Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / gpu / drm / radeon / radeon_vce.c
diff --git a/kernel/drivers/gpu/drm/radeon/radeon_vce.c b/kernel/drivers/gpu/drm/radeon/radeon_vce.c
new file mode 100644 (file)
index 0000000..0de5711
--- /dev/null
@@ -0,0 +1,804 @@
+/*
+ * 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"
+
+/* 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
+ *
+ * @rdev: radeon_device pointer
+ *
+ * First step to get VCE online, allocate memory and load the firmware
+ */
+int radeon_vce_init(struct radeon_device *rdev)
+{
+       static const char *fw_version = "[ATI LIB=VCEFW,";
+       static const char *fb_version = "[ATI LIB=VCEFWSTATS,";
+       unsigned long size;
+       const char *fw_name, *c;
+       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:
+       case CHIP_KABINI:
+       case CHIP_HAWAII:
+       case CHIP_MULLINS:
+               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;
+       }
+
+       /* search for firmware version */
+
+       size = rdev->vce_fw->size - strlen(fw_version) - 9;
+       c = rdev->vce_fw->data;
+       for (;size > 0; --size, ++c)
+               if (strncmp(c, fw_version, strlen(fw_version)) == 0)
+                       break;
+
+       if (size == 0)
+               return -EINVAL;
+
+       c += strlen(fw_version);
+       if (sscanf(c, "%2hhd.%2hhd.%2hhd]", &start, &mid, &end) != 3)
+               return -EINVAL;
+
+       /* search for feedback version */
+
+       size = rdev->vce_fw->size - strlen(fb_version) - 3;
+       c = rdev->vce_fw->data;
+       for (;size > 0; --size, ++c)
+               if (strncmp(c, fb_version, strlen(fb_version)) == 0)
+                       break;
+
+       if (size == 0)
+               return -EINVAL;
+
+       c += strlen(fb_version);
+       if (sscanf(c, "%2u]", &rdev->vce.fb_version) != 1)
+               return -EINVAL;
+
+       DRM_INFO("Found VCE firmware/feedback version %hhd.%hhd.%hhd / %d!\n",
+                start, mid, end, rdev->vce.fb_version);
+
+       rdev->vce.fw_version = (start << 24) | (mid << 16) | (end << 8);
+
+       /* we can only work with this fw version for now */
+       if (rdev->vce.fw_version != ((40 << 24) | (2 << 16) | (2 << 8)))
+               return -EINVAL;
+
+       /* allocate firmware, stack and heap BO */
+
+       size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->size) +
+              RADEON_VCE_STACK_SIZE + RADEON_VCE_HEAP_SIZE;
+       r = radeon_bo_create(rdev, size, PAGE_SIZE, true,
+                            RADEON_GEM_DOMAIN_VRAM, 0, NULL, NULL,
+                            &rdev->vce.vcpu_bo);
+       if (r) {
+               dev_err(rdev->dev, "(%d) failed to allocate VCE bo\n", r);
+               return r;
+       }
+
+       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);
+       radeon_bo_unreserve(rdev->vce.vcpu_bo);
+       if (r) {
+               radeon_bo_unref(&rdev->vce.vcpu_bo);
+               dev_err(rdev->dev, "(%d) VCE bo pin failed\n", 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)
+{
+       if (rdev->vce.vcpu_bo == NULL)
+               return;
+
+       radeon_bo_unref(&rdev->vce.vcpu_bo);
+
+       release_firmware(rdev->vce_fw);
+}
+
+/**
+ * radeon_vce_suspend - unpin VCE fw memory
+ *
+ * @rdev: radeon_device pointer
+ *
+ */
+int radeon_vce_suspend(struct radeon_device *rdev)
+{
+       int i;
+
+       if (rdev->vce.vcpu_bo == NULL)
+               return 0;
+
+       for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i)
+               if (atomic_read(&rdev->vce.handles[i]))
+                       break;
+
+       if (i == RADEON_MAX_VCE_HANDLES)
+               return 0;
+
+       /* TODO: suspending running encoding sessions isn't supported */
+       return -EINVAL;
+}
+
+/**
+ * radeon_vce_resume - pin VCE fw memory
+ *
+ * @rdev: radeon_device pointer
+ *
+ */
+int radeon_vce_resume(struct radeon_device *rdev)
+{
+       void *cpu_addr;
+       int r;
+
+       if (rdev->vce.vcpu_bo == NULL)
+               return -EINVAL;
+
+       r = radeon_bo_reserve(rdev->vce.vcpu_bo, false);
+       if (r) {
+               dev_err(rdev->dev, "(%d) failed to reserve VCE bo\n", r);
+               return r;
+       }
+
+       r = radeon_bo_kmap(rdev->vce.vcpu_bo, &cpu_addr);
+       if (r) {
+               radeon_bo_unreserve(rdev->vce.vcpu_bo);
+               dev_err(rdev->dev, "(%d) VCE map failed\n", r);
+               return r;
+       }
+
+       memcpy(cpu_addr, rdev->vce_fw->data, rdev->vce_fw->size);
+
+       radeon_bo_kunmap(rdev->vce.vcpu_bo);
+
+       radeon_bo_unreserve(rdev->vce.vcpu_bo);
+
+       return 0;
+}
+
+/**
+ * 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
+ * @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;
+
+               radeon_vce_note_usage(rdev);
+
+               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, false);
+       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, false);
+       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
+ * @size: size of checker for relocation buffer
+ *
+ * Patch relocation inside command stream with real buffer address
+ */
+int radeon_vce_cs_reloc(struct radeon_cs_parser *p, int lo, int hi,
+                       unsigned size)
+{
+       struct radeon_cs_chunk *relocs_chunk;
+       struct radeon_bo_list *reloc;
+       uint64_t start, end, offset;
+       unsigned idx;
+
+       relocs_chunk = p->chunk_relocs;
+       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;
+       }
+
+       reloc = &p->relocs[(idx / 4)];
+       start = reloc->gpu_offset;
+       end = start + radeon_bo_size(reloc->robj);
+       start += offset;
+
+       p->ib.ptr[lo] = start & 0xFFFFFFFF;
+       p->ib.ptr[hi] = start >> 32;
+
+       if (end <= start) {
+               DRM_ERROR("invalid reloc offset %llX!\n", offset);
+               return -EINVAL;
+       }
+       if ((end - start) < size) {
+               DRM_ERROR("buffer to small (%d / %d)!\n",
+                       (unsigned)(end - start), size);
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+/**
+ * radeon_vce_validate_handle - validate stream handle
+ *
+ * @p: parser context
+ * @handle: handle to validate
+ * @allocated: allocated a new handle?
+ *
+ * Validates the handle and return the found session index or -EINVAL
+ * we we don't have another free session index.
+ */
+static int radeon_vce_validate_handle(struct radeon_cs_parser *p,
+                                     uint32_t handle, bool *allocated)
+{
+       unsigned i;
+
+       *allocated = false;
+
+       /* validate the handle */
+       for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i) {
+               if (atomic_read(&p->rdev->vce.handles[i]) == handle) {
+                       if (p->rdev->vce.filp[i] != p->filp) {
+                               DRM_ERROR("VCE handle collision detected!\n");
+                               return -EINVAL;
+                       }
+                       return i;
+               }
+       }
+
+       /* 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;
+                       p->rdev->vce.img_size[i] = 0;
+                       *allocated = true;
+                       return i;
+               }
+       }
+
+       DRM_ERROR("No more free VCE handles!\n");
+       return -EINVAL;
+}
+
+/**
+ * radeon_vce_cs_parse - parse and validate the command stream
+ *
+ * @p: parser context
+ *
+ */
+int radeon_vce_cs_parse(struct radeon_cs_parser *p)
+{
+       int session_idx = -1;
+       bool destroyed = false, created = false, allocated = false;
+       uint32_t tmp, handle = 0;
+       uint32_t *size = &tmp;
+       int i, r = 0;
+
+       while (p->idx < p->chunk_ib->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);
+                       r = -EINVAL;
+                       goto out;
+               }
+
+               if (destroyed) {
+                       DRM_ERROR("No other command allowed after destroy!\n");
+                       r = -EINVAL;
+                       goto out;
+               }
+
+               switch (cmd) {
+               case 0x00000001: // session
+                       handle = radeon_get_ib_value(p, p->idx + 2);
+                       session_idx = radeon_vce_validate_handle(p, handle,
+                                                                &allocated);
+                       if (session_idx < 0)
+                               return session_idx;
+                       size = &p->rdev->vce.img_size[session_idx];
+                       break;
+
+               case 0x00000002: // task info
+                       break;
+
+               case 0x01000001: // create
+                       created = true;
+                       if (!allocated) {
+                               DRM_ERROR("Handle already in use!\n");
+                               r = -EINVAL;
+                               goto out;
+                       }
+
+                       *size = radeon_get_ib_value(p, p->idx + 8) *
+                               radeon_get_ib_value(p, p->idx + 10) *
+                               8 * 3 / 2;
+                       break;
+
+               case 0x04000001: // config extension
+               case 0x04000002: // pic control
+               case 0x04000005: // rate control
+               case 0x04000007: // motion estimation
+               case 0x04000008: // rdo
+               case 0x04000009: // vui
+                       break;
+
+               case 0x03000001: // encode
+                       r = radeon_vce_cs_reloc(p, p->idx + 10, p->idx + 9,
+                                               *size);
+                       if (r)
+                               goto out;
+
+                       r = radeon_vce_cs_reloc(p, p->idx + 12, p->idx + 11,
+                                               *size / 3);
+                       if (r)
+                               goto out;
+                       break;
+
+               case 0x02000001: // destroy
+                       destroyed = true;
+                       break;
+
+               case 0x05000001: // context buffer
+                       r = radeon_vce_cs_reloc(p, p->idx + 3, p->idx + 2,
+                                               *size * 2);
+                       if (r)
+                               goto out;
+                       break;
+
+               case 0x05000004: // video bitstream buffer
+                       tmp = radeon_get_ib_value(p, p->idx + 4);
+                       r = radeon_vce_cs_reloc(p, p->idx + 3, p->idx + 2,
+                                               tmp);
+                       if (r)
+                               goto out;
+                       break;
+
+               case 0x05000005: // feedback buffer
+                       r = radeon_vce_cs_reloc(p, p->idx + 3, p->idx + 2,
+                                               4096);
+                       if (r)
+                               goto out;
+                       break;
+
+               default:
+                       DRM_ERROR("invalid VCE command (0x%x)!\n", cmd);
+                       r = -EINVAL;
+                       goto out;
+               }
+
+               if (session_idx == -1) {
+                       DRM_ERROR("no session command at start of IB\n");
+                       r = -EINVAL;
+                       goto out;
+               }
+
+               p->idx += len / 4;
+       }
+
+       if (allocated && !created) {
+               DRM_ERROR("New session without create command!\n");
+               r = -ENOENT;
+       }
+
+out:
+       if ((!r && destroyed) || (r && allocated)) {
+               /*
+                * IB contains a destroy msg or we have allocated an
+                * handle and got an error, anyway free the handle
+                */
+               for (i = 0; i < RADEON_MAX_VCE_HANDLES; ++i)
+                       atomic_cmpxchg(&p->rdev->vce.handles[i], handle, 0);
+       }
+
+       return r;
+}
+
+/**
+ * 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];
+       uint64_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, false);
+
+       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;
+}