From: Christian König christian.koenig@amd.com

Only VCE 2.0 support so far.

v2: squashing multiple patches into this one v3: add IRQ support for CIK, major cleanups, basic code documentation v4: remove HAINAN from chipset list

Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/radeon/Makefile | 6 + drivers/gpu/drm/radeon/cik.c | 60 ++++ drivers/gpu/drm/radeon/cikd.h | 33 ++ drivers/gpu/drm/radeon/radeon.h | 56 +++- drivers/gpu/drm/radeon/radeon_asic.c | 17 + drivers/gpu/drm/radeon/radeon_asic.h | 13 + drivers/gpu/drm/radeon/radeon_cs.c | 4 + drivers/gpu/drm/radeon/radeon_kms.c | 1 + drivers/gpu/drm/radeon/radeon_ring.c | 4 + drivers/gpu/drm/radeon/radeon_test.c | 39 ++- drivers/gpu/drm/radeon/radeon_vce.c | 588 +++++++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/sid.h | 47 +++ drivers/gpu/drm/radeon/vce_v1_0.c | 187 +++++++++++ drivers/gpu/drm/radeon/vce_v2_0.c | 70 +++++ include/uapi/drm/radeon_drm.h | 1 + 15 files changed, 1117 insertions(+), 9 deletions(-) create mode 100644 drivers/gpu/drm/radeon/radeon_vce.c create mode 100644 drivers/gpu/drm/radeon/vce_v1_0.c create mode 100644 drivers/gpu/drm/radeon/vce_v2_0.c

diff --git a/drivers/gpu/drm/radeon/Makefile b/drivers/gpu/drm/radeon/Makefile index 306364a..ed60caa 100644 --- a/drivers/gpu/drm/radeon/Makefile +++ b/drivers/gpu/drm/radeon/Makefile @@ -99,6 +99,12 @@ radeon-y += \ uvd_v3_1.o \ uvd_v4_2.o

+# add VCE block +radeon-y += \ + radeon_vce.o \ + vce_v1_0.o \ + vce_v2_0.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 e6419ca..be6eb4d 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -7490,6 +7490,20 @@ restart_ih: /* reset addr and status */ WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1); break; + case 167: /* VCE */ + DRM_DEBUG("IH: VCE int: 0x%08x\n", src_data); + switch (src_data) { + case 0: + radeon_fence_process(rdev, TN_RING_TYPE_VCE1_INDEX); + break; + case 1: + radeon_fence_process(rdev, TN_RING_TYPE_VCE2_INDEX); + break; + default: + DRM_ERROR("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; case 176: /* GFX RB CP_INT */ case 177: /* GFX IB CP_INT */ radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX); @@ -7789,6 +7803,22 @@ static int cik_startup(struct radeon_device *rdev) if (r) rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_size = 0;

+ r = radeon_vce_resume(rdev); + if (!r) { + r = vce_v2_0_resume(rdev); + if (!r) + r = radeon_fence_driver_start_ring(rdev, + TN_RING_TYPE_VCE1_INDEX); + if (!r) + r = radeon_fence_driver_start_ring(rdev, + TN_RING_TYPE_VCE2_INDEX); + } + if (r) { + dev_err(rdev->dev, "VCE init error (%d).\n", r); + rdev->ring[TN_RING_TYPE_VCE1_INDEX].ring_size = 0; + rdev->ring[TN_RING_TYPE_VCE2_INDEX].ring_size = 0; + } + /* Enable IRQ */ if (!rdev->irq.installed) { r = radeon_irq_kms_init(rdev); @@ -7864,6 +7894,23 @@ static int cik_startup(struct radeon_device *rdev) DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); }

+ r = -ENOENT; + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + if (ring->ring_size) + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, + VCE_CMD_NO_OP); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + if (ring->ring_size) + r = radeon_ring_init(rdev, ring, ring->ring_size, 0, + VCE_CMD_NO_OP); + + if (!r) + r = vce_v1_0_init(rdev); + else if (r != -ENOENT) + DRM_ERROR("radeon: failed initializing VCE (%d).\n", r); + r = radeon_ib_pool_init(rdev); if (r) { dev_err(rdev->dev, "IB initialization failed (%d).\n", r); @@ -7934,6 +7981,7 @@ int cik_suspend(struct radeon_device *rdev) cik_sdma_enable(rdev, false); uvd_v1_0_fini(rdev); radeon_uvd_suspend(rdev); + radeon_vce_suspend(rdev); cik_fini_pg(rdev); cik_fini_cg(rdev); cik_irq_suspend(rdev); @@ -8066,6 +8114,17 @@ int cik_init(struct radeon_device *rdev) r600_ring_init(rdev, ring, 4096); }

+ r = radeon_vce_init(rdev); + if (!r) { + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 4096); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 4096); + } + rdev->ih.ring_obj = NULL; r600_ih_ring_init(rdev, 64 * 1024);

@@ -8127,6 +8186,7 @@ void cik_fini(struct radeon_device *rdev) radeon_irq_kms_fini(rdev); uvd_v1_0_fini(rdev); radeon_uvd_fini(rdev); + radeon_vce_fini(rdev); cik_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 98bae9d7..459ae02 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -2010,4 +2010,37 @@ /* UVD CTX indirect */ #define UVD_CGC_MEM_CTRL 0xC0

+/* VCE */ + +#define VCE_VCPU_CACHE_OFFSET0 0x20024 +#define VCE_VCPU_CACHE_SIZE0 0x20028 +#define VCE_VCPU_CACHE_OFFSET1 0x2002c +#define VCE_VCPU_CACHE_SIZE1 0x20030 +#define VCE_VCPU_CACHE_OFFSET2 0x20034 +#define VCE_VCPU_CACHE_SIZE2 0x20038 +#define VCE_RB_RPTR2 0x20178 +#define VCE_RB_WPTR2 0x2017c +#define VCE_RB_RPTR 0x2018c +#define VCE_RB_WPTR 0x20190 +#define VCE_CLOCK_GATING_A 0x202f8 +#define VCE_CLOCK_GATING_B 0x202fc +#define VCE_UENC_CLOCK_GATING 0x207bc +#define VCE_UENC_REG_CLOCK_GATING 0x207c0 +#define VCE_SYS_INT_EN 0x21300 +# define VCE_SYS_INT_TRAP_INTERRUPT_EN (1 << 3) +#define VCE_LMI_CTRL2 0x21474 +#define VCE_LMI_CTRL 0x21498 +#define VCE_LMI_VM_CTRL 0x214a0 +#define VCE_LMI_SWAP_CNTL 0x214b4 +#define VCE_LMI_SWAP_CNTL1 0x214b8 +#define VCE_LMI_CACHE_CTRL 0x214f4 + +#define VCE_CMD_NO_OP 0x00000000 +#define VCE_CMD_END 0x00000001 +#define VCE_CMD_IB 0x00000002 +#define VCE_CMD_FENCE 0x00000003 +#define VCE_CMD_TRAP 0x00000004 +#define VCE_CMD_IB_AUTO 0x00000005 +#define VCE_CMD_SEMAPHORE 0x00000006 + #endif diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 4a8ac1c..a9295b9 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -113,19 +113,16 @@ extern int radeon_hard_reset; #define RADEONFB_CONN_LIMIT 4 #define RADEON_BIOS_NUM_SCRATCH 8

-/* max number of rings */ -#define RADEON_NUM_RINGS 6 - /* fence seq are set to this number when signaled */ #define RADEON_FENCE_SIGNALED_SEQ 0LL

/* internal ring indices */ /* r1xx+ has gfx CP ring */ -#define RADEON_RING_TYPE_GFX_INDEX 0 +#define RADEON_RING_TYPE_GFX_INDEX 0

/* cayman has 2 compute CP rings */ -#define CAYMAN_RING_TYPE_CP1_INDEX 1 -#define CAYMAN_RING_TYPE_CP2_INDEX 2 +#define CAYMAN_RING_TYPE_CP1_INDEX 1 +#define CAYMAN_RING_TYPE_CP2_INDEX 2

/* R600+ has an async dma ring */ #define R600_RING_TYPE_DMA_INDEX 3 @@ -133,7 +130,14 @@ extern int radeon_hard_reset; #define CAYMAN_RING_TYPE_DMA1_INDEX 4

/* R600+ */ -#define R600_RING_TYPE_UVD_INDEX 5 +#define R600_RING_TYPE_UVD_INDEX 5 + +/* TN+ */ +#define TN_RING_TYPE_VCE1_INDEX 6 +#define TN_RING_TYPE_VCE2_INDEX 7 + +/* max number of rings */ +#define RADEON_NUM_RINGS 8

/* hardcode those limit for now */ #define RADEON_VA_IB_OFFSET (1 << 20) @@ -1589,6 +1593,42 @@ int radeon_uvd_calc_upll_dividers(struct radeon_device *rdev, int radeon_uvd_send_upll_ctlreq(struct radeon_device *rdev, unsigned cg_upll_func_cntl);

+/* + * VCE + */ +#define RADEON_MAX_VCE_HANDLES 16 +#define RADEON_VCE_STACK_SIZE (1024*1024) +#define RADEON_VCE_HEAP_SIZE (4*1024*1024) + +struct radeon_vce { + struct radeon_bo *vcpu_bo; + void *cpu_addr; + uint64_t gpu_addr; + atomic_t handles[RADEON_MAX_VCE_HANDLES]; + struct drm_file *filp[RADEON_MAX_VCE_HANDLES]; +}; + +int radeon_vce_init(struct radeon_device *rdev); +void radeon_vce_fini(struct radeon_device *rdev); +int radeon_vce_suspend(struct radeon_device *rdev); +int radeon_vce_resume(struct radeon_device *rdev); +int radeon_vce_get_create_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence); +int radeon_vce_get_destroy_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence); +void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp); +int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi); +int radeon_vce_cs_parse(struct radeon_cs_parser *p); +bool radeon_vce_semaphore_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait); +void radeon_vce_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); +void radeon_vce_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence); +int radeon_vce_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); +int radeon_vce_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); + struct r600_audio_pin { int channels; int rate; @@ -2184,6 +2224,7 @@ struct radeon_device { struct radeon_gem gem; struct radeon_pm pm; struct radeon_uvd uvd; + struct radeon_vce vce; uint32_t bios_scratch[RADEON_BIOS_NUM_SCRATCH]; struct radeon_wb wb; struct radeon_dummy_page dummy_page; @@ -2203,6 +2244,7 @@ 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 */ + const struct firmware *vce_fw; /* VCE firmware */ 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.c b/drivers/gpu/drm/radeon/radeon_asic.c index f74db43..e9eaba99 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -1987,6 +1987,19 @@ static struct radeon_asic_ring ci_dma_ring = { .set_wptr = &cik_sdma_set_wptr, };

+static struct radeon_asic_ring ci_vce_ring = { + .ib_execute = &radeon_vce_ib_execute, + .emit_fence = &radeon_vce_fence_emit, + .emit_semaphore = &radeon_vce_semaphore_emit, + .cs_parse = &radeon_vce_cs_parse, + .ring_test = &radeon_vce_ring_test, + .ib_test = &radeon_vce_ib_test, + .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &vce_v1_0_get_rptr, + .get_wptr = &vce_v1_0_get_wptr, + .set_wptr = &vce_v1_0_set_wptr, +}; + static struct radeon_asic ci_asic = { .init = &cik_init, .fini = &cik_fini, @@ -2015,6 +2028,8 @@ static struct radeon_asic ci_asic = { [R600_RING_TYPE_DMA_INDEX] = &ci_dma_ring, [CAYMAN_RING_TYPE_DMA1_INDEX] = &ci_dma_ring, [R600_RING_TYPE_UVD_INDEX] = &cayman_uvd_ring, + [TN_RING_TYPE_VCE1_INDEX] = &ci_vce_ring, + [TN_RING_TYPE_VCE2_INDEX] = &ci_vce_ring, }, .irq = { .set = &cik_irq_set, @@ -2117,6 +2132,8 @@ static struct radeon_asic kv_asic = { [R600_RING_TYPE_DMA_INDEX] = &ci_dma_ring, [CAYMAN_RING_TYPE_DMA1_INDEX] = &ci_dma_ring, [R600_RING_TYPE_UVD_INDEX] = &cayman_uvd_ring, + [TN_RING_TYPE_VCE1_INDEX] = &ci_vce_ring, + [TN_RING_TYPE_VCE2_INDEX] = &ci_vce_ring, }, .irq = { .set = &cik_irq_set, diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index b3bc433..2017efd 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -861,4 +861,17 @@ bool uvd_v3_1_semaphore_emit(struct radeon_device *rdev, /* uvd v4.2 */ int uvd_v4_2_resume(struct radeon_device *rdev);

+/* vce v1.0 */ +uint32_t vce_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +uint32_t vce_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void vce_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +int vce_v1_0_init(struct radeon_device *rdev); +int vce_v1_0_start(struct radeon_device *rdev); + +/* vce v2.0 */ +int vce_v2_0_resume(struct radeon_device *rdev); + #endif diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index dfb5a1d..701ee79 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -147,6 +147,10 @@ static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority case RADEON_CS_RING_UVD: p->ring = R600_RING_TYPE_UVD_INDEX; break; + case RADEON_CS_RING_VCE: + /* TODO: only use the low priority ring for now */ + p->ring = TN_RING_TYPE_VCE1_INDEX; + break; } return 0; } diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 114d167..0e078af 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -610,6 +610,7 @@ void radeon_driver_preclose_kms(struct drm_device *dev, if (rdev->cmask_filp == file_priv) rdev->cmask_filp = NULL; radeon_uvd_free_handles(rdev, file_priv); + radeon_vce_free_handles(rdev, file_priv); }

/* diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c index 1b783f0..149cd63 100644 --- a/drivers/gpu/drm/radeon/radeon_ring.c +++ b/drivers/gpu/drm/radeon/radeon_ring.c @@ -814,6 +814,8 @@ static int cayman_cp2_index = CAYMAN_RING_TYPE_CP2_INDEX; static int radeon_dma1_index = R600_RING_TYPE_DMA_INDEX; static int radeon_dma2_index = CAYMAN_RING_TYPE_DMA1_INDEX; static int r600_uvd_index = R600_RING_TYPE_UVD_INDEX; +static int si_vce1_index = TN_RING_TYPE_VCE1_INDEX; +static int si_vce2_index = TN_RING_TYPE_VCE2_INDEX;

static struct drm_info_list radeon_debugfs_ring_info_list[] = { {"radeon_ring_gfx", radeon_debugfs_ring_info, 0, &radeon_gfx_index}, @@ -822,6 +824,8 @@ static struct drm_info_list radeon_debugfs_ring_info_list[] = { {"radeon_ring_dma1", radeon_debugfs_ring_info, 0, &radeon_dma1_index}, {"radeon_ring_dma2", radeon_debugfs_ring_info, 0, &radeon_dma2_index}, {"radeon_ring_uvd", radeon_debugfs_ring_info, 0, &r600_uvd_index}, + {"radeon_ring_vce1", radeon_debugfs_ring_info, 0, &si_vce1_index}, + {"radeon_ring_vce2", radeon_debugfs_ring_info, 0, &si_vce2_index}, };

static int radeon_debugfs_sa_info(struct seq_file *m, void *data) diff --git a/drivers/gpu/drm/radeon/radeon_test.c b/drivers/gpu/drm/radeon/radeon_test.c index 12e8099..3a13e0d 100644 --- a/drivers/gpu/drm/radeon/radeon_test.c +++ b/drivers/gpu/drm/radeon/radeon_test.c @@ -257,20 +257,36 @@ static int radeon_test_create_and_emit_fence(struct radeon_device *rdev, struct radeon_ring *ring, struct radeon_fence **fence) { + uint32_t handle = ring->idx ^ 0xdeafbeef; int r;

if (ring->idx == R600_RING_TYPE_UVD_INDEX) { - r = radeon_uvd_get_create_msg(rdev, ring->idx, 1, NULL); + r = radeon_uvd_get_create_msg(rdev, ring->idx, handle, NULL); if (r) { DRM_ERROR("Failed to get dummy create msg\n"); return r; }

- r = radeon_uvd_get_destroy_msg(rdev, ring->idx, 1, fence); + r = radeon_uvd_get_destroy_msg(rdev, ring->idx, handle, fence); if (r) { DRM_ERROR("Failed to get dummy destroy msg\n"); return r; } + + } else if (ring->idx == TN_RING_TYPE_VCE1_INDEX || + ring->idx == TN_RING_TYPE_VCE2_INDEX) { + r = radeon_vce_get_create_msg(rdev, ring->idx, handle, NULL); + if (r) { + DRM_ERROR("Failed to get dummy create msg\n"); + return r; + } + + r = radeon_vce_get_destroy_msg(rdev, ring->idx, handle, fence); + if (r) { + DRM_ERROR("Failed to get dummy destroy msg\n"); + return r; + } + } else { r = radeon_ring_lock(rdev, ring, 64); if (r) { @@ -486,6 +502,16 @@ out_cleanup: printk(KERN_WARNING "Error while testing ring sync (%d).\n", r); }

+static bool radeon_test_sync_possible(struct radeon_ring *ringA, + struct radeon_ring *ringB) +{ + if (ringA->idx == TN_RING_TYPE_VCE2_INDEX && + ringB->idx == TN_RING_TYPE_VCE1_INDEX) + return false; + + return true; +} + void radeon_test_syncing(struct radeon_device *rdev) { int i, j, k; @@ -500,6 +526,9 @@ void radeon_test_syncing(struct radeon_device *rdev) if (!ringB->ready) continue;

+ if (!radeon_test_sync_possible(ringA, ringB)) + continue; + DRM_INFO("Testing syncing between rings %d and %d...\n", i, j); radeon_test_ring_sync(rdev, ringA, ringB);

@@ -511,6 +540,12 @@ void radeon_test_syncing(struct radeon_device *rdev) if (!ringC->ready) continue;

+ if (!radeon_test_sync_possible(ringA, ringC)) + continue; + + if (!radeon_test_sync_possible(ringB, ringC)) + continue; + DRM_INFO("Testing syncing between rings %d, %d and %d...\n", i, j, k); radeon_test_ring_sync2(rdev, ringA, ringB, ringC);

diff --git a/drivers/gpu/drm/radeon/radeon_vce.c b/drivers/gpu/drm/radeon/radeon_vce.c new file mode 100644 index 0000000..2547d8e --- /dev/null +++ b/drivers/gpu/drm/radeon/radeon_vce.c @@ -0,0 +1,588 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König christian.koenig@amd.com + */ + +#include <linux/firmware.h> +#include <linux/module.h> +#include <drm/drmP.h> +#include <drm/drm.h> + +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" + +/* Firmware Names */ +#define FIRMWARE_BONAIRE "radeon/BONAIRE_vce.bin" + +MODULE_FIRMWARE(FIRMWARE_BONAIRE); + +/** + * radeon_vce_init - allocate memory, load vce firmware + * + * @rdev: radeon_device pointer + * + * First step to get VCE online, allocate memory and load the firmware + */ +int radeon_vce_init(struct radeon_device *rdev) +{ + unsigned long bo_size; + const char *fw_name; + int i, r; + + switch (rdev->family) { + case CHIP_BONAIRE: + case CHIP_KAVERI: + case CHIP_KABINI: + fw_name = FIRMWARE_BONAIRE; + break; + + default: + return -EINVAL; + } + + r = request_firmware(&rdev->vce_fw, fw_name, rdev->dev); + if (r) { + dev_err(rdev->dev, "radeon_vce: Can't load firmware "%s"\n", + fw_name); + return r; + } + + bo_size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size) + + RADEON_VCE_STACK_SIZE + RADEON_VCE_HEAP_SIZE; + r = radeon_bo_create(rdev, bo_size, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->vce.vcpu_bo); + if (r) { + dev_err(rdev->dev, "(%d) failed to allocate VCE bo\n", r); + return r; + } + + r = radeon_vce_resume(rdev); + if (r) + return r; + + memset(rdev->vce.cpu_addr, 0, bo_size); + memcpy(rdev->vce.cpu_addr, rdev->vce_fw->data, rdev->vce_fw->size); + + r = radeon_vce_suspend(rdev); + if (r) + return r; + + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + atomic_set(&rdev->vce.handles[i], 0); + rdev->vce.filp[i] = NULL; + } + + return 0; +} + +/** + * radeon_vce_fini - free memory + * + * @rdev: radeon_device pointer + * + * Last step on VCE teardown, free firmware memory + */ +void radeon_vce_fini(struct radeon_device *rdev) +{ + radeon_vce_suspend(rdev); + radeon_bo_unref(&rdev->vce.vcpu_bo); +} + +/** + * radeon_vce_suspend - unpin VCE fw memory + * + * @rdev: radeon_device pointer + * + * TODO: Test VCE suspend/resume + */ +int radeon_vce_suspend(struct radeon_device *rdev) +{ + int r; + + if (rdev->vce.vcpu_bo == NULL) + return 0; + + r = radeon_bo_reserve(rdev->vce.vcpu_bo, false); + if (!r) { + radeon_bo_kunmap(rdev->vce.vcpu_bo); + radeon_bo_unpin(rdev->vce.vcpu_bo); + radeon_bo_unreserve(rdev->vce.vcpu_bo); + } + return r; +} + +/** + * radeon_vce_resume - pin VCE fw memory + * + * @rdev: radeon_device pointer + * + * TODO: Test VCE suspend/resume + */ +int radeon_vce_resume(struct radeon_device *rdev) +{ + int r; + + if (rdev->vce.vcpu_bo == NULL) + return -EINVAL; + + r = radeon_bo_reserve(rdev->vce.vcpu_bo, false); + if (r) { + radeon_bo_unref(&rdev->vce.vcpu_bo); + dev_err(rdev->dev, "(%d) failed to reserve VCE bo\n", r); + return r; + } + + r = radeon_bo_pin(rdev->vce.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, + &rdev->vce.gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->vce.vcpu_bo); + radeon_bo_unref(&rdev->vce.vcpu_bo); + dev_err(rdev->dev, "(%d) VCE bo pin failed\n", r); + return r; + } + + r = radeon_bo_kmap(rdev->vce.vcpu_bo, &rdev->vce.cpu_addr); + if (r) { + dev_err(rdev->dev, "(%d) VCE map failed\n", r); + return r; + } + + radeon_bo_unreserve(rdev->vce.vcpu_bo); + + return 0; +} + +/** + * radeon_vce_free_handles - free still open VCE handles + * + * @rdev: radeon_device pointer + * @filp: drm file pointer + * + * Close all VCE handles still open by this file pointer + */ +void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp) +{ + int i, r; + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + uint32_t handle = atomic_read(&rdev->vce.handles[i]); + if (!handle || rdev->vce.filp[i] != filp) + continue; + + r = radeon_vce_get_destroy_msg(rdev, TN_RING_TYPE_VCE1_INDEX, + handle, NULL); + if (r) + DRM_ERROR("Error destroying VCE handle (%d)!\n", r); + + rdev->vce.filp[i] = NULL; + atomic_set(&rdev->vce.handles[i], 0); + } +} + +/** + * radeon_vce_get_create_msg - generate a VCE create msg + * + * @rdev: radeon_device pointer + * @ring: ring we should submit the msg to + * @handle: VCE session handle to use + * @fence: optional fence to return + * + * Open up a stream for HW test + */ +int radeon_vce_get_create_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence) +{ + const unsigned ib_size_dw = 1024; + struct radeon_ib ib; + uint64_t dummy; + int i, r; + + r = radeon_ib_get(rdev, ring, &ib, NULL, ib_size_dw * 4); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + dummy = ib.gpu_addr + 1024; + + /* stitch together an VCE create msg */ + ib.length_dw = 0; + ib.ptr[ib.length_dw++] = 0x0000000c; /* len */ + ib.ptr[ib.length_dw++] = 0x00000001; /* session cmd */ + ib.ptr[ib.length_dw++] = handle; + + ib.ptr[ib.length_dw++] = 0x00000030; /* len */ + ib.ptr[ib.length_dw++] = 0x01000001; /* create cmd */ + ib.ptr[ib.length_dw++] = 0x00000000; + ib.ptr[ib.length_dw++] = 0x00000042; + ib.ptr[ib.length_dw++] = 0x0000000a; + ib.ptr[ib.length_dw++] = 0x00000001; + ib.ptr[ib.length_dw++] = 0x00000080; + ib.ptr[ib.length_dw++] = 0x00000060; + ib.ptr[ib.length_dw++] = 0x00000100; + ib.ptr[ib.length_dw++] = 0x00000100; + ib.ptr[ib.length_dw++] = 0x0000000c; + ib.ptr[ib.length_dw++] = 0x00000000; + + ib.ptr[ib.length_dw++] = 0x00000014; /* len */ + ib.ptr[ib.length_dw++] = 0x05000005; /* feedback buffer */ + ib.ptr[ib.length_dw++] = upper_32_bits(dummy); + ib.ptr[ib.length_dw++] = dummy; + ib.ptr[ib.length_dw++] = 0x00000001; + + for (i = ib.length_dw; i < ib_size_dw; ++i) + ib.ptr[i] = 0x0; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + } + + if (fence) + *fence = radeon_fence_ref(ib.fence); + + radeon_ib_free(rdev, &ib); + + return r; +} + +/** + * radeon_vce_get_destroy_msg - generate a VCE destroy msg + * + * @rdev: radeon_device pointer + * @ring: ring we should submit the msg to + * @handle: VCE session handle to use + * @fence: optional fence to return + * + * Close up a stream for HW test or if userspace failed to do so + */ +int radeon_vce_get_destroy_msg(struct radeon_device *rdev, int ring, + uint32_t handle, struct radeon_fence **fence) +{ + const unsigned ib_size_dw = 1024; + struct radeon_ib ib; + uint64_t dummy; + int i, r; + + r = radeon_ib_get(rdev, ring, &ib, NULL, ib_size_dw * 4); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + dummy = ib.gpu_addr + 1024; + + /* stitch together an VCE destroy msg */ + ib.length_dw = 0; + ib.ptr[ib.length_dw++] = 0x0000000c; /* len */ + ib.ptr[ib.length_dw++] = 0x00000001; /* session cmd */ + ib.ptr[ib.length_dw++] = handle; + + ib.ptr[ib.length_dw++] = 0x00000014; /* len */ + ib.ptr[ib.length_dw++] = 0x05000005; /* feedback buffer */ + ib.ptr[ib.length_dw++] = upper_32_bits(dummy); + ib.ptr[ib.length_dw++] = dummy; + ib.ptr[ib.length_dw++] = 0x00000001; + + ib.ptr[ib.length_dw++] = 0x00000008; /* len */ + ib.ptr[ib.length_dw++] = 0x02000001; /* destroy cmd */ + + for (i = ib.length_dw; i < ib_size_dw; ++i) + ib.ptr[i] = 0x0; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + } + + if (fence) + *fence = radeon_fence_ref(ib.fence); + + radeon_ib_free(rdev, &ib); + + return r; +} + +/** + * radeon_vce_cs_reloc - command submission relocation + * + * @p: parser context + * @lo: address of lower dword + * @hi: address of higher dword + * + * Patch relocation inside command stream with real buffer address + */ +int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi) +{ + struct radeon_cs_chunk *relocs_chunk; + uint64_t offset; + unsigned idx; + + relocs_chunk = &p->chunks[p->chunk_relocs_idx]; + offset = radeon_get_ib_value(p, lo); + idx = radeon_get_ib_value(p, hi); + + if (idx >= relocs_chunk->length_dw) { + DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", + idx, relocs_chunk->length_dw); + return -EINVAL; + } + + offset += p->relocs_ptr[(idx / 4)]->lobj.gpu_offset; + + p->ib.ptr[lo] = offset & 0xFFFFFFFF; + p->ib.ptr[hi] = offset >> 32; + + return 0; +} + +/** + * radeon_vce_cs_parse - parse and validate the command stream + * + * @p: parser context + * + */ +int radeon_vce_cs_parse(struct radeon_cs_parser *p) +{ + uint32_t handle = 0; + bool destroy = false; + int i, r; + + while (p->idx < p->chunks[p->chunk_ib_idx].length_dw) { + uint32_t len = radeon_get_ib_value(p, p->idx); + uint32_t cmd = radeon_get_ib_value(p, p->idx + 1); + + if ((len < 8) || (len & 3)) { + DRM_ERROR("invalid VCE command length (%d)!\n", len); + return -EINVAL; + } + + switch (cmd) { + case 0x00000001: // session + handle = radeon_get_ib_value(p, p->idx + 2); + break; + + case 0x00000002: // task info + case 0x01000001: // create + case 0x04000001: // config extension + case 0x04000002: // pic control + case 0x04000005: // rate control + case 0x04000007: // motion estimation + case 0x04000008: // rdo + break; + + case 0x03000001: // encode + r = radeon_vce_cs_reloc(p, p->idx + 10, p->idx + 9); + if (r) + return r; + + r = radeon_vce_cs_reloc(p, p->idx + 12, p->idx + 11); + if (r) + return r; + break; + + case 0x02000001: // destroy + destroy = true; + break; + + case 0x05000001: // context buffer + case 0x05000004: // video bitstream buffer + case 0x05000005: // feedback buffer + r = radeon_vce_cs_reloc(p, p->idx + 3, p->idx + 2); + if (r) + return r; + break; + + default: + DRM_ERROR("invalid VCE command (0x%x)!\n", cmd); + return -EINVAL; + } + + p->idx += len / 4; + } + + if (destroy) { + /* IB contains a destroy msg, free the handle */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) + atomic_cmpxchg(&p->rdev->vce.handles[i], handle, 0); + + return 0; + } + + /* create or encode, validate the handle */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + if (atomic_read(&p->rdev->vce.handles[i]) == handle) + return 0; + } + + /* handle not found try to alloc a new one */ + for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) { + if (!atomic_cmpxchg(&p->rdev->vce.handles[i], 0, handle)) { + p->rdev->vce.filp[i] = p->filp; + return 0; + } + } + + DRM_ERROR("No more free VCE handles!\n"); + return -EINVAL; +} + +/** + * radeon_vce_semaphore_emit - emit a semaphore command + * + * @rdev: radeon_device pointer + * @ring: engine to use + * @semaphore: address of semaphore + * @emit_wait: true=emit wait, false=emit signal + * + */ +bool radeon_vce_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, VCE_CMD_SEMAPHORE); + radeon_ring_write(ring, (addr >> 3) & 0x000FFFFF); + radeon_ring_write(ring, (addr >> 23) & 0x000FFFFF); + radeon_ring_write(ring, 0x01003000 | (emit_wait ? 1 : 0)); + if (!emit_wait) + radeon_ring_write(ring, VCE_CMD_END); + + return true; +} + +/** + * radeon_vce_ib_execute - execute indirect buffer + * + * @rdev: radeon_device pointer + * @ib: the IB to execute + * + */ +void radeon_vce_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + radeon_ring_write(ring, VCE_CMD_IB); + radeon_ring_write(ring, ib->gpu_addr); + radeon_ring_write(ring, upper_32_bits(ib->gpu_addr)); + radeon_ring_write(ring, ib->length_dw); +} + +/** + * radeon_vce_fence_emit - add a fence command to the ring + * + * @rdev: radeon_device pointer + * @fence: the fence + * + */ +void radeon_vce_fence_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + uint32_t addr = rdev->fence_drv[fence->ring].gpu_addr; + + radeon_ring_write(ring, VCE_CMD_FENCE); + radeon_ring_write(ring, addr); + radeon_ring_write(ring, upper_32_bits(addr)); + radeon_ring_write(ring, fence->seq); + radeon_ring_write(ring, VCE_CMD_TRAP); + radeon_ring_write(ring, VCE_CMD_END); +} + +/** + * radeon_vce_ring_test - test if VCE ring is working + * + * @rdev: radeon_device pointer + * @ring: the engine to test on + * + */ +int radeon_vce_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + uint32_t rptr = vce_v1_0_get_rptr(rdev, ring); + unsigned i; + int r; + + r = radeon_ring_lock(rdev, ring, 16); + if (r) { + DRM_ERROR("radeon: vce failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + radeon_ring_write(ring, VCE_CMD_END); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (vce_v1_0_get_rptr(rdev, ring) != rptr) + 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\n", + ring->idx); + r = -ETIMEDOUT; + } + + return r; +} + +/** + * radeon_vce_ib_test - test if VCE IBs are working + * + * @rdev: radeon_device pointer + * @ring: the engine to test on + * + */ +int radeon_vce_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_fence *fence = NULL; + int r; + + r = radeon_vce_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_vce_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); + } else { + DRM_INFO("ib test on ring %d succeeded\n", ring->idx); + } +error: + radeon_fence_unref(&fence); + return r; +} diff --git a/drivers/gpu/drm/radeon/sid.h b/drivers/gpu/drm/radeon/sid.h index 9239a6d..683532f 100644 --- a/drivers/gpu/drm/radeon/sid.h +++ b/drivers/gpu/drm/radeon/sid.h @@ -1798,4 +1798,51 @@ #define DMA_PACKET_CONSTANT_FILL 0xd #define DMA_PACKET_NOP 0xf

+#define VCE_STATUS 0x20004 +#define VCE_VCPU_CNTL 0x20014 +#define VCE_CLK_EN (1 << 0) +#define VCE_VCPU_CACHE_OFFSET0 0x20024 +#define VCE_VCPU_CACHE_SIZE0 0x20028 +#define VCE_VCPU_CACHE_OFFSET1 0x2002c +#define VCE_VCPU_CACHE_SIZE1 0x20030 +#define VCE_VCPU_CACHE_OFFSET2 0x20034 +#define VCE_VCPU_CACHE_SIZE2 0x20038 +#define VCE_SOFT_RESET 0x20120 +#define VCE_ECPU_SOFT_RESET (1 << 0) +#define VCE_FME_SOFT_RESET (1 << 2) +#define VCE_RB_BASE_LO2 0x2016c +#define VCE_RB_BASE_HI2 0x20170 +#define VCE_RB_SIZE2 0x20174 +#define VCE_RB_RPTR2 0x20178 +#define VCE_RB_WPTR2 0x2017c +#define VCE_RB_BASE_LO 0x20180 +#define VCE_RB_BASE_HI 0x20184 +#define VCE_RB_SIZE 0x20188 +#define VCE_RB_RPTR 0x2018c +#define VCE_RB_WPTR 0x20190 +#define VCE_CLOCK_GATING_A 0x202f8 +#define VCE_CLOCK_GATING_B 0x202fc +#define VCE_UENC_CLOCK_GATING 0x205bc +#define VCE_UENC_REG_CLOCK_GATING 0x205c0 +#define VCE_FW_REG_STATUS 0x20e10 +# define VCE_FW_REG_STATUS_BUSY (1 << 0) +# define VCE_FW_REG_STATUS_PASS (1 << 3) +# define VCE_FW_REG_STATUS_DONE (1 << 11) +#define VCE_LMI_FW_START_KEYSEL 0x20e18 +#define VCE_LMI_FW_PERIODIC_CTRL 0x20e20 +#define VCE_LMI_CTRL2 0x20e74 +#define VCE_LMI_CTRL 0x20e98 +#define VCE_LMI_VM_CTRL 0x20ea0 +#define VCE_LMI_SWAP_CNTL 0x20eb4 +#define VCE_LMI_SWAP_CNTL1 0x20eb8 +#define VCE_LMI_CACHE_CTRL 0x20ef4 + +#define VCE_CMD_NO_OP 0x00000000 +#define VCE_CMD_END 0x00000001 +#define VCE_CMD_IB 0x00000002 +#define VCE_CMD_FENCE 0x00000003 +#define VCE_CMD_TRAP 0x00000004 +#define VCE_CMD_IB_AUTO 0x00000005 +#define VCE_CMD_SEMAPHORE 0x00000006 + #endif diff --git a/drivers/gpu/drm/radeon/vce_v1_0.c b/drivers/gpu/drm/radeon/vce_v1_0.c new file mode 100644 index 0000000..e0c3534 --- /dev/null +++ b/drivers/gpu/drm/radeon/vce_v1_0.c @@ -0,0 +1,187 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König christian.koenig@amd.com + */ + +#include <linux/firmware.h> +#include <drm/drmP.h> +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" + +/** + * vce_v1_0_get_rptr - get read pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware read pointer + */ +uint32_t vce_v1_0_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + return RREG32(VCE_RB_RPTR); + else + return RREG32(VCE_RB_RPTR2); +} + +/** + * vce_v1_0_get_wptr - get write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Returns the current hardware write pointer + */ +uint32_t vce_v1_0_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + return RREG32(VCE_RB_WPTR); + else + return RREG32(VCE_RB_WPTR2); +} + +/** + * vce_v1_0_set_wptr - set write pointer + * + * @rdev: radeon_device pointer + * @ring: radeon_ring pointer + * + * Commits the write pointer to the hardware + */ +void vce_v1_0_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + if (ring->idx == TN_RING_TYPE_VCE1_INDEX) + WREG32(VCE_RB_WPTR, ring->wptr); + else + WREG32(VCE_RB_WPTR2, ring->wptr); +} + +/** + * vce_v1_0_start - start VCE block + * + * @rdev: radeon_device pointer + * + * Setup and start the VCE block + */ +int vce_v1_0_start(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int i, j, r; + + /* set BUSY flag */ + WREG32_P(VCE_STATUS, 1, ~1); + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + WREG32(VCE_RB_RPTR, ring->rptr); + WREG32(VCE_RB_WPTR, ring->wptr); + WREG32(VCE_RB_BASE_LO, ring->gpu_addr); + WREG32(VCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32(VCE_RB_SIZE, ring->ring_size / 4); + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + WREG32(VCE_RB_RPTR2, ring->rptr); + WREG32(VCE_RB_WPTR2, ring->wptr); + WREG32(VCE_RB_BASE_LO2, ring->gpu_addr); + WREG32(VCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32(VCE_RB_SIZE2, ring->ring_size / 4); + + WREG32_P(VCE_VCPU_CNTL, VCE_CLK_EN, ~VCE_CLK_EN); + + WREG32_P(VCE_SOFT_RESET, + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET, ~( + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET)); + + mdelay(100); + + WREG32_P(VCE_SOFT_RESET, 0, ~( + VCE_ECPU_SOFT_RESET | + VCE_FME_SOFT_RESET)); + + for (i = 0; i < 10; ++i) { + uint32_t status; + for (j = 0; j < 100; ++j) { + status = RREG32(VCE_STATUS); + if (status & 2) + break; + mdelay(10); + } + r = 0; + if (status & 2) + break; + + DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n"); + WREG32_P(VCE_SOFT_RESET, VCE_ECPU_SOFT_RESET, ~VCE_ECPU_SOFT_RESET); + mdelay(10); + WREG32_P(VCE_SOFT_RESET, 0, ~VCE_ECPU_SOFT_RESET); + mdelay(10); + r = -1; + } + + /* clear BUSY flag */ + WREG32_P(VCE_STATUS, 0, ~1); + + if (r) { + DRM_ERROR("VCE not responding, giving up!!!\n"); + return r; + } + + return 0; +} + +int vce_v1_0_init(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int r; + + r = vce_v1_0_start(rdev); + if (r) + return r; + + ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX]; + ring->ready = true; + r = radeon_ring_test(rdev, TN_RING_TYPE_VCE1_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } + + ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX]; + ring->ready = true; + r = radeon_ring_test(rdev, TN_RING_TYPE_VCE2_INDEX, ring); + if (r) { + ring->ready = false; + return r; + } + + DRM_INFO("VCE initialized successfully.\n"); + + return 0; +} diff --git a/drivers/gpu/drm/radeon/vce_v2_0.c b/drivers/gpu/drm/radeon/vce_v2_0.c new file mode 100644 index 0000000..4911d1b --- /dev/null +++ b/drivers/gpu/drm/radeon/vce_v2_0.c @@ -0,0 +1,70 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * 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, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS 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. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * Authors: Christian König christian.koenig@amd.com + */ + +#include <linux/firmware.h> +#include <drm/drmP.h> +#include "radeon.h" +#include "radeon_asic.h" +#include "cikd.h" + +int vce_v2_0_resume(struct radeon_device *rdev) +{ + uint64_t addr = rdev->vce.gpu_addr; + uint32_t size; + + WREG32_P(VCE_CLOCK_GATING_A, 0, ~(1 << 16)); + WREG32_P(VCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000); + WREG32_P(VCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F); + WREG32(VCE_CLOCK_GATING_B, 0xf7); + + WREG32(VCE_LMI_CTRL, 0x00398000); + WREG32_P(VCE_LMI_CACHE_CTRL, 0x0, ~0x1); + WREG32(VCE_LMI_SWAP_CNTL, 0); + WREG32(VCE_LMI_SWAP_CNTL1, 0); + WREG32(VCE_LMI_VM_CTRL, 0); + + size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size); + WREG32(VCE_VCPU_CACHE_OFFSET0, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE0, size); + + addr += size; + size = RADEON_VCE_STACK_SIZE; + WREG32(VCE_VCPU_CACHE_OFFSET1, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE1, size); + + addr += size; + size = RADEON_VCE_HEAP_SIZE; + WREG32(VCE_VCPU_CACHE_OFFSET2, addr & 0x7fffffff); + WREG32(VCE_VCPU_CACHE_SIZE2, size); + + WREG32_P(VCE_LMI_CTRL2, 0x0, ~0x100); + + WREG32_P(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, + ~VCE_SYS_INT_TRAP_INTERRUPT_EN); + + return 0; +} diff --git a/include/uapi/drm/radeon_drm.h b/include/uapi/drm/radeon_drm.h index d9ea3a7..6493ca5 100644 --- a/include/uapi/drm/radeon_drm.h +++ b/include/uapi/drm/radeon_drm.h @@ -919,6 +919,7 @@ struct drm_radeon_gem_va { #define RADEON_CS_RING_COMPUTE 1 #define RADEON_CS_RING_DMA 2 #define RADEON_CS_RING_UVD 3 +#define RADEON_CS_RING_VCE 4 /* The third dword of RADEON_CHUNK_ID_FLAGS is a sint32 that sets the priority */ /* 0 = normal, + = higher priority, - = lower priority */

From: Christian König christian.koenig@amd.com

Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/radeon/radeon_kms.c | 3 +++ 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c index 0e078af..ea018d5 100644 --- a/drivers/gpu/drm/radeon/radeon_kms.c +++ b/drivers/gpu/drm/radeon/radeon_kms.c @@ -433,6 +433,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file case RADEON_CS_RING_UVD: *value = rdev->ring[R600_RING_TYPE_UVD_INDEX].ready; break; + case RADEON_CS_RING_VCE: + *value = rdev->ring[TN_RING_TYPE_VCE1_INDEX].ready; + break; default: return -EINVAL; }

From: Alex Deucher alexander.deucher@amd.com

Similar to uvd clock setting.

Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/radeon/radeon.h | 2 ++ 1 file changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index bd181ec..05fa29e 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1820,6 +1820,7 @@ struct radeon_asic { void (*set_pcie_lanes)(struct radeon_device *rdev, int lanes); void (*set_clock_gating)(struct radeon_device *rdev, int enable); int (*set_uvd_clocks)(struct radeon_device *rdev, u32 vclk, u32 dclk); + int (*set_vce_clocks)(struct radeon_device *rdev, u32 evclk, u32 ecclk); int (*get_temperature)(struct radeon_device *rdev); } pm; /* dynamic power management */ @@ -2681,6 +2682,7 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_set_pcie_lanes(rdev, l) (rdev)->asic->pm.set_pcie_lanes((rdev), (l)) #define radeon_set_clock_gating(rdev, e) (rdev)->asic->pm.set_clock_gating((rdev), (e)) #define radeon_set_uvd_clocks(rdev, v, d) (rdev)->asic->pm.set_uvd_clocks((rdev), (v), (d)) +#define radeon_set_vce_clocks(rdev, ev, ec) (rdev)->asic->pm.set_vce_clocks((rdev), (ev), (ec)) #define radeon_get_temperature(rdev) (rdev)->asic->pm.get_temperature((rdev)) #define radeon_set_surface_reg(rdev, r, f, p, o, s) ((rdev)->asic->surface.set_reg((rdev), (r), (f), (p), (o), (s))) #define radeon_clear_surface_reg(rdev, r) ((rdev)->asic->surface.clear_reg((rdev), (r)))

From: Alex Deucher alexander.deucher@amd.com

Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/radeon/radeon.h | 12 ++++++++++++ drivers/gpu/drm/radeon/radeon_pm.c | 8 ++++++++ 2 files changed, 20 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 05fa29e..7edd13a 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1257,6 +1257,15 @@ enum radeon_dpm_event_src { RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL = 4 };

+enum radeon_vce_level { + RADEON_VCE_LEVEL_AC_ALL = 0, /* AC, All cases */ + RADEON_VCE_LEVEL_DC_EE = 1, /* DC, entropy encoding */ + RADEON_VCE_LEVEL_DC_LL_LOW = 2, /* DC, low latency queue, res <= 720 */ + RADEON_VCE_LEVEL_DC_LL_HIGH = 3, /* DC, low latency queue, 1080 >= res > 720 */ + RADEON_VCE_LEVEL_DC_GP_LOW = 4, /* DC, general purpose queue, res <= 720 */ + RADEON_VCE_LEVEL_DC_GP_HIGH = 5, /* DC, general purpose queue, 1080 >= res > 720 */ +}; + struct radeon_ps { u32 caps; /* vbios flags */ u32 class; /* vbios flags */ @@ -1267,6 +1276,8 @@ struct radeon_ps { /* VCE clocks */ u32 evclk; u32 ecclk; + bool vce_active; + enum radeon_vce_level vce_level; /* asic priv */ void *ps_priv; }; @@ -1478,6 +1489,7 @@ struct radeon_dpm { /* special states active */ bool thermal_active; bool uvd_active; + bool vce_active; /* thermal handling */ struct radeon_dpm_thermal thermal; /* forced levels */ diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c index 8e8153e..a4687e7 100644 --- a/drivers/gpu/drm/radeon/radeon_pm.c +++ b/drivers/gpu/drm/radeon/radeon_pm.c @@ -826,6 +826,9 @@ static void radeon_dpm_change_power_state_locked(struct radeon_device *rdev)

/* no need to reprogram if nothing changed unless we are on BTC+ */ if (rdev->pm.dpm.current_ps == rdev->pm.dpm.requested_ps) { + /* vce just modifies an existing state so force a change */ + if (ps->vce_active != rdev->pm.dpm.vce_active) + goto force; if ((rdev->family < CHIP_BARTS) || (rdev->flags & RADEON_IS_IGP)) { /* for pre-BTC and APUs if the num crtcs changed but state is the same, * all we need to do is update the display configuration. @@ -862,16 +865,21 @@ static void radeon_dpm_change_power_state_locked(struct radeon_device *rdev) } }

+force: if (radeon_dpm == 1) { printk("switching from power state:\n"); radeon_dpm_print_power_state(rdev, rdev->pm.dpm.current_ps); printk("switching to power state:\n"); radeon_dpm_print_power_state(rdev, rdev->pm.dpm.requested_ps); } + mutex_lock(&rdev->ddev->struct_mutex); down_write(&rdev->pm.mclk_lock); mutex_lock(&rdev->ring_lock);

+ /* update whether vce is active */ + ps->vce_active = rdev->pm.dpm.vce_active; + ret = radeon_dpm_pre_set_power_state(rdev); if (ret) goto done;

From: Alex Deucher alexander.deucher@amd.com

Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/radeon/cik.c | 35 +++++++++++++++++++++++++++++++++++ drivers/gpu/drm/radeon/cikd.h | 6 ++++++ drivers/gpu/drm/radeon/radeon_asic.c | 2 ++ drivers/gpu/drm/radeon/radeon_asic.h | 1 + 4 files changed, 44 insertions(+)

diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c index be6eb4d..ecb16b1 100644 --- a/drivers/gpu/drm/radeon/cik.c +++ b/drivers/gpu/drm/radeon/cik.c @@ -8925,6 +8925,41 @@ int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) return r; }

+int cik_set_vce_clocks(struct radeon_device *rdev, u32 evclk, u32 ecclk) +{ + int r, i; + struct atom_clock_dividers dividers; + u32 tmp; + + r = radeon_atom_get_clock_dividers(rdev, COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, + ecclk, false, &dividers); + if (r) + return r; + + for (i = 0; i < 100; i++) { + if (RREG32_SMC(CG_ECLK_STATUS) & ECLK_STATUS) + break; + mdelay(10); + } + if (i == 100) + return -ETIMEDOUT; + + tmp = RREG32_SMC(CG_ECLK_CNTL); + tmp &= ~(ECLK_DIR_CNTL_EN|ECLK_DIVIDER_MASK); + tmp |= dividers.post_divider; + WREG32_SMC(CG_ECLK_CNTL, tmp); + + for (i = 0; i < 100; i++) { + if (RREG32_SMC(CG_ECLK_STATUS) & ECLK_STATUS) + break; + mdelay(10); + } + if (i == 100) + return -ETIMEDOUT; + + return 0; +} + 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/cikd.h b/drivers/gpu/drm/radeon/cikd.h index 459ae02..ee16380 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -203,6 +203,12 @@ #define CTF_TEMP_MASK 0x0003fe00 #define CTF_TEMP_SHIFT 9

+#define CG_ECLK_CNTL 0xC05000AC +# define ECLK_DIVIDER_MASK 0x7f +# define ECLK_DIR_CNTL_EN (1 << 8) +#define CG_ECLK_STATUS 0xC05000B0 +# define ECLK_STATUS (1 << 0) + #define CG_SPLL_FUNC_CNTL 0xC0500140 #define SPLL_RESET (1 << 0) #define SPLL_PWRON (1 << 1) diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c index e9eaba99..db1df9a 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.c +++ b/drivers/gpu/drm/radeon/radeon_asic.c @@ -2076,6 +2076,7 @@ static struct radeon_asic ci_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &cik_set_uvd_clocks, + .set_vce_clocks = &cik_set_vce_clocks, .get_temperature = &ci_get_temp, }, .dpm = { @@ -2180,6 +2181,7 @@ static struct radeon_asic kv_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &cik_set_uvd_clocks, + .set_vce_clocks = &cik_set_vce_clocks, .get_temperature = &kv_get_temp, }, .dpm = { diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h index 2017efd..385bb6c 100644 --- a/drivers/gpu/drm/radeon/radeon_asic.h +++ b/drivers/gpu/drm/radeon/radeon_asic.h @@ -715,6 +715,7 @@ 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_set_vce_clocks(struct radeon_device *rdev, u32 evclk, u32 ecclk); void cik_sdma_fence_ring_emit(struct radeon_device *rdev, struct radeon_fence *fence); bool cik_sdma_semaphore_ring_emit(struct radeon_device *rdev,

From: Alex Deucher alexander.deucher@amd.com

VCE dpm dynamically adjusts the uvd clocks on demand.

Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/radeon/ci_dpm.c | 1 + 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/radeon/ci_dpm.c b/drivers/gpu/drm/radeon/ci_dpm.c index c91d0ee..6669d32 100644 --- a/drivers/gpu/drm/radeon/ci_dpm.c +++ b/drivers/gpu/drm/radeon/ci_dpm.c @@ -5153,6 +5153,7 @@ int ci_dpm_init(struct radeon_device *rdev) pi->caps_sclk_throttle_low_notification = false;

pi->caps_uvd_dpm = true; + pi->caps_vce_dpm = true;

ci_get_leakage_voltages(rdev); ci_patch_dependency_tables_with_leakage(rdev);

From: Alex Deucher alexander.deucher@amd.com

enable vce states when vce is active. When vce is active, it adjusts the currently selected state (performance, battery, uvd, etc.)

v2: add code comments

Signed-off-by: Alex Deucher alexander.deucher@amd.com Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/radeon/radeon.h | 3 ++ drivers/gpu/drm/radeon/radeon_cs.c | 3 ++ drivers/gpu/drm/radeon/radeon_pm.c | 17 ++++++++++ drivers/gpu/drm/radeon/radeon_vce.c | 62 +++++++++++++++++++++++++++++++++++++ 4 files changed, 85 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h index 7e34ecd..63777e4 100644 --- a/drivers/gpu/drm/radeon/radeon.h +++ b/drivers/gpu/drm/radeon/radeon.h @@ -1516,6 +1516,7 @@ struct radeon_dpm { };

void radeon_dpm_enable_uvd(struct radeon_device *rdev, bool enable); +void radeon_dpm_enable_vce(struct radeon_device *rdev, bool enable);

struct radeon_pm { struct mutex mutex; @@ -1636,6 +1637,7 @@ struct radeon_vce { unsigned fb_version; atomic_t handles[RADEON_MAX_VCE_HANDLES]; struct drm_file *filp[RADEON_MAX_VCE_HANDLES]; + struct delayed_work idle_work; };

int radeon_vce_init(struct radeon_device *rdev); @@ -1647,6 +1649,7 @@ int radeon_vce_get_create_msg(struct radeon_device *rdev, int ring, int radeon_vce_get_destroy_msg(struct radeon_device *rdev, int ring, uint32_t handle, struct radeon_fence **fence); void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp); +void radeon_vce_note_usage(struct radeon_device *rdev); int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi); int radeon_vce_cs_parse(struct radeon_cs_parser *p); bool radeon_vce_semaphore_emit(struct radeon_device *rdev, diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 701ee79..f28a8d8 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -347,6 +347,9 @@ static int radeon_cs_ib_chunk(struct radeon_device *rdev,

if (parser->ring == R600_RING_TYPE_UVD_INDEX) radeon_uvd_note_usage(rdev); + else if ((parser->ring == TN_RING_TYPE_VCE1_INDEX) || + (parser->ring == TN_RING_TYPE_VCE2_INDEX)) + radeon_vce_note_usage(rdev);

radeon_cs_sync_rings(parser); r = radeon_ib_schedule(rdev, &parser->ib, NULL); diff --git a/drivers/gpu/drm/radeon/radeon_pm.c b/drivers/gpu/drm/radeon/radeon_pm.c index a4687e7..4ad9af9 100644 --- a/drivers/gpu/drm/radeon/radeon_pm.c +++ b/drivers/gpu/drm/radeon/radeon_pm.c @@ -968,6 +968,23 @@ void radeon_dpm_enable_uvd(struct radeon_device *rdev, bool enable) } }

+void radeon_dpm_enable_vce(struct radeon_device *rdev, bool enable) +{ + if (enable) { + mutex_lock(&rdev->pm.mutex); + rdev->pm.dpm.vce_active = true; + /* XXX select vce level based on ring/task */ + rdev->pm.dpm.vce_level = RADEON_VCE_LEVEL_AC_ALL; + mutex_unlock(&rdev->pm.mutex); + } else { + mutex_lock(&rdev->pm.mutex); + rdev->pm.dpm.vce_active = false; + mutex_unlock(&rdev->pm.mutex); + } + + radeon_pm_compute_clocks(rdev); +} + static void radeon_pm_suspend_old(struct radeon_device *rdev) { mutex_lock(&rdev->pm.mutex); diff --git a/drivers/gpu/drm/radeon/radeon_vce.c b/drivers/gpu/drm/radeon/radeon_vce.c index f46563b..d130432 100644 --- a/drivers/gpu/drm/radeon/radeon_vce.c +++ b/drivers/gpu/drm/radeon/radeon_vce.c @@ -34,11 +34,16 @@ #include "radeon_asic.h" #include "sid.h"

+/* 1 second timeout */ +#define VCE_IDLE_TIMEOUT_MS 1000 + /* Firmware Names */ #define FIRMWARE_BONAIRE "radeon/BONAIRE_vce.bin"

MODULE_FIRMWARE(FIRMWARE_BONAIRE);

+static void radeon_vce_idle_work_handler(struct work_struct *work); + /** * radeon_vce_init - allocate memory, load vce firmware * @@ -55,6 +60,8 @@ int radeon_vce_init(struct radeon_device *rdev) uint8_t start, mid, end; int i, r;

+ INIT_DELAYED_WORK(&rdev->vce.idle_work, radeon_vce_idle_work_handler); + switch (rdev->family) { case CHIP_BONAIRE: case CHIP_KAVERI: @@ -220,6 +227,59 @@ int radeon_vce_resume(struct radeon_device *rdev) }

/** + * radeon_vce_idle_work_handler - power off VCE + * + * @work: pointer to work structure + * + * power of VCE when it's not used any more + */ +static void radeon_vce_idle_work_handler(struct work_struct *work) +{ + struct radeon_device *rdev = + container_of(work, struct radeon_device, vce.idle_work.work); + + if ((radeon_fence_count_emitted(rdev, TN_RING_TYPE_VCE1_INDEX) == 0) && + (radeon_fence_count_emitted(rdev, TN_RING_TYPE_VCE2_INDEX) == 0)) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_dpm_enable_vce(rdev, false); + } else { + radeon_set_vce_clocks(rdev, 0, 0); + } + } else { + schedule_delayed_work(&rdev->vce.idle_work, + msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); + } +} + +/** + * radeon_vce_note_usage - power up VCE + * + * @rdev: radeon_device pointer + * + * Make sure VCE is powerd up when we want to use it + */ +void radeon_vce_note_usage(struct radeon_device *rdev) +{ + bool streams_changed = false; + bool set_clocks = !cancel_delayed_work_sync(&rdev->vce.idle_work); + set_clocks &= schedule_delayed_work(&rdev->vce.idle_work, + msecs_to_jiffies(VCE_IDLE_TIMEOUT_MS)); + + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + /* XXX figure out if the streams changed */ + streams_changed = false; + } + + if (set_clocks || streams_changed) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + radeon_dpm_enable_vce(rdev, true); + } else { + radeon_set_vce_clocks(rdev, 53300, 40000); + } + } +} + +/** * radeon_vce_free_handles - free still open VCE handles * * @rdev: radeon_device pointer @@ -235,6 +295,8 @@ void radeon_vce_free_handles(struct radeon_device *rdev, struct drm_file *filp) if (!handle || rdev->vce.filp[i] != filp) continue;

+ radeon_vce_note_usage(rdev); + r = radeon_vce_get_destroy_msg(rdev, TN_RING_TYPE_VCE1_INDEX, handle, NULL); if (r)

From: Alex Deucher alexander.deucher@amd.com

Signed-off-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/radeon/cikd.h | 10 ++++ drivers/gpu/drm/radeon/vce_v2_0.c | 111 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 121 insertions(+)

diff --git a/drivers/gpu/drm/radeon/cikd.h b/drivers/gpu/drm/radeon/cikd.h index ee16380..2138732 100644 --- a/drivers/gpu/drm/radeon/cikd.h +++ b/drivers/gpu/drm/radeon/cikd.h @@ -2029,8 +2029,18 @@ #define VCE_RB_RPTR 0x2018c #define VCE_RB_WPTR 0x20190 #define VCE_CLOCK_GATING_A 0x202f8 +# define CGC_CLK_GATE_DLY_TIMER_MASK (0xf << 0) +# define CGC_CLK_GATE_DLY_TIMER(x) ((x) << 0) +# define CGC_CLK_GATER_OFF_DLY_TIMER_MASK (0xff << 4) +# define CGC_CLK_GATER_OFF_DLY_TIMER(x) ((x) << 4) +# define CGC_UENC_WAIT_AWAKE (1 << 18) #define VCE_CLOCK_GATING_B 0x202fc +#define VCE_CGTT_CLK_OVERRIDE 0x207a0 #define VCE_UENC_CLOCK_GATING 0x207bc +# define CLOCK_ON_DELAY_MASK (0xf << 0) +# define CLOCK_ON_DELAY(x) ((x) << 0) +# define CLOCK_OFF_DELAY_MASK (0xff << 4) +# define CLOCK_OFF_DELAY(x) ((x) << 4) #define VCE_UENC_REG_CLOCK_GATING 0x207c0 #define VCE_SYS_INT_EN 0x21300 # define VCE_SYS_INT_TRAP_INTERRUPT_EN (1 << 3) diff --git a/drivers/gpu/drm/radeon/vce_v2_0.c b/drivers/gpu/drm/radeon/vce_v2_0.c index 4911d1b..1ac7bb8 100644 --- a/drivers/gpu/drm/radeon/vce_v2_0.c +++ b/drivers/gpu/drm/radeon/vce_v2_0.c @@ -31,6 +31,115 @@ #include "radeon_asic.h" #include "cikd.h"

+static void vce_v2_0_set_sw_cg(struct radeon_device *rdev, bool gated) +{ + u32 tmp; + + if (gated) { + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0xe70000; + WREG32(VCE_CLOCK_GATING_B, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp |= 0xff000000; + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp &= ~0x3fc; + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + + WREG32(VCE_CGTT_CLK_OVERRIDE, 0); + } else { + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0xe7; + tmp &= ~0xe70000; + WREG32(VCE_CLOCK_GATING_B, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp |= 0x1fe000; + tmp &= ~0xff000000; + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp |= 0x3fc; + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + } +} + +static void vce_v2_0_set_dyn_cg(struct radeon_device *rdev, bool gated) +{ + u32 orig, tmp; + + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp &= ~0x00060006; + if (gated) { + tmp |= 0xe10000; + } else { + tmp |= 0xe1; + tmp &= ~0xe10000; + } + WREG32(VCE_CLOCK_GATING_B, tmp); + + orig = tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp &= ~0x1fe000; + tmp &= ~0xff000000; + if (tmp != orig) + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + orig = tmp = RREG32(VCE_UENC_REG_CLOCK_GATING); + tmp &= ~0x3fc; + if (tmp != orig) + WREG32(VCE_UENC_REG_CLOCK_GATING, tmp); + + if (gated) + WREG32(VCE_CGTT_CLK_OVERRIDE, 0); +} + +static void vce_v2_0_disable_cg(struct radeon_device *rdev) +{ + WREG32(VCE_CGTT_CLK_OVERRIDE, 7); +} + +void vce_v2_0_enable_mgcg(struct radeon_device *rdev, bool enable) +{ + bool sw_cg = false; + + if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_VCE_MGCG)) { + if (sw_cg) + vce_v2_0_set_sw_cg(rdev, true); + else + vce_v2_0_set_dyn_cg(rdev, true); + } else { + vce_v2_0_disable_cg(rdev); + + if (sw_cg) + vce_v2_0_set_sw_cg(rdev, false); + else + vce_v2_0_set_dyn_cg(rdev, false); + } +} + +static void vce_v2_0_init_cg(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32(VCE_CLOCK_GATING_A); + tmp &= ~(CGC_CLK_GATE_DLY_TIMER_MASK | CGC_CLK_GATER_OFF_DLY_TIMER_MASK); + tmp |= (CGC_CLK_GATE_DLY_TIMER(0) | CGC_CLK_GATER_OFF_DLY_TIMER(4)); + tmp |= CGC_UENC_WAIT_AWAKE; + WREG32(VCE_CLOCK_GATING_A, tmp); + + tmp = RREG32(VCE_UENC_CLOCK_GATING); + tmp &= ~(CLOCK_ON_DELAY_MASK | CLOCK_OFF_DELAY_MASK); + tmp |= (CLOCK_ON_DELAY(0) | CLOCK_OFF_DELAY(4)); + WREG32(VCE_UENC_CLOCK_GATING, tmp); + + tmp = RREG32(VCE_CLOCK_GATING_B); + tmp |= 0x10; + tmp &= ~0x100000; + WREG32(VCE_CLOCK_GATING_B, tmp); +} + int vce_v2_0_resume(struct radeon_device *rdev) { uint64_t addr = rdev->vce.gpu_addr; @@ -66,5 +175,7 @@ int vce_v2_0_resume(struct radeon_device *rdev) WREG32_P(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, ~VCE_SYS_INT_TRAP_INTERRUPT_EN);

+ vce_v2_0_init_cg(rdev); + return 0; }