTER_OFFSET; + else + reg_offset = SDMA1_REGISTER_OFFSET; + rb_cntl = RREG32(SDMA0_GFX_RB_CNTL + reg_offset); + rb_cntl &= ~SDMA_RB_ENABLE; + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); + WREG32(SDMA0_GFX_IB_CNTL + reg_offset, 0); + } +} + +/** + * cik_sdma_rlc_stop - stop the compute async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the compute async dma queues (CIK). + */ +static void cik_sdma_rlc_stop(struct radeon_device *rdev) +{ + /* XXX todo */ +} + +/** + * cik_sdma_enable - stop the async dma engines + * + * @rdev: radeon_device pointer + * @enable: enable/disable the DMA MEs. + * + * Halt or unhalt the async dma engines (CIK). + */ +void cik_sdma_enable(struct radeon_device *rdev, bool enable) +{ + u32 me_cntl, reg_offset; + int i; + + for (i = 0; i < 2; i++) { + if (i == 0) + reg_offset = SDMA0_REGISTER_OFFSET; + else + reg_offset = SDMA1_REGISTER_OFFSET; + me_cntl = RREG32(SDMA0_ME_CNTL + reg_offset); + if (enable) + me_cntl &= ~SDMA_HALT; + else + me_cntl |= SDMA_HALT; + WREG32(SDMA0_ME_CNTL + reg_offset, me_cntl); + } +} + +/** + * cik_sdma_gfx_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the gfx DMA ring buffers and enable them (CIK). + * Returns 0 for success, error for failure. + */ +static int cik_sdma_gfx_resume(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + u32 rb_cntl, ib_cntl; + u32 rb_bufsz; + u32 reg_offset, wb_offset; + int i, r; + + for (i = 0; i < 2; i++) { + if (i == 0) { + ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + reg_offset = SDMA0_REGISTER_OFFSET; + wb_offset = R600_WB_DMA_RPTR_OFFSET; + } else { + ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; + reg_offset = SDMA1_REGISTER_OFFSET; + wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET; + } + + WREG32(SDMA0_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0); + WREG32(SDMA0_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0); + + /* Set ring buffer size in dwords */ + rb_bufsz = drm_order(ring->ring_size / 4); + rb_cntl = rb_bufsz << 1; +#ifdef __BIG_ENDIAN + rb_cntl |= SDMA_RB_SWAP_ENABLE | SDMA_RPTR_WRITEBACK_SWAP_ENABLE; +#endif + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(SDMA0_GFX_RB_RPTR + reg_offset, 0); + WREG32(SDMA0_GFX_RB_WPTR + reg_offset, 0); + + /* set the wb address whether it's enabled or not */ + WREG32(SDMA0_GFX_RB_RPTR_ADDR_HI + reg_offset, + upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + WREG32(SDMA0_GFX_RB_RPTR_ADDR_LO + reg_offset, + ((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); + + if (rdev->wb.enabled) + rb_cntl |= SDMA_RPTR_WRITEBACK_ENABLE; + + WREG32(SDMA0_GFX_RB_BASE + reg_offset, ring->gpu_addr >> 8); + WREG32(SDMA0_GFX_RB_BASE_HI + reg_offset, ring->gpu_addr >> 40); + + ring->wptr = 0; + WREG32(SDMA0_GFX_RB_WPTR + reg_offset, ring->wptr << 2); + + ring->rptr = RREG32(SDMA0_GFX_RB_RPTR + reg_offset) >> 2; + + /* enable DMA RB */ + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl | SDMA_RB_ENABLE); + + ib_cntl = SDMA_IB_ENABLE; +#ifdef __BIG_ENDIAN + ib_cntl |= SDMA_IB_SWAP_ENABLE; +#endif + /* enable DMA IBs */ + WREG32(SDMA0_GFX_IB_CNTL + reg_offset, ib_cntl); + + ring->ready = true; + + r = radeon_ring_test(rdev, ring->idx, ring); + if (r) { + ring->ready = false; + return r; + } + } + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); + + return 0; +} + +/** + * cik_sdma_rlc_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the compute DMA queues and enable them (CIK). + * Returns 0 for success, error for failure. + */ +static int cik_sdma_rlc_resume(struct radeon_device *rdev) +{ + /* XXX todo */ + return 0; +} + +/** + * cik_sdma_load_microcode - load the sDMA ME ucode + * + * @rdev: radeon_device pointer + * + * Loads the sDMA0/1 ucode. + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static int cik_sdma_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + int i; + + if (!rdev->sdma_fw) + return -EINVAL; + + /* stop the gfx rings and rlc compute queues */ + cik_sdma_gfx_stop(rdev); + cik_sdma_rlc_stop(rdev); + + /* halt the MEs */ + cik_sdma_enable(rdev, false); + + /* sdma0 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + /* sdma1 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + return 0; +} + +/** + * cik_sdma_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the DMA engines and enable them (CIK). + * Returns 0 for success, error for failure. + */ +int cik_sdma_resume(struct radeon_device *rdev) +{ + int r; + + /* Reset dma */ + WREG32(SRBM_SOFT_RESET, SOFT_RESET_SDMA | SOFT_RESET_SDMA1); + RREG32(SRBM_SOFT_RESET); + udelay(50); + WREG32(SRBM_SOFT_RESET, 0); + RREG32(SRBM_SOFT_RESET); + + r = cik_sdma_load_microcode(rdev); + if (r) + return r; + + /* unhalt the MEs */ + cik_sdma_enable(rdev, true); + + /* start the gfx rings and rlc compute queues */ + r = cik_sdma_gfx_resume(rdev); + if (r) + return r; + r = cik_sdma_rlc_resume(rdev); + if (r) + return r; + + return 0; +} + +/** + * cik_sdma_fini - tear down the async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the async dma engines and free the rings (CIK). + */ +void cik_sdma_fini(struct radeon_device *rdev) +{ + /* stop the gfx rings and rlc compute queues */ + cik_sdma_gfx_stop(rdev); + cik_sdma_rlc_stop(rdev); + /* halt the MEs */ + cik_sdma_enable(rdev, false); + radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); + radeon_ring_fini(rdev, &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]); + /* XXX - compute dma queue tear down */ +} + +/** + * cik_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (CIK). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int cik_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_bytes, cur_size_in_bytes; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); + num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); + r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_bytes = size_in_bytes; + if (cur_size_in_bytes > 0x1fffff) + cur_size_in_bytes = 0x1fffff; + size_in_bytes -= cur_size_in_bytes; + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, cur_size_in_bytes); + radeon_ring_write(ring, 0); /* src/dst endian swap */ + radeon_ring_write(ring, src_offset & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(src_offset) & 0xffffffff); + radeon_ring_write(ring, dst_offset & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xffffffff); + src_offset += cur_size_in_bytes; + dst_offset += cur_size_in_bytes; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} + +/** + * cik_sdma_ring_test - simple async dma engine test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test the DMA engine by writing using it to write an + * value to memory. (CIK). + * Returns 0 for success, error for failure. + */ +int cik_sdma_ring_test(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + unsigned i; + int r; + void __iomem *ptr = (void *)rdev->vram_scratch.ptr; + u32 tmp; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ring_lock(rdev, ring, 4); + if (r) { + DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r); + return r; + } + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff); + radeon_ring_write(ring, 1); /* number of DWs to follow */ + radeon_ring_write(ring, 0xDEADBEEF); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", + ring->idx, tmp); + r = -EINVAL; + } + return r; +} + +/** + * cik_sdma_ib_test - test an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test a simple IB in the DMA ring (CIK). + * Returns 0 on success, error on failure. + */ +int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_ib ib; + unsigned i; + int r; + void __iomem *ptr = (void *)rdev->vram_scratch.ptr; + u32 tmp = 0; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; + ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff; + ib.ptr[3] = 1; + ib.ptr[4] = 0xDEADBEEF; + ib.length_dw = 5; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + radeon_ib_free(rdev, &ib); + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + return r; + } + r = radeon_fence_wait(ib.fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + return r; + } + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); + } else { + DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp); + r = -EINVAL; + } + radeon_ib_free(rdev, &ib); + return r; +} + +/** + * cik_sdma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up (CIK). + * Returns true if the engine appears to be locked up, false if not. + */ +bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = cik_gpu_check_soft_reset(rdev); + u32 mask; + + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + mask = RADEON_RESET_DMA; + else + mask = RADEON_RESET_DMA1; + + if (!(reset_mask & mask)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + +/** + * cik_sdma_vm_set_page - update the page tables using sDMA + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update the page tables using sDMA (CIK). + */ +void cik_sdma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) +{ + uint32_t r600_flags = cayman_vm_page_flags(rdev, flags); + uint64_t value; + unsigned ndw; + + if (flags & RADEON_VM_PAGE_SYSTEM) { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + /* for non-physically contiguous pages (system) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + if (flags & RADEON_VM_PAGE_SYSTEM) { + value = radeon_vm_map_gart(rdev, addr); + value &= 0xFFFFFFFFFFFFF000ULL; + } else if (flags & RADEON_VM_PAGE_VALID) { + value = addr; + } else { + value = 0; + } + addr += incr; + value |= r600_flags; + ib->ptr[ib->length_dw++] = value; + ib->ptr[ib->length_dw++] = upper_32_bits(value); + } + } + } else { + while (count) { + ndw = count; + if (ndw > 0x7FFFF) + ndw = 0x7FFFF; + + if (flags & RADEON_VM_PAGE_VALID) + value = addr; + else + value = 0; + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); + ib->ptr[ib->length_dw++] = pe; /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = r600_flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = value; /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(value); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = ndw; /* number of entries */ + pe += ndw * 8; + addr += ndw * incr; + count -= ndw; + } + } + while (ib->length_dw & 0x7) + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0); +} + +/** + * cik_dma_vm_flush - cik vm flush using sDMA + * + * @rdev: radeon_device pointer + * + * Update the page table base and flush the VM TLB + * using sDMA (CIK). + */ +void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +{ + struct radeon_ring *ring = &rdev->ring[ridx]; + u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | + SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ + u32 ref_and_mask; + + if (vm == NULL) + return; + + if (ridx == R600_RING_TYPE_DMA_INDEX) + ref_and_mask = SDMA0; + else + ref_and_mask = SDMA1; + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + if (vm->id < 8) { + radeon_ring_write(ring, (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + } else { + radeon_ring_write(ring, (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + } + radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + + /* update SH_MEM_* regs */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, VMID(vm->id)); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_BASES >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_CONFIG >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_APE1_BASE >> 2); + radeon_ring_write(ring, 1); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_APE1_LIMIT >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, VMID(0)); + + /* flush HDP */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); + radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); + radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); + radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ + radeon_ring_write(ring, ref_and_mask); /* MASK */ + radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ + + /* flush TLB */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); + radeon_ring_write(ring, 1 << vm->id); +} + diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 52ed22333f0d..bbaa4f2056ce 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -3613,7 +3613,7 @@ bool evergreen_is_display_hung(struct radeon_device *rdev) return true; } -static u32 evergreen_gpu_check_soft_reset(struct radeon_device *rdev) +u32 evergreen_gpu_check_soft_reset(struct radeon_device *rdev) { u32 reset_mask = 0; u32 tmp; @@ -3839,28 +3839,6 @@ bool evergreen_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin return radeon_ring_test_lockup(rdev, ring); } -/** - * evergreen_dma_is_lockup - Check if the DMA engine is locked up - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Check if the async DMA engine is locked up. - * Returns true if the engine appears to be locked up, false if not. - */ -bool evergreen_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 reset_mask = evergreen_gpu_check_soft_reset(rdev); - - if (!(reset_mask & RADEON_RESET_DMA)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - /* * RLC */ @@ -5024,143 +5002,6 @@ restart_ih: return IRQ_HANDLED; } -/** - * evergreen_dma_fence_ring_emit - emit a fence on the DMA ring - * - * @rdev: radeon_device pointer - * @fence: radeon fence object - * - * Add a DMA fence packet to the ring to write - * the fence seq number and DMA trap packet to generate - * an interrupt if needed (evergreen-SI). - */ -void evergreen_dma_fence_ring_emit(struct radeon_device *rdev, - struct radeon_fence *fence) -{ - struct radeon_ring *ring = &rdev->ring[fence->ring]; - u64 addr = rdev->fence_drv[fence->ring].gpu_addr; - /* write the fence */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0)); - radeon_ring_write(ring, addr & 0xfffffffc); - radeon_ring_write(ring, (upper_32_bits(addr) & 0xff)); - radeon_ring_write(ring, fence->seq); - /* generate an interrupt */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0)); - /* flush HDP */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); - radeon_ring_write(ring, 1); -} - -/** - * evergreen_dma_ring_ib_execute - schedule an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ib: IB object to schedule - * - * Schedule an IB in the DMA ring (evergreen). - */ -void evergreen_dma_ring_ib_execute(struct radeon_device *rdev, - struct radeon_ib *ib) -{ - struct radeon_ring *ring = &rdev->ring[ib->ring]; - - if (rdev->wb.enabled) { - u32 next_rptr = ring->wptr + 4; - while ((next_rptr & 7) != 5) - next_rptr++; - next_rptr += 3; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 1)); - radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); - radeon_ring_write(ring, next_rptr); - } - - /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. - * Pad as necessary with NOPs. - */ - while ((ring->wptr & 7) != 5) - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0)); - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_INDIRECT_BUFFER, 0, 0)); - radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); - radeon_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF)); - -} - -/** - * evergreen_copy_dma - copy pages using the DMA engine - * - * @rdev: radeon_device pointer - * @src_offset: src GPU address - * @dst_offset: dst GPU address - * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object - * - * Copy GPU paging using the DMA engine (evergreen-cayman). - * Used by the radeon ttm implementation to move pages if - * registered as the asic copy callback. - */ -int evergreen_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - int ring_index = rdev->asic->copy.dma_ring_index; - struct radeon_ring *ring = &rdev->ring[ring_index]; - u32 size_in_dw, cur_size_in_dw; - int i, num_loops; - int r = 0; - - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; - } - - size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; - num_loops = DIV_ROUND_UP(size_in_dw, 0xfffff); - r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, ring->idx)) { - radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, - ring->idx); - radeon_fence_note_sync(*fence, ring->idx); - } else { - radeon_semaphore_free(rdev, &sem, NULL); - } - - for (i = 0; i < num_loops; i++) { - cur_size_in_dw = size_in_dw; - if (cur_size_in_dw > 0xFFFFF) - cur_size_in_dw = 0xFFFFF; - size_in_dw -= cur_size_in_dw; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, cur_size_in_dw)); - radeon_ring_write(ring, dst_offset & 0xfffffffc); - radeon_ring_write(ring, src_offset & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); - radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); - src_offset += cur_size_in_dw * 4; - dst_offset += cur_size_in_dw * 4; - } - - r = radeon_fence_emit(rdev, fence, ring->idx); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return r; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); - - return r; -} - static int evergreen_startup(struct radeon_device *rdev) { struct radeon_ring *ring; diff --git a/drivers/gpu/drm/radeon/evergreen_dma.c b/drivers/gpu/drm/radeon/evergreen_dma.c new file mode 100644 index 000000000000..6a0656d00ed0 --- /dev/null +++ b/drivers/gpu/drm/radeon/evergreen_dma.c @@ -0,0 +1,190 @@ +/* + * Copyright 2010 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "evergreend.h" + +u32 evergreen_gpu_check_soft_reset(struct radeon_device *rdev); + +/** + * evergreen_dma_fence_ring_emit - emit a fence on the DMA ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Add a DMA fence packet to the ring to write + * the fence seq number and DMA trap packet to generate + * an interrupt if needed (evergreen-SI). + */ +void evergreen_dma_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + /* write the fence */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0)); + radeon_ring_write(ring, addr & 0xfffffffc); + radeon_ring_write(ring, (upper_32_bits(addr) & 0xff)); + radeon_ring_write(ring, fence->seq); + /* generate an interrupt */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0)); + /* flush HDP */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); + radeon_ring_write(ring, 1); +} + +/** + * evergreen_dma_ring_ib_execute - schedule an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring (evergreen). + */ +void evergreen_dma_ring_ib_execute(struct radeon_device *rdev, + struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + + if (rdev->wb.enabled) { + u32 next_rptr = ring->wptr + 4; + while ((next_rptr & 7) != 5) + next_rptr++; + next_rptr += 3; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 1)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); + radeon_ring_write(ring, next_rptr); + } + + /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. + * Pad as necessary with NOPs. + */ + while ((ring->wptr & 7) != 5) + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0)); + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_INDIRECT_BUFFER, 0, 0)); + radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); + radeon_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF)); + +} + +/** + * evergreen_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (evergreen-cayman). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int evergreen_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_dw, cur_size_in_dw; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; + num_loops = DIV_ROUND_UP(size_in_dw, 0xfffff); + r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_dw = size_in_dw; + if (cur_size_in_dw > 0xFFFFF) + cur_size_in_dw = 0xFFFFF; + size_in_dw -= cur_size_in_dw; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, cur_size_in_dw)); + radeon_ring_write(ring, dst_offset & 0xfffffffc); + radeon_ring_write(ring, src_offset & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); + radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); + src_offset += cur_size_in_dw * 4; + dst_offset += cur_size_in_dw * 4; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} + +/** + * evergreen_dma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up. + * Returns true if the engine appears to be locked up, false if not. + */ +bool evergreen_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = evergreen_gpu_check_soft_reset(rdev); + + if (!(reset_mask & RADEON_RESET_DMA)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + + diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 0205fa1594fa..2db8ce0023ac 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -174,6 +174,11 @@ extern void evergreen_pcie_gen2_enable(struct radeon_device *rdev); extern void evergreen_program_aspm(struct radeon_device *rdev); extern void sumo_rlc_fini(struct radeon_device *rdev); extern int sumo_rlc_init(struct radeon_device *rdev); +extern void cayman_dma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags); /* Firmware Names */ MODULE_FIRMWARE("radeon/BARTS_pfp.bin"); @@ -1595,186 +1600,7 @@ static int cayman_cp_resume(struct radeon_device *rdev) return 0; } -/* - * DMA - * Starting with R600, the GPU has an asynchronous - * DMA engine. The programming model is very similar - * to the 3D engine (ring buffer, IBs, etc.), but the - * DMA controller has it's own packet format that is - * different form the PM4 format used by the 3D engine. - * It supports copying data, writing embedded data, - * solid fills, and a number of other things. It also - * has support for tiling/detiling of buffers. - * Cayman and newer support two asynchronous DMA engines. - */ -/** - * cayman_dma_ring_ib_execute - Schedule an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ib: IB object to schedule - * - * Schedule an IB in the DMA ring (cayman-SI). - */ -void cayman_dma_ring_ib_execute(struct radeon_device *rdev, - struct radeon_ib *ib) -{ - struct radeon_ring *ring = &rdev->ring[ib->ring]; - - if (rdev->wb.enabled) { - u32 next_rptr = ring->wptr + 4; - while ((next_rptr & 7) != 5) - next_rptr++; - next_rptr += 3; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); - radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); - radeon_ring_write(ring, next_rptr); - } - - /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. - * Pad as necessary with NOPs. - */ - while ((ring->wptr & 7) != 5) - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); - radeon_ring_write(ring, DMA_IB_PACKET(DMA_PACKET_INDIRECT_BUFFER, ib->vm ? ib->vm->id : 0, 0)); - radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); - radeon_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF)); - -} - -/** - * cayman_dma_stop - stop the async dma engines - * - * @rdev: radeon_device pointer - * - * Stop the async dma engines (cayman-SI). - */ -void cayman_dma_stop(struct radeon_device *rdev) -{ - u32 rb_cntl; - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); - - /* dma0 */ - rb_cntl = RREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET); - rb_cntl &= ~DMA_RB_ENABLE; - WREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET, rb_cntl); - - /* dma1 */ - rb_cntl = RREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET); - rb_cntl &= ~DMA_RB_ENABLE; - WREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET, rb_cntl); - - rdev->ring[R600_RING_TYPE_DMA_INDEX].ready = false; - rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX].ready = false; -} - -/** - * cayman_dma_resume - setup and start the async dma engines - * - * @rdev: radeon_device pointer - * - * Set up the DMA ring buffers and enable them. (cayman-SI). - * Returns 0 for success, error for failure. - */ -int cayman_dma_resume(struct radeon_device *rdev) -{ - struct radeon_ring *ring; - u32 rb_cntl, dma_cntl, ib_cntl; - u32 rb_bufsz; - u32 reg_offset, wb_offset; - int i, r; - - /* Reset dma */ - WREG32(SRBM_SOFT_RESET, SOFT_RESET_DMA | SOFT_RESET_DMA1); - RREG32(SRBM_SOFT_RESET); - udelay(50); - WREG32(SRBM_SOFT_RESET, 0); - - for (i = 0; i < 2; i++) { - if (i == 0) { - ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; - reg_offset = DMA0_REGISTER_OFFSET; - wb_offset = R600_WB_DMA_RPTR_OFFSET; - } else { - ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; - reg_offset = DMA1_REGISTER_OFFSET; - wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET; - } - - WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0); - WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0); - - /* Set ring buffer size in dwords */ - rb_bufsz = drm_order(ring->ring_size / 4); - rb_cntl = rb_bufsz << 1; -#ifdef __BIG_ENDIAN - rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE; -#endif - WREG32(DMA_RB_CNTL + reg_offset, rb_cntl); - - /* Initialize the ring buffer's read and write pointers */ - WREG32(DMA_RB_RPTR + reg_offset, 0); - WREG32(DMA_RB_WPTR + reg_offset, 0); - - /* set the wb address whether it's enabled or not */ - WREG32(DMA_RB_RPTR_ADDR_HI + reg_offset, - upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFF); - WREG32(DMA_RB_RPTR_ADDR_LO + reg_offset, - ((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); - - if (rdev->wb.enabled) - rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE; - - WREG32(DMA_RB_BASE + reg_offset, ring->gpu_addr >> 8); - - /* enable DMA IBs */ - ib_cntl = DMA_IB_ENABLE | CMD_VMID_FORCE; -#ifdef __BIG_ENDIAN - ib_cntl |= DMA_IB_SWAP_ENABLE; -#endif - WREG32(DMA_IB_CNTL + reg_offset, ib_cntl); - - dma_cntl = RREG32(DMA_CNTL + reg_offset); - dma_cntl &= ~CTXEMPTY_INT_ENABLE; - WREG32(DMA_CNTL + reg_offset, dma_cntl); - - ring->wptr = 0; - WREG32(DMA_RB_WPTR + reg_offset, ring->wptr << 2); - - ring->rptr = RREG32(DMA_RB_RPTR + reg_offset) >> 2; - - WREG32(DMA_RB_CNTL + reg_offset, rb_cntl | DMA_RB_ENABLE); - - ring->ready = true; - - r = radeon_ring_test(rdev, ring->idx, ring); - if (r) { - ring->ready = false; - return r; - } - } - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); - - return 0; -} - -/** - * cayman_dma_fini - tear down the async dma engines - * - * @rdev: radeon_device pointer - * - * Stop the async dma engines and free the rings (cayman-SI). - */ -void cayman_dma_fini(struct radeon_device *rdev) -{ - cayman_dma_stop(rdev); - radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); - radeon_ring_fini(rdev, &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]); -} - -static u32 cayman_gpu_check_soft_reset(struct radeon_device *rdev) +u32 cayman_gpu_check_soft_reset(struct radeon_device *rdev) { u32 reset_mask = 0; u32 tmp; @@ -2027,34 +1853,6 @@ bool cayman_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) return radeon_ring_test_lockup(rdev, ring); } -/** - * cayman_dma_is_lockup - Check if the DMA engine is locked up - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Check if the async DMA engine is locked up. - * Returns true if the engine appears to be locked up, false if not. - */ -bool cayman_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 reset_mask = cayman_gpu_check_soft_reset(rdev); - u32 mask; - - if (ring->idx == R600_RING_TYPE_DMA_INDEX) - mask = RADEON_RESET_DMA; - else - mask = RADEON_RESET_DMA1; - - if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - static int cayman_startup(struct radeon_device *rdev) { struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; @@ -2658,61 +2456,7 @@ void cayman_vm_set_page(struct radeon_device *rdev, } } } else { - if ((flags & RADEON_VM_PAGE_SYSTEM) || - (count == 1)) { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, ndw); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - if (flags & RADEON_VM_PAGE_SYSTEM) { - value = radeon_vm_map_gart(rdev, addr); - value &= 0xFFFFFFFFFFFFF000ULL; - } else if (flags & RADEON_VM_PAGE_VALID) { - value = addr; - } else { - value = 0; - } - addr += incr; - value |= r600_flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } - } - while (ib->length_dw & 0x7) - ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0); - } else { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - if (flags & RADEON_VM_PAGE_VALID) - value = addr; - else - value = 0; - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; - ib->ptr[ib->length_dw++] = r600_flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - pe += ndw * 4; - addr += (ndw / 2) * incr; - count -= ndw / 2; - } - } - while (ib->length_dw & 0x7) - ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0); + cayman_dma_vm_set_page(rdev, ib, pe, addr, count, incr, flags); } } @@ -2746,26 +2490,3 @@ void cayman_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)); radeon_ring_write(ring, 0x0); } - -void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) -{ - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); - - /* flush hdp cache */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); - radeon_ring_write(ring, 1); - - /* bits 0-7 are the VM contexts0-7 */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); - radeon_ring_write(ring, 1 << vm->id); -} - diff --git a/drivers/gpu/drm/radeon/ni_dma.c b/drivers/gpu/drm/radeon/ni_dma.c new file mode 100644 index 000000000000..0f3c0baea4a6 --- /dev/null +++ b/drivers/gpu/drm/radeon/ni_dma.c @@ -0,0 +1,338 @@ +/* + * Copyright 2010 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "nid.h" + +u32 cayman_gpu_check_soft_reset(struct radeon_device *rdev); + +/* + * DMA + * Starting with R600, the GPU has an asynchronous + * DMA engine. The programming model is very similar + * to the 3D engine (ring buffer, IBs, etc.), but the + * DMA controller has it's own packet format that is + * different form the PM4 format used by the 3D engine. + * It supports copying data, writing embedded data, + * solid fills, and a number of other things. It also + * has support for tiling/detiling of buffers. + * Cayman and newer support two asynchronous DMA engines. + */ + +/** + * cayman_dma_ring_ib_execute - Schedule an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring (cayman-SI). + */ +void cayman_dma_ring_ib_execute(struct radeon_device *rdev, + struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + + if (rdev->wb.enabled) { + u32 next_rptr = ring->wptr + 4; + while ((next_rptr & 7) != 5) + next_rptr++; + next_rptr += 3; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); + radeon_ring_write(ring, next_rptr); + } + + /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. + * Pad as necessary with NOPs. + */ + while ((ring->wptr & 7) != 5) + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + radeon_ring_write(ring, DMA_IB_PACKET(DMA_PACKET_INDIRECT_BUFFER, ib->vm ? ib->vm->id : 0, 0)); + radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); + radeon_ring_write(ring, (ib->length_dw << 12) | (upper_32_bits(ib->gpu_addr) & 0xFF)); + +} + +/** + * cayman_dma_stop - stop the async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the async dma engines (cayman-SI). + */ +void cayman_dma_stop(struct radeon_device *rdev) +{ + u32 rb_cntl; + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); + + /* dma0 */ + rb_cntl = RREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET); + rb_cntl &= ~DMA_RB_ENABLE; + WREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET, rb_cntl); + + /* dma1 */ + rb_cntl = RREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET); + rb_cntl &= ~DMA_RB_ENABLE; + WREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET, rb_cntl); + + rdev->ring[R600_RING_TYPE_DMA_INDEX].ready = false; + rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX].ready = false; +} + +/** + * cayman_dma_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the DMA ring buffers and enable them. (cayman-SI). + * Returns 0 for success, error for failure. + */ +int cayman_dma_resume(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + u32 rb_cntl, dma_cntl, ib_cntl; + u32 rb_bufsz; + u32 reg_offset, wb_offset; + int i, r; + + /* Reset dma */ + WREG32(SRBM_SOFT_RESET, SOFT_RESET_DMA | SOFT_RESET_DMA1); + RREG32(SRBM_SOFT_RESET); + udelay(50); + WREG32(SRBM_SOFT_RESET, 0); + + for (i = 0; i < 2; i++) { + if (i == 0) { + ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + reg_offset = DMA0_REGISTER_OFFSET; + wb_offset = R600_WB_DMA_RPTR_OFFSET; + } else { + ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; + reg_offset = DMA1_REGISTER_OFFSET; + wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET; + } + + WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0); + WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0); + + /* Set ring buffer size in dwords */ + rb_bufsz = drm_order(ring->ring_size / 4); + rb_cntl = rb_bufsz << 1; +#ifdef __BIG_ENDIAN + rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE; +#endif + WREG32(DMA_RB_CNTL + reg_offset, rb_cntl); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(DMA_RB_RPTR + reg_offset, 0); + WREG32(DMA_RB_WPTR + reg_offset, 0); + + /* set the wb address whether it's enabled or not */ + WREG32(DMA_RB_RPTR_ADDR_HI + reg_offset, + upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFF); + WREG32(DMA_RB_RPTR_ADDR_LO + reg_offset, + ((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); + + if (rdev->wb.enabled) + rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE; + + WREG32(DMA_RB_BASE + reg_offset, ring->gpu_addr >> 8); + + /* enable DMA IBs */ + ib_cntl = DMA_IB_ENABLE | CMD_VMID_FORCE; +#ifdef __BIG_ENDIAN + ib_cntl |= DMA_IB_SWAP_ENABLE; +#endif + WREG32(DMA_IB_CNTL + reg_offset, ib_cntl); + + dma_cntl = RREG32(DMA_CNTL + reg_offset); + dma_cntl &= ~CTXEMPTY_INT_ENABLE; + WREG32(DMA_CNTL + reg_offset, dma_cntl); + + ring->wptr = 0; + WREG32(DMA_RB_WPTR + reg_offset, ring->wptr << 2); + + ring->rptr = RREG32(DMA_RB_RPTR + reg_offset) >> 2; + + WREG32(DMA_RB_CNTL + reg_offset, rb_cntl | DMA_RB_ENABLE); + + ring->ready = true; + + r = radeon_ring_test(rdev, ring->idx, ring); + if (r) { + ring->ready = false; + return r; + } + } + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); + + return 0; +} + +/** + * cayman_dma_fini - tear down the async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the async dma engines and free the rings (cayman-SI). + */ +void cayman_dma_fini(struct radeon_device *rdev) +{ + cayman_dma_stop(rdev); + radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); + radeon_ring_fini(rdev, &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]); +} + +/** + * cayman_dma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up. + * Returns true if the engine appears to be locked up, false if not. + */ +bool cayman_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = cayman_gpu_check_soft_reset(rdev); + u32 mask; + + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + mask = RADEON_RESET_DMA; + else + mask = RADEON_RESET_DMA1; + + if (!(reset_mask & mask)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + +/** + * cayman_dma_vm_set_page - update the page tables using the DMA + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * @r600_flags: hw access flags + * + * Update the page tables using the DMA (cayman/TN). + */ +void cayman_dma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) +{ + uint32_t r600_flags = cayman_vm_page_flags(rdev, flags); + uint64_t value; + unsigned ndw; + + if ((flags & RADEON_VM_PAGE_SYSTEM) || (count == 1)) { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + /* for non-physically contiguous pages (system) */ + ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, ndw); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + if (flags & RADEON_VM_PAGE_SYSTEM) { + value = radeon_vm_map_gart(rdev, addr); + value &= 0xFFFFFFFFFFFFF000ULL; + } else if (flags & RADEON_VM_PAGE_VALID) { + value = addr; + } else { + value = 0; + } + addr += incr; + value |= r600_flags; + ib->ptr[ib->length_dw++] = value; + ib->ptr[ib->length_dw++] = upper_32_bits(value); + } + } + } else { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + if (flags & RADEON_VM_PAGE_VALID) + value = addr; + else + value = 0; + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw); + ib->ptr[ib->length_dw++] = pe; /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; + ib->ptr[ib->length_dw++] = r600_flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = value; /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(value); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + pe += ndw * 4; + addr += (ndw / 2) * incr; + count -= ndw / 2; + } + } + while (ib->length_dw & 0x7) + ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0); +} + +void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +{ + struct radeon_ring *ring = &rdev->ring[ridx]; + + if (vm == NULL) + return; + + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); + radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + + /* flush hdp cache */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); + radeon_ring_write(ring, 1); + + /* bits 0-7 are the VM contexts0-7 */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); + radeon_ring_write(ring, 1 << vm->id); +} + diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 3a08ef92d33f..087cff444ba2 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -1374,7 +1374,7 @@ static bool r600_is_display_hung(struct radeon_device *rdev) return true; } -static u32 r600_gpu_check_soft_reset(struct radeon_device *rdev) +u32 r600_gpu_check_soft_reset(struct radeon_device *rdev) { u32 reset_mask = 0; u32 tmp; @@ -1622,28 +1622,6 @@ bool r600_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) return radeon_ring_test_lockup(rdev, ring); } -/** - * r600_dma_is_lockup - Check if the DMA engine is locked up - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Check if the async DMA engine is locked up. - * Returns true if the engine appears to be locked up, false if not. - */ -bool r600_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 reset_mask = r600_gpu_check_soft_reset(rdev); - - if (!(reset_mask & RADEON_RESET_DMA)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - u32 r6xx_remap_render_backend(struct radeon_device *rdev, u32 tiling_pipe_num, u32 max_rb_num, @@ -2493,176 +2471,6 @@ void r600_cp_fini(struct radeon_device *rdev) radeon_scratch_free(rdev, ring->rptr_save_reg); } -/* - * DMA - * Starting with R600, the GPU has an asynchronous - * DMA engine. The programming model is very similar - * to the 3D engine (ring buffer, IBs, etc.), but the - * DMA controller has it's own packet format that is - * different form the PM4 format used by the 3D engine. - * It supports copying data, writing embedded data, - * solid fills, and a number of other things. It also - * has support for tiling/detiling of buffers. - */ - -/** - * r600_dma_get_rptr - get the current read pointer - * - * @rdev: radeon_device pointer - * @ring: radeon ring pointer - * - * Get the current rptr from the hardware (r6xx+). - */ -uint32_t r600_dma_get_rptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - return (radeon_ring_generic_get_rptr(rdev, ring) & 0x3fffc) >> 2; -} - -/** - * r600_dma_get_wptr - get the current write pointer - * - * @rdev: radeon_device pointer - * @ring: radeon ring pointer - * - * Get the current wptr from the hardware (r6xx+). - */ -uint32_t r600_dma_get_wptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - return (RREG32(ring->wptr_reg) & 0x3fffc) >> 2; -} - -/** - * r600_dma_set_wptr - commit the write pointer - * - * @rdev: radeon_device pointer - * @ring: radeon ring pointer - * - * Write the wptr back to the hardware (r6xx+). - */ -void r600_dma_set_wptr(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - WREG32(ring->wptr_reg, (ring->wptr << 2) & 0x3fffc); -} - -/** - * r600_dma_stop - stop the async dma engine - * - * @rdev: radeon_device pointer - * - * Stop the async dma engine (r6xx-evergreen). - */ -void r600_dma_stop(struct radeon_device *rdev) -{ - u32 rb_cntl = RREG32(DMA_RB_CNTL); - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); - - rb_cntl &= ~DMA_RB_ENABLE; - WREG32(DMA_RB_CNTL, rb_cntl); - - rdev->ring[R600_RING_TYPE_DMA_INDEX].ready = false; -} - -/** - * r600_dma_resume - setup and start the async dma engine - * - * @rdev: radeon_device pointer - * - * Set up the DMA ring buffer and enable it. (r6xx-evergreen). - * Returns 0 for success, error for failure. - */ -int r600_dma_resume(struct radeon_device *rdev) -{ - struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; - u32 rb_cntl, dma_cntl, ib_cntl; - u32 rb_bufsz; - int r; - - /* Reset dma */ - if (rdev->family >= CHIP_RV770) - WREG32(SRBM_SOFT_RESET, RV770_SOFT_RESET_DMA); - else - WREG32(SRBM_SOFT_RESET, SOFT_RESET_DMA); - RREG32(SRBM_SOFT_RESET); - udelay(50); - WREG32(SRBM_SOFT_RESET, 0); - - WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL, 0); - WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL, 0); - - /* Set ring buffer size in dwords */ - rb_bufsz = drm_order(ring->ring_size / 4); - rb_cntl = rb_bufsz << 1; -#ifdef __BIG_ENDIAN - rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE; -#endif - WREG32(DMA_RB_CNTL, rb_cntl); - - /* Initialize the ring buffer's read and write pointers */ - WREG32(DMA_RB_RPTR, 0); - WREG32(DMA_RB_WPTR, 0); - - /* set the wb address whether it's enabled or not */ - WREG32(DMA_RB_RPTR_ADDR_HI, - upper_32_bits(rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFF); - WREG32(DMA_RB_RPTR_ADDR_LO, - ((rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFFFFFFFC)); - - if (rdev->wb.enabled) - rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE; - - WREG32(DMA_RB_BASE, ring->gpu_addr >> 8); - - /* enable DMA IBs */ - ib_cntl = DMA_IB_ENABLE; -#ifdef __BIG_ENDIAN - ib_cntl |= DMA_IB_SWAP_ENABLE; -#endif - WREG32(DMA_IB_CNTL, ib_cntl); - - dma_cntl = RREG32(DMA_CNTL); - dma_cntl &= ~CTXEMPTY_INT_ENABLE; - WREG32(DMA_CNTL, dma_cntl); - - if (rdev->family >= CHIP_RV770) - WREG32(DMA_MODE, 1); - - ring->wptr = 0; - WREG32(DMA_RB_WPTR, ring->wptr << 2); - - ring->rptr = RREG32(DMA_RB_RPTR) >> 2; - - WREG32(DMA_RB_CNTL, rb_cntl | DMA_RB_ENABLE); - - ring->ready = true; - - r = radeon_ring_test(rdev, R600_RING_TYPE_DMA_INDEX, ring); - if (r) { - ring->ready = false; - return r; - } - - radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); - - return 0; -} - -/** - * r600_dma_fini - tear down the async dma engine - * - * @rdev: radeon_device pointer - * - * Stop the async dma engine and free the ring (r6xx-evergreen). - */ -void r600_dma_fini(struct radeon_device *rdev) -{ - r600_dma_stop(rdev); - radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); -} - /* * GPU scratch registers helpers function. */ @@ -2718,60 +2526,6 @@ int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) return r; } -/** - * r600_dma_ring_test - simple async dma engine test - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Test the DMA engine by writing using it to write an - * value to memory. (r6xx-SI). - * Returns 0 for success, error for failure. - */ -int r600_dma_ring_test(struct radeon_device *rdev, - struct radeon_ring *ring) -{ - unsigned i; - int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; - u32 tmp; - - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } - - tmp = 0xCAFEDEAD; - writel(tmp, ptr); - - r = radeon_ring_lock(rdev, ring, 4); - if (r) { - DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r); - return r; - } - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); - radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff); - radeon_ring_write(ring, 0xDEADBEEF); - radeon_ring_unlock_commit(rdev, ring); - - for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - - if (i < rdev->usec_timeout) { - DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); - } else { - DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", - ring->idx, tmp); - r = -EINVAL; - } - return r; -} - /* * CP fences/semaphores */ @@ -2839,59 +2593,6 @@ void r600_semaphore_ring_emit(struct radeon_device *rdev, radeon_ring_write(ring, (upper_32_bits(addr) & 0xff) | sel); } -/* - * DMA fences/semaphores - */ - -/** - * r600_dma_fence_ring_emit - emit a fence on the DMA ring - * - * @rdev: radeon_device pointer - * @fence: radeon fence object - * - * Add a DMA fence packet to the ring to write - * the fence seq number and DMA trap packet to generate - * an interrupt if needed (r6xx-r7xx). - */ -void r600_dma_fence_ring_emit(struct radeon_device *rdev, - struct radeon_fence *fence) -{ - struct radeon_ring *ring = &rdev->ring[fence->ring]; - u64 addr = rdev->fence_drv[fence->ring].gpu_addr; - - /* write the fence */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0, 0)); - radeon_ring_write(ring, addr & 0xfffffffc); - radeon_ring_write(ring, (upper_32_bits(addr) & 0xff)); - radeon_ring_write(ring, lower_32_bits(fence->seq)); - /* generate an interrupt */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0, 0)); -} - -/** - * r600_dma_semaphore_ring_emit - emit a semaphore on the dma ring - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * @semaphore: radeon semaphore object - * @emit_wait: wait or signal semaphore - * - * Add a DMA semaphore packet to the ring wait on or signal - * other rings (r6xx-SI). - */ -void r600_dma_semaphore_ring_emit(struct radeon_device *rdev, - struct radeon_ring *ring, - struct radeon_semaphore *semaphore, - bool emit_wait) -{ - u64 addr = semaphore->gpu_addr; - u32 s = emit_wait ? 0 : 1; - - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SEMAPHORE, 0, s, 0)); - radeon_ring_write(ring, addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(addr) & 0xff); -} - /** * r600_copy_cpdma - copy pages using the CP DMA engine * @@ -2976,80 +2677,6 @@ int r600_copy_cpdma(struct radeon_device *rdev, return r; } -/** - * r600_copy_dma - copy pages using the DMA engine - * - * @rdev: radeon_device pointer - * @src_offset: src GPU address - * @dst_offset: dst GPU address - * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object - * - * Copy GPU paging using the DMA engine (r6xx). - * Used by the radeon ttm implementation to move pages if - * registered as the asic copy callback. - */ -int r600_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - int ring_index = rdev->asic->copy.dma_ring_index; - struct radeon_ring *ring = &rdev->ring[ring_index]; - u32 size_in_dw, cur_size_in_dw; - int i, num_loops; - int r = 0; - - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; - } - - size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; - num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFE); - r = radeon_ring_lock(rdev, ring, num_loops * 4 + 8); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, ring->idx)) { - radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, - ring->idx); - radeon_fence_note_sync(*fence, ring->idx); - } else { - radeon_semaphore_free(rdev, &sem, NULL); - } - - for (i = 0; i < num_loops; i++) { - cur_size_in_dw = size_in_dw; - if (cur_size_in_dw > 0xFFFE) - cur_size_in_dw = 0xFFFE; - size_in_dw -= cur_size_in_dw; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw)); - radeon_ring_write(ring, dst_offset & 0xfffffffc); - radeon_ring_write(ring, src_offset & 0xfffffffc); - radeon_ring_write(ring, (((upper_32_bits(dst_offset) & 0xff) << 16) | - (upper_32_bits(src_offset) & 0xff))); - src_offset += cur_size_in_dw * 4; - dst_offset += cur_size_in_dw * 4; - } - - r = radeon_fence_emit(rdev, fence, ring->idx); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return r; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); - - return r; -} - int r600_set_surface_reg(struct radeon_device *rdev, int reg, uint32_t tiling_flags, uint32_t pitch, uint32_t offset, uint32_t obj_size) @@ -3409,104 +3036,6 @@ free_scratch: return r; } -/** - * r600_dma_ib_test - test an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Test a simple IB in the DMA ring (r6xx-SI). - * Returns 0 on success, error on failure. - */ -int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) -{ - struct radeon_ib ib; - unsigned i; - int r; - void __iomem *ptr = (void *)rdev->vram_scratch.ptr; - u32 tmp = 0; - - if (!ptr) { - DRM_ERROR("invalid vram scratch pointer\n"); - return -EINVAL; - } - - tmp = 0xCAFEDEAD; - writel(tmp, ptr); - - r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); - if (r) { - DRM_ERROR("radeon: failed to get ib (%d).\n", r); - return r; - } - - ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1); - ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; - ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff; - ib.ptr[3] = 0xDEADBEEF; - ib.length_dw = 4; - - r = radeon_ib_schedule(rdev, &ib, NULL); - if (r) { - radeon_ib_free(rdev, &ib); - DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); - return r; - } - r = radeon_fence_wait(ib.fence, false); - if (r) { - DRM_ERROR("radeon: fence wait failed (%d).\n", r); - return r; - } - for (i = 0; i < rdev->usec_timeout; i++) { - tmp = readl(ptr); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i < rdev->usec_timeout) { - DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); - } else { - DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp); - r = -EINVAL; - } - radeon_ib_free(rdev, &ib); - return r; -} - -/** - * r600_dma_ring_ib_execute - Schedule an IB on the DMA engine - * - * @rdev: radeon_device pointer - * @ib: IB object to schedule - * - * Schedule an IB in the DMA ring (r6xx-r7xx). - */ -void r600_dma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) -{ - struct radeon_ring *ring = &rdev->ring[ib->ring]; - - if (rdev->wb.enabled) { - u32 next_rptr = ring->wptr + 4; - while ((next_rptr & 7) != 5) - next_rptr++; - next_rptr += 3; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); - radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); - radeon_ring_write(ring, next_rptr); - } - - /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. - * Pad as necessary with NOPs. - */ - while ((ring->wptr & 7) != 5) - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_INDIRECT_BUFFER, 0, 0, 0)); - radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); - radeon_ring_write(ring, (ib->length_dw << 16) | (upper_32_bits(ib->gpu_addr) & 0xFF)); - -} - /* * Interrupts * diff --git a/drivers/gpu/drm/radeon/r600_dma.c b/drivers/gpu/drm/radeon/r600_dma.c new file mode 100644 index 000000000000..bff05576266b --- /dev/null +++ b/drivers/gpu/drm/radeon/r600_dma.c @@ -0,0 +1,497 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "r600d.h" + +u32 r600_gpu_check_soft_reset(struct radeon_device *rdev); + +/* + * DMA + * Starting with R600, the GPU has an asynchronous + * DMA engine. The programming model is very similar + * to the 3D engine (ring buffer, IBs, etc.), but the + * DMA controller has it's own packet format that is + * different form the PM4 format used by the 3D engine. + * It supports copying data, writing embedded data, + * solid fills, and a number of other things. It also + * has support for tiling/detiling of buffers. + */ + +/** + * r600_dma_get_rptr - get the current read pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Get the current rptr from the hardware (r6xx+). + */ +uint32_t r600_dma_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return (radeon_ring_generic_get_rptr(rdev, ring) & 0x3fffc) >> 2; +} + +/** + * r600_dma_get_wptr - get the current write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Get the current wptr from the hardware (r6xx+). + */ +uint32_t r600_dma_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + return (RREG32(ring->wptr_reg) & 0x3fffc) >> 2; +} + +/** + * r600_dma_set_wptr - commit the write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon ring pointer + * + * Write the wptr back to the hardware (r6xx+). + */ +void r600_dma_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + WREG32(ring->wptr_reg, (ring->wptr << 2) & 0x3fffc); +} + +/** + * r600_dma_stop - stop the async dma engine + * + * @rdev: radeon_device pointer + * + * Stop the async dma engine (r6xx-evergreen). + */ +void r600_dma_stop(struct radeon_device *rdev) +{ + u32 rb_cntl = RREG32(DMA_RB_CNTL); + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); + + rb_cntl &= ~DMA_RB_ENABLE; + WREG32(DMA_RB_CNTL, rb_cntl); + + rdev->ring[R600_RING_TYPE_DMA_INDEX].ready = false; +} + +/** + * r600_dma_resume - setup and start the async dma engine + * + * @rdev: radeon_device pointer + * + * Set up the DMA ring buffer and enable it. (r6xx-evergreen). + * Returns 0 for success, error for failure. + */ +int r600_dma_resume(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + u32 rb_cntl, dma_cntl, ib_cntl; + u32 rb_bufsz; + int r; + + /* Reset dma */ + if (rdev->family >= CHIP_RV770) + WREG32(SRBM_SOFT_RESET, RV770_SOFT_RESET_DMA); + else + WREG32(SRBM_SOFT_RESET, SOFT_RESET_DMA); + RREG32(SRBM_SOFT_RESET); + udelay(50); + WREG32(SRBM_SOFT_RESET, 0); + + WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL, 0); + WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL, 0); + + /* Set ring buffer size in dwords */ + rb_bufsz = drm_order(ring->ring_size / 4); + rb_cntl = rb_bufsz << 1; +#ifdef __BIG_ENDIAN + rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE; +#endif + WREG32(DMA_RB_CNTL, rb_cntl); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(DMA_RB_RPTR, 0); + WREG32(DMA_RB_WPTR, 0); + + /* set the wb address whether it's enabled or not */ + WREG32(DMA_RB_RPTR_ADDR_HI, + upper_32_bits(rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFF); + WREG32(DMA_RB_RPTR_ADDR_LO, + ((rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFFFFFFFC)); + + if (rdev->wb.enabled) + rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE; + + WREG32(DMA_RB_BASE, ring->gpu_addr >> 8); + + /* enable DMA IBs */ + ib_cntl = DMA_IB_ENABLE; +#ifdef __BIG_ENDIAN + ib_cntl |= DMA_IB_SWAP_ENABLE; +#endif + WREG32(DMA_IB_CNTL, ib_cntl); + + dma_cntl = RREG32(DMA_CNTL); + dma_cntl &= ~CTXEMPTY_INT_ENABLE; + WREG32(DMA_CNTL, dma_cntl); + + if (rdev->family >= CHIP_RV770) + WREG32(DMA_MODE, 1); + + ring->wptr = 0; + WREG32(DMA_RB_WPTR, ring->wptr << 2); + + ring->rptr = RREG32(DMA_RB_RPTR) >> 2; + + WREG32(DMA_RB_CNTL, rb_cntl | DMA_RB_ENABLE); + + ring->ready = true; + + r = radeon_ring_test(rdev, R600_RING_TYPE_DMA_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); + + return 0; +} + +/** + * r600_dma_fini - tear down the async dma engine + * + * @rdev: radeon_device pointer + * + * Stop the async dma engine and free the ring (r6xx-evergreen). + */ +void r600_dma_fini(struct radeon_device *rdev) +{ + r600_dma_stop(rdev); + radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); +} + +/** + * r600_dma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up. + * Returns true if the engine appears to be locked up, false if not. + */ +bool r600_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = r600_gpu_check_soft_reset(rdev); + + if (!(reset_mask & RADEON_RESET_DMA)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + + +/** + * r600_dma_ring_test - simple async dma engine test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test the DMA engine by writing using it to write an + * value to memory. (r6xx-SI). + * Returns 0 for success, error for failure. + */ +int r600_dma_ring_test(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + unsigned i; + int r; + void __iomem *ptr = (void *)rdev->vram_scratch.ptr; + u32 tmp; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ring_lock(rdev, ring, 4); + if (r) { + DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r); + return r; + } + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); + radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff); + radeon_ring_write(ring, 0xDEADBEEF); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", + ring->idx, tmp); + r = -EINVAL; + } + return r; +} + +/** + * r600_dma_fence_ring_emit - emit a fence on the DMA ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Add a DMA fence packet to the ring to write + * the fence seq number and DMA trap packet to generate + * an interrupt if needed (r6xx-r7xx). + */ +void r600_dma_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + + /* write the fence */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_FENCE, 0, 0, 0)); + radeon_ring_write(ring, addr & 0xfffffffc); + radeon_ring_write(ring, (upper_32_bits(addr) & 0xff)); + radeon_ring_write(ring, lower_32_bits(fence->seq)); + /* generate an interrupt */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_TRAP, 0, 0, 0)); +} + +/** + * r600_dma_semaphore_ring_emit - emit a semaphore on the dma ring + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * @semaphore: radeon semaphore object + * @emit_wait: wait or signal semaphore + * + * Add a DMA semaphore packet to the ring wait on or signal + * other rings (r6xx-SI). + */ +void r600_dma_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + u64 addr = semaphore->gpu_addr; + u32 s = emit_wait ? 0 : 1; + + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SEMAPHORE, 0, s, 0)); + radeon_ring_write(ring, addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(addr) & 0xff); +} + +/** + * r600_dma_ib_test - test an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test a simple IB in the DMA ring (r6xx-SI). + * Returns 0 on success, error on failure. + */ +int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_ib ib; + unsigned i; + int r; + void __iomem *ptr = (void *)rdev->vram_scratch.ptr; + u32 tmp = 0; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1); + ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; + ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff; + ib.ptr[3] = 0xDEADBEEF; + ib.length_dw = 4; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + radeon_ib_free(rdev, &ib); + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + return r; + } + r = radeon_fence_wait(ib.fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + return r; + } + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); + } else { + DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp); + r = -EINVAL; + } + radeon_ib_free(rdev, &ib); + return r; +} + +/** + * r600_dma_ring_ib_execute - Schedule an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring (r6xx-r7xx). + */ +void r600_dma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + + if (rdev->wb.enabled) { + u32 next_rptr = ring->wptr + 4; + while ((next_rptr & 7) != 5) + next_rptr++; + next_rptr += 3; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 1)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xff); + radeon_ring_write(ring, next_rptr); + } + + /* The indirect buffer packet must end on an 8 DW boundary in the DMA ring. + * Pad as necessary with NOPs. + */ + while ((ring->wptr & 7) != 5) + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0)); + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_INDIRECT_BUFFER, 0, 0, 0)); + radeon_ring_write(ring, (ib->gpu_addr & 0xFFFFFFE0)); + radeon_ring_write(ring, (ib->length_dw << 16) | (upper_32_bits(ib->gpu_addr) & 0xFF)); + +} + +/** + * r600_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (r6xx). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int r600_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_dw, cur_size_in_dw; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; + num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFE); + r = radeon_ring_lock(rdev, ring, num_loops * 4 + 8); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_dw = size_in_dw; + if (cur_size_in_dw > 0xFFFE) + cur_size_in_dw = 0xFFFE; + size_in_dw -= cur_size_in_dw; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw)); + radeon_ring_write(ring, dst_offset & 0xfffffffc); + radeon_ring_write(ring, src_offset & 0xfffffffc); + radeon_ring_write(ring, (((upper_32_bits(dst_offset) & 0xff) << 16) | + (upper_32_bits(src_offset) & 0xff))); + src_offset += cur_size_in_dw * 4; + dst_offset += cur_size_in_dw * 4; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index aaab7b1bba27..b811296462a3 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1650,80 +1650,6 @@ static int rv770_mc_init(struct radeon_device *rdev) return 0; } -/** - * rv770_copy_dma - copy pages using the DMA engine - * - * @rdev: radeon_device pointer - * @src_offset: src GPU address - * @dst_offset: dst GPU address - * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object - * - * Copy GPU paging using the DMA engine (r7xx). - * Used by the radeon ttm implementation to move pages if - * registered as the asic copy callback. - */ -int rv770_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - int ring_index = rdev->asic->copy.dma_ring_index; - struct radeon_ring *ring = &rdev->ring[ring_index]; - u32 size_in_dw, cur_size_in_dw; - int i, num_loops; - int r = 0; - - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; - } - - size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; - num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFF); - r = radeon_ring_lock(rdev, ring, num_loops * 5 + 8); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, ring->idx)) { - radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, - ring->idx); - radeon_fence_note_sync(*fence, ring->idx); - } else { - radeon_semaphore_free(rdev, &sem, NULL); - } - - for (i = 0; i < num_loops; i++) { - cur_size_in_dw = size_in_dw; - if (cur_size_in_dw > 0xFFFF) - cur_size_in_dw = 0xFFFF; - size_in_dw -= cur_size_in_dw; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw)); - radeon_ring_write(ring, dst_offset & 0xfffffffc); - radeon_ring_write(ring, src_offset & 0xfffffffc); - radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); - radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); - src_offset += cur_size_in_dw * 4; - dst_offset += cur_size_in_dw * 4; - } - - r = radeon_fence_emit(rdev, fence, ring->idx); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return r; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); - - return r; -} - static int rv770_startup(struct radeon_device *rdev) { struct radeon_ring *ring; diff --git a/drivers/gpu/drm/radeon/rv770_dma.c b/drivers/gpu/drm/radeon/rv770_dma.c new file mode 100644 index 000000000000..f9b02e3d6830 --- /dev/null +++ b/drivers/gpu/drm/radeon/rv770_dma.c @@ -0,0 +1,101 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "rv770d.h" + +/** + * rv770_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (r7xx). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int rv770_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_dw, cur_size_in_dw; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4; + num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFF); + r = radeon_ring_lock(rdev, ring, num_loops * 5 + 8); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_dw = size_in_dw; + if (cur_size_in_dw > 0xFFFF) + cur_size_in_dw = 0xFFFF; + size_in_dw -= cur_size_in_dw; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw)); + radeon_ring_write(ring, dst_offset & 0xfffffffc); + radeon_ring_write(ring, src_offset & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); + radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); + src_offset += cur_size_in_dw * 4; + dst_offset += cur_size_in_dw * 4; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index f3f79089405e..f5307e6bb92b 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -78,6 +78,11 @@ extern void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_ extern u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev); extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev); extern bool evergreen_is_display_hung(struct radeon_device *rdev); +extern void si_dma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags); static const u32 verde_rlc_save_restore_register_list[] = { @@ -3495,7 +3500,7 @@ static int si_cp_resume(struct radeon_device *rdev) return 0; } -static u32 si_gpu_check_soft_reset(struct radeon_device *rdev) +u32 si_gpu_check_soft_reset(struct radeon_device *rdev) { u32 reset_mask = 0; u32 tmp; @@ -3744,34 +3749,6 @@ bool si_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) return radeon_ring_test_lockup(rdev, ring); } -/** - * si_dma_is_lockup - Check if the DMA engine is locked up - * - * @rdev: radeon_device pointer - * @ring: radeon_ring structure holding ring information - * - * Check if the async DMA engine is locked up. - * Returns true if the engine appears to be locked up, false if not. - */ -bool si_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) -{ - u32 reset_mask = si_gpu_check_soft_reset(rdev); - u32 mask; - - if (ring->idx == R600_RING_TYPE_DMA_INDEX) - mask = RADEON_RESET_DMA; - else - mask = RADEON_RESET_DMA1; - - if (!(reset_mask & mask)) { - radeon_ring_lockup_update(ring); - return false; - } - /* force ring activities */ - radeon_ring_force_activity(rdev, ring); - return radeon_ring_test_lockup(rdev, ring); -} - /* MC */ static void si_mc_program(struct radeon_device *rdev) { @@ -4710,58 +4687,7 @@ void si_vm_set_page(struct radeon_device *rdev, } } else { /* DMA */ - if (flags & RADEON_VM_PAGE_SYSTEM) { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - /* for non-physically contiguous pages (system) */ - ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, ndw); - ib->ptr[ib->length_dw++] = pe; - ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; - for (; ndw > 0; ndw -= 2, --count, pe += 8) { - if (flags & RADEON_VM_PAGE_SYSTEM) { - value = radeon_vm_map_gart(rdev, addr); - value &= 0xFFFFFFFFFFFFF000ULL; - } else if (flags & RADEON_VM_PAGE_VALID) { - value = addr; - } else { - value = 0; - } - addr += incr; - value |= r600_flags; - ib->ptr[ib->length_dw++] = value; - ib->ptr[ib->length_dw++] = upper_32_bits(value); - } - } - } else { - while (count) { - ndw = count * 2; - if (ndw > 0xFFFFE) - ndw = 0xFFFFE; - - if (flags & RADEON_VM_PAGE_VALID) - value = addr; - else - value = 0; - /* for physically contiguous pages (vram) */ - ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw); - ib->ptr[ib->length_dw++] = pe; /* dst addr */ - ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; - ib->ptr[ib->length_dw++] = r600_flags; /* mask */ - ib->ptr[ib->length_dw++] = 0; - ib->ptr[ib->length_dw++] = value; /* value */ - ib->ptr[ib->length_dw++] = upper_32_bits(value); - ib->ptr[ib->length_dw++] = incr; /* increment size */ - ib->ptr[ib->length_dw++] = 0; - pe += ndw * 4; - addr += (ndw / 2) * incr; - count -= ndw / 2; - } - } - while (ib->length_dw & 0x7) - ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0); + si_dma_vm_set_page(rdev, ib, pe, addr, count, incr, flags); } } @@ -4808,32 +4734,6 @@ void si_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) radeon_ring_write(ring, 0x0); } -void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) -{ - struct radeon_ring *ring = &rdev->ring[ridx]; - - if (vm == NULL) - return; - - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - if (vm->id < 8) { - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); - } else { - radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2)); - } - radeon_ring_write(ring, vm->pd_gpu_addr >> 12); - - /* flush hdp cache */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); - radeon_ring_write(ring, 1); - - /* bits 0-7 are the VM contexts0-7 */ - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); - radeon_ring_write(ring, 1 << vm->id); -} - /* * Power and clock gating */ @@ -6177,80 +6077,6 @@ restart_ih: return IRQ_HANDLED; } -/** - * si_copy_dma - copy pages using the DMA engine - * - * @rdev: radeon_device pointer - * @src_offset: src GPU address - * @dst_offset: dst GPU address - * @num_gpu_pages: number of GPU pages to xfer - * @fence: radeon fence object - * - * Copy GPU paging using the DMA engine (SI). - * Used by the radeon ttm implementation to move pages if - * registered as the asic copy callback. - */ -int si_copy_dma(struct radeon_device *rdev, - uint64_t src_offset, uint64_t dst_offset, - unsigned num_gpu_pages, - struct radeon_fence **fence) -{ - struct radeon_semaphore *sem = NULL; - int ring_index = rdev->asic->copy.dma_ring_index; - struct radeon_ring *ring = &rdev->ring[ring_index]; - u32 size_in_bytes, cur_size_in_bytes; - int i, num_loops; - int r = 0; - - r = radeon_semaphore_create(rdev, &sem); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - return r; - } - - size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); - num_loops = DIV_ROUND_UP(size_in_bytes, 0xfffff); - r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); - if (r) { - DRM_ERROR("radeon: moving bo (%d).\n", r); - radeon_semaphore_free(rdev, &sem, NULL); - return r; - } - - if (radeon_fence_need_sync(*fence, ring->idx)) { - radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, - ring->idx); - radeon_fence_note_sync(*fence, ring->idx); - } else { - radeon_semaphore_free(rdev, &sem, NULL); - } - - for (i = 0; i < num_loops; i++) { - cur_size_in_bytes = size_in_bytes; - if (cur_size_in_bytes > 0xFFFFF) - cur_size_in_bytes = 0xFFFFF; - size_in_bytes -= cur_size_in_bytes; - radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 1, 0, 0, cur_size_in_bytes)); - radeon_ring_write(ring, dst_offset & 0xffffffff); - radeon_ring_write(ring, src_offset & 0xffffffff); - radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); - radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); - src_offset += cur_size_in_bytes; - dst_offset += cur_size_in_bytes; - } - - r = radeon_fence_emit(rdev, fence, ring->idx); - if (r) { - radeon_ring_unlock_undo(rdev, ring); - return r; - } - - radeon_ring_unlock_commit(rdev, ring); - radeon_semaphore_free(rdev, &sem, *fence); - - return r; -} - /* * startup/shutdown callbacks */ diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c new file mode 100644 index 000000000000..49909d23dfce --- /dev/null +++ b/drivers/gpu/drm/radeon/si_dma.c @@ -0,0 +1,235 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" + +u32 si_gpu_check_soft_reset(struct radeon_device *rdev); + +/** + * si_dma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up. + * Returns true if the engine appears to be locked up, false if not. + */ +bool si_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = si_gpu_check_soft_reset(rdev); + u32 mask; + + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + mask = RADEON_RESET_DMA; + else + mask = RADEON_RESET_DMA1; + + if (!(reset_mask & mask)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + +/** + * si_dma_vm_set_page - update the page tables using the DMA + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update the page tables using the DMA (SI). + */ +void si_dma_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) +{ + uint32_t r600_flags = cayman_vm_page_flags(rdev, flags); + uint64_t value; + unsigned ndw; + + if (flags & RADEON_VM_PAGE_SYSTEM) { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + /* for non-physically contiguous pages (system) */ + ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, ndw); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + if (flags & RADEON_VM_PAGE_SYSTEM) { + value = radeon_vm_map_gart(rdev, addr); + value &= 0xFFFFFFFFFFFFF000ULL; + } else if (flags & RADEON_VM_PAGE_VALID) { + value = addr; + } else { + value = 0; + } + addr += incr; + value |= r600_flags; + ib->ptr[ib->length_dw++] = value; + ib->ptr[ib->length_dw++] = upper_32_bits(value); + } + } + } else { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + if (flags & RADEON_VM_PAGE_VALID) + value = addr; + else + value = 0; + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw); + ib->ptr[ib->length_dw++] = pe; /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff; + ib->ptr[ib->length_dw++] = r600_flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = value; /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(value); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + pe += ndw * 4; + addr += (ndw / 2) * incr; + count -= ndw / 2; + } + } + while (ib->length_dw & 0x7) + ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0); +} + +void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +{ + struct radeon_ring *ring = &rdev->ring[ridx]; + + if (vm == NULL) + return; + + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); + if (vm->id < 8) { + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2)); + } else { + radeon_ring_write(ring, (0xf << 16) | ((VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2)); + } + radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + + /* flush hdp cache */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL >> 2)); + radeon_ring_write(ring, 1); + + /* bits 0-7 are the VM contexts0-7 */ + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); + radeon_ring_write(ring, (0xf << 16) | (VM_INVALIDATE_REQUEST >> 2)); + radeon_ring_write(ring, 1 << vm->id); +} + +/** + * si_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (SI). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int si_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_bytes, cur_size_in_bytes; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); + num_loops = DIV_ROUND_UP(size_in_bytes, 0xfffff); + r = radeon_ring_lock(rdev, ring, num_loops * 5 + 11); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_bytes = size_in_bytes; + if (cur_size_in_bytes > 0xFFFFF) + cur_size_in_bytes = 0xFFFFF; + size_in_bytes -= cur_size_in_bytes; + radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 1, 0, 0, cur_size_in_bytes)); + radeon_ring_write(ring, dst_offset & 0xffffffff); + radeon_ring_write(ring, src_offset & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); + radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff); + src_offset += cur_size_in_bytes; + dst_offset += cur_size_in_bytes; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} + -- cgit v1.2.3 From 59a82d0e65539812652dffea160ef483c1d310f5 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 13 Aug 2013 12:48:06 -0400 Subject: drm/radeon/si: properly set up the clearstate buffer for pg (v2) The format of the clearstate buffer used for pg (powergating) changed between NI and SI. This formats it properly for what the hardware expects on SI. v2: fix addresses Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/evergreen.c | 93 ++++++++++++++++++++++---------------- drivers/gpu/drm/radeon/si.c | 91 +++++++++++++++++++++++++++++++++++++ 2 files changed, 145 insertions(+), 39 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index bbaa4f2056ce..536908109001 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -142,6 +142,9 @@ extern void cayman_vm_decode_fault(struct radeon_device *rdev, u32 status, u32 addr); void cik_init_cp_pg_table(struct radeon_device *rdev); +extern u32 si_get_csb_size(struct radeon_device *rdev); +extern void si_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer); + static const u32 evergreen_golden_registers[] = { 0x3f90, 0xffff0000, 0xff000000, @@ -3893,7 +3896,7 @@ int sumo_rlc_init(struct radeon_device *rdev) const u32 *src_ptr; volatile u32 *dst_ptr; u32 dws, data, i, j, k, reg_num; - u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index; + u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index = 0; u64 reg_list_mc_addr; const struct cs_section_def *cs_data; int r; @@ -3937,7 +3940,7 @@ int sumo_rlc_init(struct radeon_device *rdev) dst_ptr = rdev->rlc.sr_ptr; if (rdev->family >= CHIP_TAHITI) { /* SI */ - for (i = 0; i < dws; i++) + for (i = 0; i < rdev->rlc.reg_list_size; i++) dst_ptr[i] = src_ptr[i]; } else { /* ON/LN/TN */ @@ -3963,20 +3966,25 @@ int sumo_rlc_init(struct radeon_device *rdev) if (cs_data) { /* clear state block */ - reg_list_num = 0; - dws = 0; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_list_num++; - dws += cs_data[i].section[j].reg_count; + if (rdev->family >= CHIP_TAHITI) { + rdev->rlc.clear_state_size = si_get_csb_size(rdev); + dws = rdev->rlc.clear_state_size + (256 / 4); + } else { + reg_list_num = 0; + dws = 0; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_list_num++; + dws += cs_data[i].section[j].reg_count; + } } + reg_list_blk_index = (3 * reg_list_num + 2); + dws += reg_list_blk_index; + rdev->rlc.clear_state_size = dws; } - reg_list_blk_index = (3 * reg_list_num + 2); - dws += reg_list_blk_index; - rdev->rlc.clear_state_size = dws; if (rdev->rlc.clear_state_obj == NULL) { - r = radeon_bo_create(rdev, rdev->rlc.clear_state_size * 4, PAGE_SIZE, true, + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); if (r) { dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); @@ -4006,36 +4014,43 @@ int sumo_rlc_init(struct radeon_device *rdev) } /* set up the cs buffer */ dst_ptr = rdev->rlc.cs_ptr; - reg_list_hdr_blk_index = 0; - reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); - data = upper_32_bits(reg_list_mc_addr); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - for (i = 0; cs_data[i].section != NULL; i++) { - for (j = 0; cs_data[i].section[j].extent != NULL; j++) { - reg_num = cs_data[i].section[j].reg_count; - data = reg_list_mc_addr & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - data = 0x08000000 | (reg_num * 4); - dst_ptr[reg_list_hdr_blk_index] = data; - reg_list_hdr_blk_index++; - - for (k = 0; k < reg_num; k++) { - data = cs_data[i].section[j].extent[k]; - dst_ptr[reg_list_blk_index + k] = data; + if (rdev->family >= CHIP_TAHITI) { + reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + 256; + dst_ptr[0] = upper_32_bits(reg_list_mc_addr); + dst_ptr[1] = lower_32_bits(reg_list_mc_addr); + dst_ptr[2] = rdev->rlc.clear_state_size; + si_get_csb_buffer(rdev, &dst_ptr[(256/4)]); + } else { + reg_list_hdr_blk_index = 0; + reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); + data = upper_32_bits(reg_list_mc_addr); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_num = cs_data[i].section[j].reg_count; + data = reg_list_mc_addr & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = 0x08000000 | (reg_num * 4); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + for (k = 0; k < reg_num; k++) { + data = cs_data[i].section[j].extent[k]; + dst_ptr[reg_list_blk_index + k] = data; + } + reg_list_mc_addr += reg_num * 4; + reg_list_blk_index += reg_num; } - reg_list_mc_addr += reg_num * 4; - reg_list_blk_index += reg_num; } + dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; } - dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; - radeon_bo_kunmap(rdev->rlc.clear_state_obj); radeon_bo_unreserve(rdev->rlc.clear_state_obj); } diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index ff48c88c9133..ebe04e45e5c8 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -5286,6 +5286,97 @@ void si_update_pg(struct radeon_device *rdev, si_enable_gfx_cgpg(rdev, enable); } +u32 si_get_csb_size(struct radeon_device *rdev) +{ + u32 count = 0; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + + if (rdev->rlc.cs_data == NULL) + return 0; + + /* begin clear state */ + count += 2; + /* context control state */ + count += 3; + + for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) + count += 2 + ext->reg_count; + else + return 0; + } + } + /* pa_sc_raster_config */ + count += 3; + /* end clear state */ + count += 2; + /* clear state */ + count += 2; + + return count; +} + +void si_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer) +{ + u32 count = 0, i; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + + if (rdev->rlc.cs_data == NULL) + return; + if (buffer == NULL) + return; + + buffer[count++] = PACKET3(PACKET3_PREAMBLE_CNTL, 0); + buffer[count++] = PACKET3_PREAMBLE_BEGIN_CLEAR_STATE; + + buffer[count++] = PACKET3(PACKET3_CONTEXT_CONTROL, 1); + buffer[count++] = 0x80000000; + buffer[count++] = 0x80000000; + + for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) { + buffer[count++] = PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count); + buffer[count++] = ext->reg_index - 0xa000; + for (i = 0; i < ext->reg_count; i++) + buffer[count++] = ext->extent[i]; + } else { + return; + } + } + } + + buffer[count++] = PACKET3(PACKET3_SET_CONTEXT_REG, 1); + buffer[count++] = PA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START; + switch (rdev->family) { + case CHIP_TAHITI: + case CHIP_PITCAIRN: + buffer[count++] = 0x2a00126a; + break; + case CHIP_VERDE: + buffer[count++] = 0x0000124a; + break; + case CHIP_OLAND: + buffer[count++] = 0x00000082; + break; + case CHIP_HAINAN: + buffer[count++] = 0x00000000; + break; + default: + buffer[count++] = 0x00000000; + break; + } + + buffer[count++] = PACKET3(PACKET3_PREAMBLE_CNTL, 0); + buffer[count++] = PACKET3_PREAMBLE_END_CLEAR_STATE; + + buffer[count++] = PACKET3(PACKET3_CLEAR_STATE, 0); + buffer[count++] = 0; +} + static void si_init_pg(struct radeon_device *rdev) { if (rdev->pg_flags) { -- cgit v1.2.3 From a0f38609c9870fe0e3d5c10b1e6926a5750d0a7a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 22 Aug 2013 11:57:46 -0400 Subject: drm/radeon/cik: properly set up the clearstate buffer for pg (v2) The format of the clearstate buffer used for pg (powergating) changed between NI and SI. This formats it properly for what the hardware expects on SI+. v2: fix addresses Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/cik.c | 93 +++++++++++++++++++++++++++++++++++++- drivers/gpu/drm/radeon/evergreen.c | 13 +++++- 2 files changed, 103 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 1f088800295d..582f8e4f36d4 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -5622,7 +5622,7 @@ static void cik_init_gfx_cgpg(struct radeon_device *rdev) if (rdev->rlc.cs_data) { WREG32(RLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET); WREG32(RLC_GPM_SCRATCH_DATA, upper_32_bits(rdev->rlc.clear_state_gpu_addr)); - WREG32(RLC_GPM_SCRATCH_DATA, rdev->rlc.clear_state_gpu_addr); + WREG32(RLC_GPM_SCRATCH_DATA, lower_32_bits(rdev->rlc.clear_state_gpu_addr)); WREG32(RLC_GPM_SCRATCH_DATA, rdev->rlc.clear_state_size); } else { WREG32(RLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET); @@ -5670,6 +5670,97 @@ static void cik_update_gfx_pg(struct radeon_device *rdev, bool enable) cik_enable_gfx_dynamic_mgpg(rdev, enable); } +u32 cik_get_csb_size(struct radeon_device *rdev) +{ + u32 count = 0; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + + if (rdev->rlc.cs_data == NULL) + return 0; + + /* begin clear state */ + count += 2; + /* context control state */ + count += 3; + + for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) + count += 2 + ext->reg_count; + else + return 0; + } + } + /* pa_sc_raster_config/pa_sc_raster_config1 */ + count += 4; + /* end clear state */ + count += 2; + /* clear state */ + count += 2; + + return count; +} + +void cik_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer) +{ + u32 count = 0, i; + const struct cs_section_def *sect = NULL; + const struct cs_extent_def *ext = NULL; + + if (rdev->rlc.cs_data == NULL) + return; + if (buffer == NULL) + return; + + buffer[count++] = PACKET3(PACKET3_PREAMBLE_CNTL, 0); + buffer[count++] = PACKET3_PREAMBLE_BEGIN_CLEAR_STATE; + + buffer[count++] = PACKET3(PACKET3_CONTEXT_CONTROL, 1); + buffer[count++] = 0x80000000; + buffer[count++] = 0x80000000; + + for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) { + for (ext = sect->section; ext->extent != NULL; ++ext) { + if (sect->id == SECT_CONTEXT) { + buffer[count++] = PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count); + buffer[count++] = ext->reg_index - 0xa000; + for (i = 0; i < ext->reg_count; i++) + buffer[count++] = ext->extent[i]; + } else { + return; + } + } + } + + buffer[count++] = PACKET3(PACKET3_SET_CONTEXT_REG, 2); + buffer[count++] = PA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START; + switch (rdev->family) { + case CHIP_BONAIRE: + buffer[count++] = 0x16000012; + buffer[count++] = 0x00000000; + break; + case CHIP_KAVERI: + buffer[count++] = 0x00000000; /* XXX */ + buffer[count++] = 0x00000000; + break; + case CHIP_KABINI: + buffer[count++] = 0x00000000; /* XXX */ + buffer[count++] = 0x00000000; + break; + default: + buffer[count++] = 0x00000000; + buffer[count++] = 0x00000000; + break; + } + + buffer[count++] = PACKET3(PACKET3_PREAMBLE_CNTL, 0); + buffer[count++] = PACKET3_PREAMBLE_END_CLEAR_STATE; + + buffer[count++] = PACKET3(PACKET3_CLEAR_STATE, 0); + buffer[count++] = 0; +} + static void cik_init_pg(struct radeon_device *rdev) { if (rdev->pg_flags) { diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 536908109001..2ca9f13f2c79 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -144,6 +144,8 @@ void cik_init_cp_pg_table(struct radeon_device *rdev); extern u32 si_get_csb_size(struct radeon_device *rdev); extern void si_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer); +extern u32 cik_get_csb_size(struct radeon_device *rdev); +extern void cik_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer); static const u32 evergreen_golden_registers[] = { @@ -3903,6 +3905,9 @@ int sumo_rlc_init(struct radeon_device *rdev) src_ptr = rdev->rlc.reg_list; dws = rdev->rlc.reg_list_size; + if (rdev->family >= CHIP_BONAIRE) { + dws += (5 * 16) + 48 + 48 + 64; + } cs_data = rdev->rlc.cs_data; if (src_ptr) { @@ -3966,7 +3971,9 @@ int sumo_rlc_init(struct radeon_device *rdev) if (cs_data) { /* clear state block */ - if (rdev->family >= CHIP_TAHITI) { + if (rdev->family >= CHIP_BONAIRE) { + rdev->rlc.clear_state_size = dws = cik_get_csb_size(rdev); + } else if (rdev->family >= CHIP_TAHITI) { rdev->rlc.clear_state_size = si_get_csb_size(rdev); dws = rdev->rlc.clear_state_size + (256 / 4); } else { @@ -4014,7 +4021,9 @@ int sumo_rlc_init(struct radeon_device *rdev) } /* set up the cs buffer */ dst_ptr = rdev->rlc.cs_ptr; - if (rdev->family >= CHIP_TAHITI) { + if (rdev->family >= CHIP_BONAIRE) { + cik_get_csb_buffer(rdev, dst_ptr); + } else if (rdev->family >= CHIP_TAHITI) { reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + 256; dst_ptr[0] = upper_32_bits(reg_list_mc_addr); dst_ptr[1] = lower_32_bits(reg_list_mc_addr); -- cgit v1.2.3 From 0b31e02363b0db4e7931561bc6c141436e729d9f Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 19 Aug 2013 11:06:50 -0400 Subject: drm/radeon: update line buffer allocation for dce4.1/5 We need to allocate line buffer to each display when setting up the watermarks. Failure to do so can lead to a blank screen. This fixes blank screen problems on dce4.1/5 asics. Based on an initial fix from: Jay Cornwall Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/radeon/evergreen.c | 25 +++++++++++++++++++++---- drivers/gpu/drm/radeon/evergreend.h | 4 ++++ 2 files changed, 25 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 2ca9f13f2c79..183213689478 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -1812,7 +1812,8 @@ static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev, struct drm_display_mode *mode, struct drm_display_mode *other_mode) { - u32 tmp; + u32 tmp, buffer_alloc, i; + u32 pipe_offset = radeon_crtc->crtc_id * 0x20; /* * Line Buffer Setup * There are 3 line buffers, each one shared by 2 display controllers. @@ -1835,18 +1836,34 @@ static u32 evergreen_line_buffer_adjust(struct radeon_device *rdev, * non-linked crtcs for maximum line buffer allocation. */ if (radeon_crtc->base.enabled && mode) { - if (other_mode) + if (other_mode) { tmp = 0; /* 1/2 */ - else + buffer_alloc = 1; + } else { tmp = 2; /* whole */ - } else + buffer_alloc = 2; + } + } else { tmp = 0; + buffer_alloc = 0; + } /* 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); + if (ASIC_IS_DCE41(rdev) || ASIC_IS_DCE5(rdev)) { + WREG32(PIPE0_DMIF_BUFFER_CONTROL + pipe_offset, + DMIF_BUFFERS_ALLOCATED(buffer_alloc)); + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(PIPE0_DMIF_BUFFER_CONTROL + pipe_offset) & + DMIF_BUFFERS_ALLOCATED_COMPLETED) + break; + udelay(1); + } + } + if (radeon_crtc->base.enabled && mode) { switch (tmp) { case 0: diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h index 430997a70acc..8768fd6a1e27 100644 --- a/drivers/gpu/drm/radeon/evergreend.h +++ b/drivers/gpu/drm/radeon/evergreend.h @@ -1160,6 +1160,10 @@ # define LATENCY_LOW_WATERMARK(x) ((x) << 0) # define LATENCY_HIGH_WATERMARK(x) ((x) << 16) +#define PIPE0_DMIF_BUFFER_CONTROL 0x0ca0 +# define DMIF_BUFFERS_ALLOCATED(x) ((x) << 0) +# define DMIF_BUFFERS_ALLOCATED_COMPLETED (1 << 4) + #define IH_RB_CNTL 0x3e00 # define IH_RB_ENABLE (1 << 0) # define IH_IB_SIZE(x) ((x) << 1) /* log2 */ -- cgit v1.2.3 From e5903d399a7b0e5c14673c1206f4aeec2859c730 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 30 Aug 2013 08:58:20 -0400 Subject: drm/radeon: fix init ordering for r600+ The vram scratch buffer needs to be initialized before the mc is programmed otherwise we program 0 as the GPU address of the default GPU fault page. In most cases we put vram at zero anyway and reserve a page for the legacy vga buffer so in practice this shouldn't cause any problems, but better to make it correct. Was changed in: 6fab3febf6d949b0a12b1e4e73db38e4a177a79e Reported-by: FrankR Huang Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/radeon/cik.c | 9 +++++---- drivers/gpu/drm/radeon/evergreen.c | 9 +++++---- drivers/gpu/drm/radeon/ni.c | 9 +++++---- drivers/gpu/drm/radeon/r600.c | 9 +++++---- drivers/gpu/drm/radeon/rv770.c | 9 +++++---- drivers/gpu/drm/radeon/si.c | 9 +++++---- 6 files changed, 30 insertions(+), 24 deletions(-) (limited to 'drivers/gpu/drm/radeon/evergreen.c') diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index 79124f81c00e..148c539684bb 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -6951,6 +6951,11 @@ static int cik_startup(struct radeon_device *rdev) /* enable aspm */ cik_program_aspm(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + cik_mc_program(rdev); if (rdev->flags & RADEON_IS_IGP) { @@ -6980,10 +6985,6 @@ static int cik_startup(struct radeon_device *rdev) } } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - r = cik_pcie_gart_enable(rdev); if (r) return r; diff --git a/drivers/gpu/drm/radeon/evergreen.c b/drivers/gpu/drm/radeon/evergreen.c index 183213689478..6398c1f76fb8 100644 --- a/drivers/gpu/drm/radeon/evergreen.c +++ b/drivers/gpu/drm/radeon/evergreen.c @@ -5053,6 +5053,11 @@ static int evergreen_startup(struct radeon_device *rdev) /* enable aspm */ evergreen_program_aspm(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + evergreen_mc_program(rdev); if (ASIC_IS_DCE5(rdev)) { @@ -5078,10 +5083,6 @@ static int evergreen_startup(struct radeon_device *rdev) } } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - if (rdev->flags & RADEON_IS_AGP) { evergreen_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c index 69499fff06b0..d60049efd7ac 100644 --- a/drivers/gpu/drm/radeon/ni.c +++ b/drivers/gpu/drm/radeon/ni.c @@ -1863,6 +1863,11 @@ static int cayman_startup(struct radeon_device *rdev) /* enable aspm */ evergreen_program_aspm(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + evergreen_mc_program(rdev); if (rdev->flags & RADEON_IS_IGP) { @@ -1889,10 +1894,6 @@ static int cayman_startup(struct radeon_device *rdev) } } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - r = cayman_pcie_gart_enable(rdev); if (r) return r; diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c index 087cff444ba2..b72d4d717a72 100644 --- a/drivers/gpu/drm/radeon/r600.c +++ b/drivers/gpu/drm/radeon/r600.c @@ -2698,6 +2698,11 @@ static int r600_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ r600_pcie_gen2_enable(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + r600_mc_program(rdev); if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { @@ -2708,10 +2713,6 @@ static int r600_startup(struct radeon_device *rdev) } } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - if (rdev->flags & RADEON_IS_AGP) { r600_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/rv770.c b/drivers/gpu/drm/radeon/rv770.c index b811296462a3..9f5846743c9e 100644 --- a/drivers/gpu/drm/radeon/rv770.c +++ b/drivers/gpu/drm/radeon/rv770.c @@ -1658,6 +1658,11 @@ static int rv770_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ rv770_pcie_gen2_enable(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + rv770_mc_program(rdev); if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { @@ -1668,10 +1673,6 @@ static int rv770_startup(struct radeon_device *rdev) } } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - if (rdev->flags & RADEON_IS_AGP) { rv770_agp_enable(rdev); } else { diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c index 89393ed593fa..fe8bca686900 100644 --- a/drivers/gpu/drm/radeon/si.c +++ b/drivers/gpu/drm/radeon/si.c @@ -6343,6 +6343,11 @@ static int si_startup(struct radeon_device *rdev) /* enable aspm */ si_program_aspm(rdev); + /* scratch needs to be initialized before MC */ + r = r600_vram_scratch_init(rdev); + if (r) + return r; + si_mc_program(rdev); if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || @@ -6360,10 +6365,6 @@ static int si_startup(struct radeon_device *rdev) return r; } - r = r600_vram_scratch_init(rdev); - if (r) - return r; - r = si_pcie_gart_enable(rdev); if (r) return r; -- cgit v1.2.3