/**
* uvd_v6_0_ring_test_ring - register write test
*
* @ring: amdgpu_ring pointer
*
* Test if we can successfully write to the context register
*/
static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
uint32_t tmp = 0;
unsigned i;
int r;
WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
ring->idx, r);
return r;
}
amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
for (i = 0; i < adev->usec_timeout; i++) {
tmp = RREG32(mmUVD_CONTEXT_ID);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_INFO("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
ring->idx, tmp);
r = -EINVAL;
}
return r;
}
/**
* uvd_v6_0_ring_test_ring - register write test
*
* @ring: amdgpu_ring pointer
*
* Test if we can successfully write to the context register
*/
static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
uint32_t tmp = 0;
unsigned i;
int r;
WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r)
return r;
amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
for (i = 0; i < adev->usec_timeout; i++) {
tmp = RREG32(mmUVD_CONTEXT_ID);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
{
signed long r;
uint32_t val;
struct dma_fence *f;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
BUG_ON(!ring->funcs->emit_rreg);
mutex_lock(&adev->virt.lock_kiq);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_rreg(ring, reg);
amdgpu_fence_emit(ring, &f);
amdgpu_ring_commit(ring);
mutex_unlock(&adev->virt.lock_kiq);
r = dma_fence_wait(f, false);
if (r)
DRM_ERROR("wait for kiq fence error: %ld.\n", r);
dma_fence_put(f);
val = adev->wb.wb[adev->virt.reg_val_offs];
return val;
}
uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
{
signed long r;
unsigned long flags;
uint32_t val, seq;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
BUG_ON(!ring->funcs->emit_rreg);
spin_lock_irqsave(&kiq->ring_lock, flags);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_rreg(ring, reg);
amdgpu_fence_emit_polling(ring, &seq);
amdgpu_ring_commit(ring);
spin_unlock_irqrestore(&kiq->ring_lock, flags);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
if (r < 1) {
DRM_ERROR("wait for kiq fence error: %ld\n", r);
return ~0;
}
val = adev->wb.wb[adev->virt.reg_val_offs];
return val;
}
/**
* uvd_v5_0_hw_init - start and test UVD block
*
* @adev: amdgpu_device pointer
*
* Initialize the hardware, boot up the VCPU and do some testing
*/
static int uvd_v5_0_hw_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring = &adev->uvd.ring;
uint32_t tmp;
int r;
/* raise clocks while booting up the VCPU */
amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
r = uvd_v5_0_start(adev);
if (r)
goto done;
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
goto done;
}
tmp = PACKET0(mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
/* Clear timeout status bits */
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_TIMEOUT_STATUS, 0));
amdgpu_ring_write(ring, 0x8);
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_CNTL, 0));
amdgpu_ring_write(ring, 3);
amdgpu_ring_commit(ring);
done:
/* lower clocks again */
amdgpu_asic_set_uvd_clocks(adev, 0, 0);
if (!r)
DRM_INFO("UVD initialized successfully.\n");
return r;
}
/**
* uvd_v5_0_hw_init - start and test UVD block
*
* @adev: amdgpu_device pointer
*
* Initialize the hardware, boot up the VCPU and do some testing
*/
static int uvd_v5_0_hw_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring = &adev->uvd.ring;
uint32_t tmp;
int r;
amdgpu_asic_set_uvd_clocks(adev, 10000, 10000);
uvd_v5_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
uvd_v5_0_enable_mgcg(adev, true);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
goto done;
}
tmp = PACKET0(mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
/* Clear timeout status bits */
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_TIMEOUT_STATUS, 0));
amdgpu_ring_write(ring, 0x8);
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_CNTL, 0));
amdgpu_ring_write(ring, 3);
amdgpu_ring_commit(ring);
done:
if (!r)
DRM_INFO("UVD initialized successfully.\n");
return r;
}
/**
* cik_sdma_ring_test_ring - simple async dma engine test
*
* @ring: amdgpu_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.
*/
static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
unsigned i;
unsigned index;
int r;
u32 tmp;
u64 gpu_addr;
r = amdgpu_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
r = amdgpu_ring_alloc(ring, 5);
if (r) {
DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
amdgpu_wb_free(adev, index);
return r;
}
amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0));
amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
amdgpu_ring_write(ring, upper_32_bits(gpu_addr));
amdgpu_ring_write(ring, 1); /* number of DWs to follow */
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
for (i = 0; i < adev->usec_timeout; i++) {
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
ring->idx, tmp);
r = -EINVAL;
}
amdgpu_wb_free(adev, index);
return r;
}
void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
{
signed long r, cnt = 0;
unsigned long flags;
uint32_t seq;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
BUG_ON(!ring->funcs->emit_wreg);
spin_lock_irqsave(&kiq->ring_lock, flags);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_wreg(ring, reg, v);
amdgpu_fence_emit_polling(ring, &seq);
amdgpu_ring_commit(ring);
spin_unlock_irqrestore(&kiq->ring_lock, flags);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
/* don't wait anymore for gpu reset case because this way may
* block gpu_recover() routine forever, e.g. this virt_kiq_rreg
* is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
* never return if we keep waiting in virt_kiq_rreg, which cause
* gpu_recover() hang there.
*
* also don't wait anymore for IRQ context
* */
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}
if (cnt > MAX_KIQ_REG_TRY)
goto failed_kiq_write;
return;
failed_kiq_write:
pr_err("failed to write reg:%x\n", reg);
}
/**
* amdgpu_ring_restore - append saved commands to the ring again
*
* @ring: ring to append commands to
* @size: number of dwords we want to write
* @data: saved commands
*
* Allocates space on the ring and restore the previously saved commands.
*/
int amdgpu_ring_restore(struct amdgpu_ring *ring,
unsigned size, uint32_t *data)
{
int i, r;
if (!size || !data)
return 0;
/* restore the saved ring content */
r = amdgpu_ring_alloc(ring, size);
if (r)
return r;
for (i = 0; i < size; ++i) {
amdgpu_ring_write(ring, data[i]);
}
amdgpu_ring_commit(ring);
kfree(data);
return 0;
}
void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
{
signed long r;
unsigned long flags;
uint32_t seq;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
BUG_ON(!ring->funcs->emit_wreg);
spin_lock_irqsave(&kiq->ring_lock, flags);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_wreg(ring, reg, v);
amdgpu_fence_emit_polling(ring, &seq);
amdgpu_ring_commit(ring);
spin_unlock_irqrestore(&kiq->ring_lock, flags);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
if (r < 1)
DRM_ERROR("wait for kiq fence error: %ld\n", r);
}
void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
uint32_t reg0, uint32_t reg1,
uint32_t ref, uint32_t mask)
{
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
signed long r, cnt = 0;
unsigned long flags;
uint32_t seq;
spin_lock_irqsave(&kiq->ring_lock, flags);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1,
ref, mask);
amdgpu_fence_emit_polling(ring, &seq);
amdgpu_ring_commit(ring);
spin_unlock_irqrestore(&kiq->ring_lock, flags);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
/* don't wait anymore for IRQ context */
if (r < 1 && in_interrupt())
goto failed_kiq;
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
}
if (cnt > MAX_KIQ_REG_TRY)
goto failed_kiq;
return;
failed_kiq:
pr_err("failed to write reg %x wait reg %x\n", reg0, reg1);
}
void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
{
signed long r;
struct dma_fence *f;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
BUG_ON(!ring->funcs->emit_wreg);
mutex_lock(&adev->virt.lock);
amdgpu_ring_alloc(ring, 32);
amdgpu_ring_emit_hdp_flush(ring);
amdgpu_ring_emit_wreg(ring, reg, v);
amdgpu_ring_emit_hdp_invalidate(ring);
amdgpu_fence_emit(ring, &f);
amdgpu_ring_commit(ring);
mutex_unlock(&adev->virt.lock);
r = dma_fence_wait(f, false);
if (r)
DRM_ERROR("wait for kiq fence error: %ld.\n", r);
dma_fence_put(f);
}
/**
* uvd_v6_0_enc_ring_test_ring - test if UVD ENC ring is working
*
* @ring: the engine to test on
*
*/
static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
uint32_t rptr = amdgpu_ring_get_rptr(ring);
unsigned i;
int r;
r = amdgpu_ring_alloc(ring, 16);
if (r)
return r;
amdgpu_ring_write(ring, HEVC_ENC_CMD_END);
amdgpu_ring_commit(ring);
for (i = 0; i < adev->usec_timeout; i++) {
if (amdgpu_ring_get_rptr(ring) != rptr)
break;
DRM_UDELAY(1);
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
/**
* uvd_v6_0_hw_init - start and test UVD block
*
* @adev: amdgpu_device pointer
*
* Initialize the hardware, boot up the VCPU and do some testing
*/
static int uvd_v6_0_hw_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_ring *ring = &adev->uvd.inst->ring;
uint32_t tmp;
int i, r;
amdgpu_asic_set_uvd_clocks(adev, 10000, 10000);
uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
uvd_v6_0_enable_mgcg(adev, true);
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
r = amdgpu_ring_alloc(ring, 10);
if (r) {
DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
goto done;
}
tmp = PACKET0(mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
tmp = PACKET0(mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0);
amdgpu_ring_write(ring, tmp);
amdgpu_ring_write(ring, 0xFFFFF);
/* Clear timeout status bits */
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_TIMEOUT_STATUS, 0));
amdgpu_ring_write(ring, 0x8);
amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_CNTL, 0));
amdgpu_ring_write(ring, 3);
amdgpu_ring_commit(ring);
if (uvd_v6_0_enc_support(adev)) {
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst->ring_enc[i];
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
}
done:
if (!r) {
if (uvd_v6_0_enc_support(adev))
DRM_INFO("UVD and UVD ENC initialized successfully.\n");
else
DRM_INFO("UVD initialized successfully.\n");
}
return r;
}
/**
* amdgpu_ib_schedule - schedule an IB (Indirect Buffer) on the ring
*
* @adev: amdgpu_device pointer
* @num_ibs: number of IBs to schedule
* @ibs: IB objects to schedule
* @f: fence created during this submission
*
* Schedule an IB on the associated ring (all asics).
* Returns 0 on success, error on failure.
*
* On SI, there are two parallel engines fed from the primary ring,
* the CE (Constant Engine) and the DE (Drawing Engine). Since
* resource descriptors have moved to memory, the CE allows you to
* prime the caches while the DE is updating register state so that
* the resource descriptors will be already in cache when the draw is
* processed. To accomplish this, the userspace driver submits two
* IBs, one for the CE and one for the DE. If there is a CE IB (called
* a CONST_IB), it will be put on the ring prior to the DE IB. Prior
* to SI there was just a DE IB.
*/
int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
struct amdgpu_ib *ibs, struct amdgpu_job *job,
struct dma_fence **f)
{
struct amdgpu_device *adev = ring->adev;
struct amdgpu_ib *ib = &ibs[0];
struct dma_fence *tmp = NULL;
bool skip_preamble, need_ctx_switch;
unsigned patch_offset = ~0;
struct amdgpu_vm *vm;
uint64_t fence_ctx;
uint32_t status = 0, alloc_size;
unsigned fence_flags = 0;
unsigned i;
int r = 0;
bool need_pipe_sync = false;
if (num_ibs == 0)
return -EINVAL;
/* ring tests don't use a job */
if (job) {
vm = job->vm;
fence_ctx = job->base.s_fence->scheduled.context;
} else {
vm = NULL;
fence_ctx = 0;
}
if (!ring->sched.ready) {
dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name);
return -EINVAL;
}
if (vm && !job->vmid) {
dev_err(adev->dev, "VM IB without ID\n");
return -EINVAL;
}
alloc_size = ring->funcs->emit_frame_size + num_ibs *
ring->funcs->emit_ib_size;
r = amdgpu_ring_alloc(ring, alloc_size);
if (r) {
dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
return r;
}
need_ctx_switch = ring->current_ctx != fence_ctx;
if (ring->funcs->emit_pipeline_sync && job &&
((tmp = amdgpu_sync_get_fence(&job->sched_sync, NULL)) ||
(amdgpu_sriov_vf(adev) && need_ctx_switch) ||
amdgpu_vm_need_pipeline_sync(ring, job))) {
need_pipe_sync = true;
if (tmp)
trace_amdgpu_ib_pipe_sync(job, tmp);
dma_fence_put(tmp);
}
if (ring->funcs->insert_start)
ring->funcs->insert_start(ring);
if (job) {
r = amdgpu_vm_flush(ring, job, need_pipe_sync);
if (r) {
amdgpu_ring_undo(ring);
return r;
}
}
if (job && ring->funcs->init_cond_exec)
patch_offset = amdgpu_ring_init_cond_exec(ring);
#ifdef CONFIG_X86_64
if (!(adev->flags & AMD_IS_APU))
#endif
{
if (ring->funcs->emit_hdp_flush)
amdgpu_ring_emit_hdp_flush(ring);
else
amdgpu_asic_flush_hdp(adev, ring);
}
if (need_ctx_switch)
status |= AMDGPU_HAVE_CTX_SWITCH;
skip_preamble = ring->current_ctx == fence_ctx;
if (job && ring->funcs->emit_cntxcntl) {
status |= job->preamble_status;
amdgpu_ring_emit_cntxcntl(ring, status);
}
for (i = 0; i < num_ibs; ++i) {
ib = &ibs[i];
/* drop preamble IBs if we don't have a context switch */
if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) &&
skip_preamble &&
!(status & AMDGPU_PREAMBLE_IB_PRESENT_FIRST) &&
!amdgpu_sriov_vf(adev)) /* for SRIOV preemption, Preamble CE ib must be inserted anyway */
continue;
amdgpu_ring_emit_ib(ring, job, ib, status);
status &= ~AMDGPU_HAVE_CTX_SWITCH;
}
if (ring->funcs->emit_tmz)
amdgpu_ring_emit_tmz(ring, false);
#ifdef CONFIG_X86_64
if (!(adev->flags & AMD_IS_APU))
#endif
amdgpu_asic_invalidate_hdp(adev, ring);
if (ib->flags & AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE)
fence_flags |= AMDGPU_FENCE_FLAG_TC_WB_ONLY;
/* wrap the last IB with fence */
if (job && job->uf_addr) {
amdgpu_ring_emit_fence(ring, job->uf_addr, job->uf_sequence,
fence_flags | AMDGPU_FENCE_FLAG_64BIT);
}
r = amdgpu_fence_emit(ring, f, fence_flags);
if (r) {
dev_err(adev->dev, "failed to emit fence (%d)\n", r);
if (job && job->vmid)
amdgpu_vmid_reset(adev, ring->funcs->vmhub, job->vmid);
amdgpu_ring_undo(ring);
return r;
}
if (ring->funcs->insert_end)
ring->funcs->insert_end(ring);
if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
amdgpu_ring_patch_cond_exec(ring, patch_offset);
ring->current_ctx = fence_ctx;
if (vm && ring->funcs->emit_switch_buffer)
amdgpu_ring_emit_switch_buffer(ring);
amdgpu_ring_commit(ring);
return 0;
}
/**
* amdgpu_ib_schedule - schedule an IB (Indirect Buffer) on the ring
*
* @adev: amdgpu_device pointer
* @num_ibs: number of IBs to schedule
* @ibs: IB objects to schedule
* @f: fence created during this submission
*
* Schedule an IB on the associated ring (all asics).
* Returns 0 on success, error on failure.
*
* On SI, there are two parallel engines fed from the primary ring,
* the CE (Constant Engine) and the DE (Drawing Engine). Since
* resource descriptors have moved to memory, the CE allows you to
* prime the caches while the DE is updating register state so that
* the resource descriptors will be already in cache when the draw is
* processed. To accomplish this, the userspace driver submits two
* IBs, one for the CE and one for the DE. If there is a CE IB (called
* a CONST_IB), it will be put on the ring prior to the DE IB. Prior
* to SI there was just a DE IB.
*/
int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
struct amdgpu_ib *ibs, struct fence *last_vm_update,
struct amdgpu_job *job, struct fence **f)
{
struct amdgpu_device *adev = ring->adev;
struct amdgpu_ib *ib = &ibs[0];
bool skip_preamble, need_ctx_switch;
unsigned patch_offset = ~0;
struct amdgpu_vm *vm;
struct fence *hwf;
uint64_t ctx;
unsigned i;
int r = 0;
if (num_ibs == 0)
return -EINVAL;
/* ring tests don't use a job */
if (job) {
vm = job->vm;
ctx = job->ctx;
} else {
vm = NULL;
ctx = 0;
}
if (!ring->ready) {
dev_err(adev->dev, "couldn't schedule ib\n");
return -EINVAL;
}
if (vm && !job->vm_id) {
dev_err(adev->dev, "VM IB without ID\n");
return -EINVAL;
}
r = amdgpu_ring_alloc(ring, 256 * num_ibs);
if (r) {
dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
return r;
}
if (ring->type == AMDGPU_RING_TYPE_SDMA && ring->funcs->init_cond_exec)
patch_offset = amdgpu_ring_init_cond_exec(ring);
if (vm) {
r = amdgpu_vm_flush(ring, job->vm_id, job->vm_pd_addr,
job->gds_base, job->gds_size,
job->gws_base, job->gws_size,
job->oa_base, job->oa_size);
if (r) {
amdgpu_ring_undo(ring);
return r;
}
}
if (ring->funcs->emit_hdp_flush)
amdgpu_ring_emit_hdp_flush(ring);
/* always set cond_exec_polling to CONTINUE */
*ring->cond_exe_cpu_addr = 1;
skip_preamble = ring->current_ctx == ctx;
need_ctx_switch = ring->current_ctx != ctx;
for (i = 0; i < num_ibs; ++i) {
ib = &ibs[i];
/* drop preamble IBs if we don't have a context switch */
if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && skip_preamble)
continue;
amdgpu_ring_emit_ib(ring, ib, job ? job->vm_id : 0,
need_ctx_switch);
need_ctx_switch = false;
}
if (ring->funcs->emit_hdp_invalidate)
amdgpu_ring_emit_hdp_invalidate(ring);
r = amdgpu_fence_emit(ring, &hwf);
if (r) {
dev_err(adev->dev, "failed to emit fence (%d)\n", r);
if (job && job->vm_id)
amdgpu_vm_reset_id(adev, job->vm_id);
amdgpu_ring_undo(ring);
return r;
}
/* wrap the last IB with fence */
if (job && job->uf_bo) {
uint64_t addr = amdgpu_bo_gpu_offset(job->uf_bo);
addr += job->uf_offset;
amdgpu_ring_emit_fence(ring, addr, job->uf_sequence,
AMDGPU_FENCE_FLAG_64BIT);
}
if (f)
*f = fence_get(hwf);
if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
amdgpu_ring_patch_cond_exec(ring, patch_offset);
ring->current_ctx = ctx;
amdgpu_ring_commit(ring);
return 0;
}
