mirrors/linux-stable
1
0
Fork 0
mirror of https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git synced 2024-09-27 21:03:32 +00:00
Code Issues Releases Wiki Activity

drm/amdgpu: move vmhub out of amdgpu_ring_funcs (v4)

Browse source 
It looks better to place this field in ring
structure. Also drop the repeated ring funcs definitions
if there's no difference except for vmhub field.

v2: rename the field to vm_hub like others (Le)
v3: apply the changes to new ip blocks (Hawking)
v4: fix vcn sw ring (Alex)

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Le Ma 2022-05-19 19:49:59 +08:00 committed by Alex Deucher
parent b805d8d785
commit 0530553ba8
31 changed files with 93 additions and 220 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
View file

@ -305,6 +305,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
ring->ring_obj = NULL;
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.kiq;
ring->vm_hub = AMDGPU_GFXHUB_0;
r = amdgpu_gfx_kiq_acquire(adev, ring);
if (r)

4
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
View file

@ -554,7 +554,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
for (i = 0; i < adev->num_rings; ++i) {
ring = adev->rings[i];
vmhub = ring->funcs->vmhub;
vmhub = ring->vm_hub;
if (ring == &adev->mes.ring)
continue;
@ -570,7 +570,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng);
dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n",
ring->name, ring->vm_inv_eng, ring->funcs->vmhub);
ring->name, ring->vm_inv_eng, ring->vm_hub);
}
return 0;

2
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
View file

@ -267,7 +267,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
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_vmid_reset(adev, ring->vm_hub, job->vmid);
amdgpu_ring_undo(ring);
return r;
}

8
drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
View file

@ -202,7 +202,7 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
struct dma_fence **fence)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
struct dma_fence **fences;
unsigned i;
@ -277,7 +277,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
struct dma_fence **fence)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
uint64_t fence_context = adev->fence_context + ring->idx;
bool needs_flush = vm->use_cpu_for_update;
@ -338,7 +338,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
struct dma_fence **fence)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
uint64_t fence_context = adev->fence_context + ring->idx;
uint64_t updates = amdgpu_vm_tlb_seq(vm);
@ -398,7 +398,7 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
struct amdgpu_job *job, struct dma_fence **fence)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
struct amdgpu_vmid *idle = NULL;
struct amdgpu_vmid *id = NULL;

2
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
View file

@ -165,7 +165,6 @@ struct amdgpu_ring_funcs {
bool support_64bit_ptrs;
bool no_user_fence;
bool secure_submission_supported;
unsigned vmhub;
unsigned extra_dw;
/* ring read/write ptr handling */
@ -275,6 +274,7 @@ struct amdgpu_ring {
unsigned cond_exe_offs;
u64 cond_exe_gpu_addr;
volatile u32 *cond_exe_cpu_addr;
unsigned vm_hub;
unsigned vm_inv_eng;
struct dma_fence *vmid_wait;
bool has_compute_vm_bug;

4
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
View file

@ -233,7 +233,7 @@ TRACE_EVENT(amdgpu_vm_grab_id,
__entry->pasid = vm->pasid;
__assign_str(ring, ring->name);
__entry->vmid = job->vmid;
__entry->vm_hub = ring->funcs->vmhub,
__entry->vm_hub = ring->vm_hub,
__entry->pd_addr = job->vm_pd_addr;
__entry->needs_flush = job->vm_needs_flush;
),
@ -427,7 +427,7 @@ TRACE_EVENT(amdgpu_vm_flush,
TP_fast_assign(
__assign_str(ring, ring->name);
__entry->vmid = vmid;
__entry->vm_hub = ring->funcs->vmhub;
__entry->vm_hub = ring->vm_hub;
__entry->pd_addr = pd_addr;
),
TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",

4
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
View file

@ -483,7 +483,7 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
struct amdgpu_job *job)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
if (job->vmid == 0)
@ -517,7 +517,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
bool need_pipe_sync)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmhub = ring->funcs->vmhub;
unsigned vmhub = ring->vm_hub;
struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
struct amdgpu_vmid *id = &id_mgr->ids[job->vmid];
bool spm_update_needed = job->spm_update_needed;

5
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
View file

@ -4461,6 +4461,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@ -4489,6 +4490,7 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX10_MEC_HPD_SIZE);
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@ -9249,7 +9251,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v10_0_ring_get_rptr_gfx,
.get_wptr = gfx_v10_0_ring_get_wptr_gfx,
.set_wptr = gfx_v10_0_ring_set_wptr_gfx,
@ -9304,7 +9305,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v10_0_ring_get_rptr_compute,
.get_wptr = gfx_v10_0_ring_get_wptr_compute,
.set_wptr = gfx_v10_0_ring_set_wptr_compute,
@ -9340,7 +9340,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v10_0_ring_get_rptr_compute,
.get_wptr = gfx_v10_0_ring_get_wptr_compute,
.set_wptr = gfx_v10_0_ring_set_wptr_compute,

5
drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
View file

@ -866,6 +866,7 @@ static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@ -896,6 +897,7 @@ static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX11_MEC_HPD_SIZE);
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@ -6204,7 +6206,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v11_0_ring_get_rptr_gfx,
.get_wptr = gfx_v11_0_ring_get_wptr_gfx,
.set_wptr = gfx_v11_0_ring_set_wptr_gfx,
@ -6252,7 +6253,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v11_0_ring_get_rptr_compute,
.get_wptr = gfx_v11_0_ring_get_wptr_compute,
.set_wptr = gfx_v11_0_ring_set_wptr_compute,
@ -6288,7 +6288,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v11_0_ring_get_rptr_compute,
.get_wptr = gfx_v11_0_ring_get_wptr_compute,
.set_wptr = gfx_v11_0_ring_set_wptr_compute,

7
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
View file

@ -2005,6 +2005,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX9_MEC_HPD_SIZE);
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@ -2104,6 +2105,7 @@ static int gfx_v9_0_sw_init(void *handle)
/* disable scheduler on the real ring */
ring->no_scheduler = true;
ring->vm_hub = AMDGPU_GFXHUB_0;
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -2121,6 +2123,7 @@ static int gfx_v9_0_sw_init(void *handle)
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
ring->is_sw_ring = true;
hw_prio = amdgpu_sw_ring_priority(i);
ring->vm_hub = AMDGPU_GFXHUB_0;
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio,
NULL);
@ -6790,7 +6793,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v9_0_ring_get_rptr_gfx,
.get_wptr = gfx_v9_0_ring_get_wptr_gfx,
.set_wptr = gfx_v9_0_ring_set_wptr_gfx,
@ -6844,7 +6846,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = amdgpu_sw_ring_get_rptr_gfx,
.get_wptr = amdgpu_sw_ring_get_wptr_gfx,
.set_wptr = amdgpu_sw_ring_set_wptr_gfx,
@ -6898,7 +6899,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v9_0_ring_get_rptr_compute,
.get_wptr = gfx_v9_0_ring_get_wptr_compute,
.set_wptr = gfx_v9_0_ring_set_wptr_compute,
@ -6937,7 +6937,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
.align_mask = 0xff,
.nop = PACKET3(PACKET3_NOP, 0x3FFF),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = gfx_v9_0_ring_get_rptr_compute,
.get_wptr = gfx_v9_0_ring_get_wptr_compute,
.set_wptr = gfx_v9_0_ring_set_wptr_compute,

6
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
View file

@ -479,8 +479,8 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
unsigned eng = ring->vm_inv_eng;
@ -534,7 +534,7 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
if (ring->vm_hub == AMDGPU_GFXHUB_0)
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;

6
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
View file

@ -378,8 +378,8 @@ static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
unsigned eng = ring->vm_inv_eng;
@ -433,7 +433,7 @@ static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
if (ring->vm_hub == AMDGPU_GFXHUB_0)
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;

8
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
View file

@ -1007,9 +1007,9 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
struct amdgpu_device *adev = ring->adev;
struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &adev->vmhub[ring->vm_hub];
uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0);
unsigned eng = ring->vm_inv_eng;
@ -1060,10 +1060,10 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
uint32_t reg;
/* Do nothing because there's no lut register for mmhub1. */
if (ring->funcs->vmhub == AMDGPU_MMHUB_1)
if (ring->vm_hub == AMDGPU_MMHUB_1)
return;
if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
if (ring->vm_hub == AMDGPU_GFXHUB_0)
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;

4
drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
View file

@ -376,7 +376,7 @@ static void jpeg_v1_0_decode_ring_emit_reg_wait(struct amdgpu_ring *ring,
static void jpeg_v1_0_decode_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -485,6 +485,7 @@ int jpeg_v1_0_sw_init(void *handle)
return r;
ring = &adev->jpeg.inst->ring_dec;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -548,7 +549,6 @@ static const struct amdgpu_ring_funcs jpeg_v1_0_decode_ring_vm_funcs = {
.nop = PACKET0(0x81ff, 0),
.support_64bit_ptrs = false,
.no_user_fence = true,
.vmhub = AMDGPU_MMHUB_0,
.extra_dw = 64,
.get_rptr = jpeg_v1_0_decode_ring_get_rptr,
.get_wptr = jpeg_v1_0_decode_ring_get_wptr,

4
drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
View file

@ -86,6 +86,7 @@ static int jpeg_v2_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -613,7 +614,7 @@ void jpeg_v2_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
void jpeg_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -762,7 +763,6 @@ static const struct amd_ip_funcs jpeg_v2_0_ip_funcs = {
static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_JPEG,
.align_mask = 0xf,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = jpeg_v2_0_dec_ring_get_rptr,
.get_wptr = jpeg_v2_0_dec_ring_get_wptr,
.set_wptr = jpeg_v2_0_dec_ring_set_wptr,

6
drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
View file

@ -127,6 +127,10 @@ static int jpeg_v2_5_sw_init(void *handle)
ring = &adev->jpeg.inst[i].ring_dec;
ring->use_doorbell = true;
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
ring->vm_hub = AMDGPU_MMHUB_1;
else
ring->vm_hub = AMDGPU_MMHUB_0;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + 8 * i;
sprintf(ring->name, "jpeg_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst[i].irq,
@ -645,7 +649,6 @@ static const struct amd_ip_funcs jpeg_v2_6_ip_funcs = {
static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_JPEG,
.align_mask = 0xf,
.vmhub = AMDGPU_MMHUB_1,
.get_rptr = jpeg_v2_5_dec_ring_get_rptr,
.get_wptr = jpeg_v2_5_dec_ring_get_wptr,
.set_wptr = jpeg_v2_5_dec_ring_set_wptr,
@ -675,7 +678,6 @@ static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = {
static const struct amdgpu_ring_funcs jpeg_v2_6_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_JPEG,
.align_mask = 0xf,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = jpeg_v2_5_dec_ring_get_rptr,
.get_wptr = jpeg_v2_5_dec_ring_get_wptr,
.set_wptr = jpeg_v2_5_dec_ring_set_wptr,

2
drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
View file

@ -100,6 +100,7 @@ static int jpeg_v3_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -559,7 +560,6 @@ static const struct amd_ip_funcs jpeg_v3_0_ip_funcs = {
static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_JPEG,
.align_mask = 0xf,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = jpeg_v3_0_dec_ring_get_rptr,
.get_wptr = jpeg_v3_0_dec_ring_get_wptr,
.set_wptr = jpeg_v3_0_dec_ring_set_wptr,

2
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
View file

@ -108,6 +108,7 @@ static int jpeg_v4_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = amdgpu_sriov_vf(adev) ? (((adev->doorbell_index.vcn.vcn_ring0_1) << 1) + 4) : ((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1);
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
@ -715,7 +716,6 @@ static const struct amd_ip_funcs jpeg_v4_0_ip_funcs = {
static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_JPEG,
.align_mask = 0xf,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = jpeg_v4_0_dec_ring_get_rptr,
.get_wptr = jpeg_v4_0_dec_ring_get_wptr,
.set_wptr = jpeg_v4_0_dec_ring_set_wptr,

99
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
View file

@ -1823,6 +1823,15 @@ static int sdma_v4_0_sw_init(void *handle)
/* doorbell size is 2 dwords, get DWORD offset */
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
/*
* On Arcturus, SDMA instance 5~7 has a different vmhub
* type(AMDGPU_MMHUB_1).
*/
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
ring->vm_hub = AMDGPU_MMHUB_1;
else
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
AMDGPU_SDMA_IRQ_INSTANCE0 + i,
@ -1841,6 +1850,11 @@ static int sdma_v4_0_sw_init(void *handle)
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
ring->doorbell_index += 0x400;
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
ring->vm_hub = AMDGPU_MMHUB_1;
else
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
&adev->sdma.trap_irq,
@ -2294,44 +2308,6 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = {
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = sdma_v4_0_ring_get_rptr,
.get_wptr = sdma_v4_0_ring_get_wptr,
.set_wptr = sdma_v4_0_ring_set_wptr,
.emit_frame_size =
6 + /* sdma_v4_0_ring_emit_hdp_flush */
3 + /* hdp invalidate */
6 + /* sdma_v4_0_ring_emit_pipeline_sync */
/* sdma_v4_0_ring_emit_vm_flush */
SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6 +
10 + 10 + 10, /* sdma_v4_0_ring_emit_fence x3 for user fence, vm fence */
.emit_ib_size = 7 + 6, /* sdma_v4_0_ring_emit_ib */
.emit_ib = sdma_v4_0_ring_emit_ib,
.emit_fence = sdma_v4_0_ring_emit_fence,
.emit_pipeline_sync = sdma_v4_0_ring_emit_pipeline_sync,
.emit_vm_flush = sdma_v4_0_ring_emit_vm_flush,
.emit_hdp_flush = sdma_v4_0_ring_emit_hdp_flush,
.test_ring = sdma_v4_0_ring_test_ring,
.test_ib = sdma_v4_0_ring_test_ib,
.insert_nop = sdma_v4_0_ring_insert_nop,
.pad_ib = sdma_v4_0_ring_pad_ib,
.emit_wreg = sdma_v4_0_ring_emit_wreg,
.emit_reg_wait = sdma_v4_0_ring_emit_reg_wait,
.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
/*
* On Arcturus, SDMA instance 5~7 has a different vmhub type(AMDGPU_MMHUB_1).
* So create a individual constant ring_funcs for those instances.
*/
static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs_2nd_mmhub = {
.type = AMDGPU_RING_TYPE_SDMA,
.align_mask = 0xf,
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_1,
.get_rptr = sdma_v4_0_ring_get_rptr,
.get_wptr = sdma_v4_0_ring_get_wptr,
.set_wptr = sdma_v4_0_ring_set_wptr,
@ -2364,40 +2340,6 @@ static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs = {
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = sdma_v4_0_ring_get_rptr,
.get_wptr = sdma_v4_0_page_ring_get_wptr,
.set_wptr = sdma_v4_0_page_ring_set_wptr,
.emit_frame_size =
6 + /* sdma_v4_0_ring_emit_hdp_flush */
3 + /* hdp invalidate */
6 + /* sdma_v4_0_ring_emit_pipeline_sync */
/* sdma_v4_0_ring_emit_vm_flush */
SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6 +
10 + 10 + 10, /* sdma_v4_0_ring_emit_fence x3 for user fence, vm fence */
.emit_ib_size = 7 + 6, /* sdma_v4_0_ring_emit_ib */
.emit_ib = sdma_v4_0_ring_emit_ib,
.emit_fence = sdma_v4_0_ring_emit_fence,
.emit_pipeline_sync = sdma_v4_0_ring_emit_pipeline_sync,
.emit_vm_flush = sdma_v4_0_ring_emit_vm_flush,
.emit_hdp_flush = sdma_v4_0_ring_emit_hdp_flush,
.test_ring = sdma_v4_0_ring_test_ring,
.test_ib = sdma_v4_0_ring_test_ib,
.insert_nop = sdma_v4_0_ring_insert_nop,
.pad_ib = sdma_v4_0_ring_pad_ib,
.emit_wreg = sdma_v4_0_ring_emit_wreg,
.emit_reg_wait = sdma_v4_0_ring_emit_reg_wait,
.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
static const struct amdgpu_ring_funcs sdma_v4_0_page_ring_funcs_2nd_mmhub = {
.type = AMDGPU_RING_TYPE_SDMA,
.align_mask = 0xf,
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_1,
.get_rptr = sdma_v4_0_ring_get_rptr,
.get_wptr = sdma_v4_0_page_ring_get_wptr,
.set_wptr = sdma_v4_0_page_ring_set_wptr,
@ -2429,19 +2371,10 @@ static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev)
int i;
for (i = 0; i < adev->sdma.num_instances; i++) {
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
adev->sdma.instance[i].ring.funcs =
&sdma_v4_0_ring_funcs_2nd_mmhub;
else
adev->sdma.instance[i].ring.funcs =
&sdma_v4_0_ring_funcs;
adev->sdma.instance[i].ring.funcs = &sdma_v4_0_ring_funcs;
adev->sdma.instance[i].ring.me = i;
if (adev->sdma.has_page_queue) {
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
adev->sdma.instance[i].page.funcs =
&sdma_v4_0_page_ring_funcs_2nd_mmhub;
else
adev->sdma.instance[i].page.funcs =
adev->sdma.instance[i].page.funcs =
&sdma_v4_0_page_ring_funcs;
adev->sdma.instance[i].page.me = i;
}

4
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
View file

@ -1309,6 +1309,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
/* doorbell size is 2 dwords, get DWORD offset */
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@ -1327,6 +1328,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
*/
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
ring->doorbell_index += 0x400;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
@ -1741,7 +1743,6 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_ring_funcs = {
.align_mask = 0xf,
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = sdma_v4_4_2_ring_get_rptr,
.get_wptr = sdma_v4_4_2_ring_get_wptr,
.set_wptr = sdma_v4_4_2_ring_set_wptr,
@ -1773,7 +1774,6 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = {
.align_mask = 0xf,
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = sdma_v4_4_2_ring_get_rptr,
.get_wptr = sdma_v4_4_2_page_ring_get_wptr,
.set_wptr = sdma_v4_4_2_page_ring_set_wptr,

2
drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
View file

@ -1389,6 +1389,7 @@ static int sdma_v5_0_sw_init(void *handle)
(adev->doorbell_index.sdma_engine[0] << 1) //get DWORD offset
: (adev->doorbell_index.sdma_engine[1] << 1); // get DWORD offset
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
(i == 0) ? AMDGPU_SDMA_IRQ_INSTANCE0 :
@ -1765,7 +1766,6 @@ static const struct amdgpu_ring_funcs sdma_v5_0_ring_funcs = {
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = sdma_v5_0_ring_get_rptr,
.get_wptr = sdma_v5_0_ring_get_wptr,
.set_wptr = sdma_v5_0_ring_set_wptr,

2
drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
View file

@ -1253,6 +1253,7 @@ static int sdma_v5_2_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); //get DWORD offset
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
AMDGPU_SDMA_IRQ_INSTANCE0 + i,
@ -1653,7 +1654,6 @@ static const struct amdgpu_ring_funcs sdma_v5_2_ring_funcs = {
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = sdma_v5_2_ring_get_rptr,
.get_wptr = sdma_v5_2_ring_get_wptr,
.set_wptr = sdma_v5_2_ring_set_wptr,

4
drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
View file

@ -1181,7 +1181,7 @@ static void sdma_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
static void sdma_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
/* Update the PD address for this VMID. */
@ -1301,6 +1301,7 @@ static int sdma_v6_0_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset
ring->vm_hub = AMDGPU_GFXHUB_0;
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
&adev->sdma.trap_irq,
@ -1557,7 +1558,6 @@ static const struct amdgpu_ring_funcs sdma_v6_0_ring_funcs = {
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_GFXHUB_0,
.get_rptr = sdma_v6_0_ring_get_rptr,
.get_wptr = sdma_v6_0_ring_get_wptr,
.set_wptr = sdma_v6_0_ring_set_wptr,

8
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
View file

@ -444,6 +444,7 @@ static int uvd_v7_0_sw_init(void *handle)
continue;
if (!amdgpu_sriov_vf(adev)) {
ring = &adev->uvd.inst[j].ring;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "uvd_%d", ring->me);
r = amdgpu_ring_init(adev, ring, 512,
&adev->uvd.inst[j].irq, 0,
@ -454,6 +455,7 @@ static int uvd_v7_0_sw_init(void *handle)
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst[j].ring_enc[i];
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "uvd_enc_%d.%d", ring->me, i);
if (amdgpu_sriov_vf(adev)) {
ring->use_doorbell = true;
@ -1397,7 +1399,7 @@ static void uvd_v7_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1440,7 +1442,7 @@ static void uvd_v7_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring,
static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned int vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1802,7 +1804,6 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = {
.align_mask = 0xf,
.support_64bit_ptrs = false,
.no_user_fence = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = uvd_v7_0_ring_get_rptr,
.get_wptr = uvd_v7_0_ring_get_wptr,
.set_wptr = uvd_v7_0_ring_set_wptr,
@ -1835,7 +1836,6 @@ static const struct amdgpu_ring_funcs uvd_v7_0_enc_ring_vm_funcs = {
.nop = HEVC_ENC_CMD_NO_OP,
.support_64bit_ptrs = false,
.no_user_fence = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = uvd_v7_0_enc_ring_get_rptr,
.get_wptr = uvd_v7_0_enc_ring_get_wptr,
.set_wptr = uvd_v7_0_enc_ring_set_wptr,

4
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
View file

@ -466,6 +466,7 @@ static int vce_v4_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(i);
ring = &adev->vce.ring[i];
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vce%d", i);
if (amdgpu_sriov_vf(adev)) {
/* DOORBELL only works under SRIOV */
@ -1021,7 +1022,7 @@ static void vce_v4_0_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring,
unsigned int vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1103,7 +1104,6 @@ static const struct amdgpu_ring_funcs vce_v4_0_ring_vm_funcs = {
.nop = VCE_CMD_NO_OP,
.support_64bit_ptrs = false,
.no_user_fence = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vce_v4_0_ring_get_rptr,
.get_wptr = vce_v4_0_ring_get_wptr,
.set_wptr = vce_v4_0_ring_set_wptr,

2
drivers/gpu/drm/amd/amdgpu/vcn_sw_ring.c
View file

@ -65,7 +65,7 @@ void vcn_dec_sw_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
void vcn_dec_sw_ring_emit_vm_flush(struct amdgpu_ring *ring,
uint32_t vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);

8
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
View file

@ -120,6 +120,7 @@ static int vcn_v1_0_sw_init(void *handle)
return r;
ring = &adev->vcn.inst->ring_dec;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -141,6 +142,7 @@ static int vcn_v1_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
ring = &adev->vcn.inst->ring_enc[i];
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_enc%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
hw_prio, NULL);
@ -1548,7 +1550,7 @@ static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring,
static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1693,7 +1695,7 @@ static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring,
static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned int vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1977,7 +1979,6 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = {
.support_64bit_ptrs = false,
.no_user_fence = true,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v1_0_dec_ring_get_rptr,
.get_wptr = vcn_v1_0_dec_ring_get_wptr,
.set_wptr = vcn_v1_0_dec_ring_set_wptr,
@ -2012,7 +2013,6 @@ static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = {
.nop = VCN_ENC_CMD_NO_OP,
.support_64bit_ptrs = false,
.no_user_fence = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v1_0_enc_ring_get_rptr,
.get_wptr = vcn_v1_0_enc_ring_get_wptr,
.set_wptr = vcn_v1_0_enc_ring_set_wptr,

8
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
View file

@ -129,6 +129,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1 << 1;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
@ -159,6 +160,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring = &adev->vcn.inst->ring_enc[i];
ring->use_doorbell = true;
ring->vm_hub = AMDGPU_MMHUB_0;
if (!amdgpu_sriov_vf(adev))
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
else
@ -1511,7 +1513,7 @@ void vcn_v2_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
void vcn_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
uint32_t data0, data1, mask;
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -1671,7 +1673,7 @@ void vcn_v2_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
void vcn_v2_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned int vmid, uint64_t pd_addr)
{
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
@ -2014,7 +2016,6 @@ static const struct amdgpu_ring_funcs vcn_v2_0_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_DEC,
.align_mask = 0xf,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v2_0_dec_ring_get_rptr,
.get_wptr = vcn_v2_0_dec_ring_get_wptr,
.set_wptr = vcn_v2_0_dec_ring_set_wptr,
@ -2045,7 +2046,6 @@ static const struct amdgpu_ring_funcs vcn_v2_0_enc_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_ENC,
.align_mask = 0x3f,
.nop = VCN_ENC_CMD_NO_OP,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v2_0_enc_ring_get_rptr,
.get_wptr = vcn_v2_0_enc_ring_get_wptr,
.set_wptr = vcn_v2_0_enc_ring_set_wptr,

84
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
View file

@ -186,6 +186,12 @@ static int vcn_v2_5_sw_init(void *handle)
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
(amdgpu_sriov_vf(adev) ? 2*j : 8*j);
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
ring->vm_hub = AMDGPU_MMHUB_1;
else
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_dec_%d", j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
@ -201,6 +207,11 @@ static int vcn_v2_5_sw_init(void *handle)
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) +
(amdgpu_sriov_vf(adev) ? (1 + i + 2*j) : (2 + i + 8*j));
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
ring->vm_hub = AMDGPU_MMHUB_1;
else
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_enc_%d.%d", j, i);
r = amdgpu_ring_init(adev, ring, 512,
&adev->vcn.inst[j].irq, 0,
@ -1562,38 +1573,6 @@ static const struct amdgpu_ring_funcs vcn_v2_5_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_DEC,
.align_mask = 0xf,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_1,
.get_rptr = vcn_v2_5_dec_ring_get_rptr,
.get_wptr = vcn_v2_5_dec_ring_get_wptr,
.set_wptr = vcn_v2_5_dec_ring_set_wptr,
.emit_frame_size =
SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 +
SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 +
8 + /* vcn_v2_0_dec_ring_emit_vm_flush */
14 + 14 + /* vcn_v2_0_dec_ring_emit_fence x2 vm fence */
6,
.emit_ib_size = 8, /* vcn_v2_0_dec_ring_emit_ib */
.emit_ib = vcn_v2_0_dec_ring_emit_ib,
.emit_fence = vcn_v2_0_dec_ring_emit_fence,
.emit_vm_flush = vcn_v2_0_dec_ring_emit_vm_flush,
.test_ring = vcn_v2_0_dec_ring_test_ring,
.test_ib = amdgpu_vcn_dec_ring_test_ib,
.insert_nop = vcn_v2_0_dec_ring_insert_nop,
.insert_start = vcn_v2_0_dec_ring_insert_start,
.insert_end = vcn_v2_0_dec_ring_insert_end,
.pad_ib = amdgpu_ring_generic_pad_ib,
.begin_use = amdgpu_vcn_ring_begin_use,
.end_use = amdgpu_vcn_ring_end_use,
.emit_wreg = vcn_v2_0_dec_ring_emit_wreg,
.emit_reg_wait = vcn_v2_0_dec_ring_emit_reg_wait,
.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
static const struct amdgpu_ring_funcs vcn_v2_6_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_DEC,
.align_mask = 0xf,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v2_5_dec_ring_get_rptr,
.get_wptr = vcn_v2_5_dec_ring_get_wptr,
.set_wptr = vcn_v2_5_dec_ring_set_wptr,
@ -1693,7 +1672,6 @@ static const struct amdgpu_ring_funcs vcn_v2_5_enc_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_ENC,
.align_mask = 0x3f,
.nop = VCN_ENC_CMD_NO_OP,
.vmhub = AMDGPU_MMHUB_1,
.get_rptr = vcn_v2_5_enc_ring_get_rptr,
.get_wptr = vcn_v2_5_enc_ring_get_wptr,
.set_wptr = vcn_v2_5_enc_ring_set_wptr,
@ -1719,36 +1697,6 @@ static const struct amdgpu_ring_funcs vcn_v2_5_enc_ring_vm_funcs = {
.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
static const struct amdgpu_ring_funcs vcn_v2_6_enc_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_ENC,
.align_mask = 0x3f,
.nop = VCN_ENC_CMD_NO_OP,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v2_5_enc_ring_get_rptr,
.get_wptr = vcn_v2_5_enc_ring_get_wptr,
.set_wptr = vcn_v2_5_enc_ring_set_wptr,
.emit_frame_size =
SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 4 +
4 + /* vcn_v2_0_enc_ring_emit_vm_flush */
5 + 5 + /* vcn_v2_0_enc_ring_emit_fence x2 vm fence */
1, /* vcn_v2_0_enc_ring_insert_end */
.emit_ib_size = 5, /* vcn_v2_0_enc_ring_emit_ib */
.emit_ib = vcn_v2_0_enc_ring_emit_ib,
.emit_fence = vcn_v2_0_enc_ring_emit_fence,
.emit_vm_flush = vcn_v2_0_enc_ring_emit_vm_flush,
.test_ring = amdgpu_vcn_enc_ring_test_ring,
.test_ib = amdgpu_vcn_enc_ring_test_ib,
.insert_nop = amdgpu_ring_insert_nop,
.insert_end = vcn_v2_0_enc_ring_insert_end,
.pad_ib = amdgpu_ring_generic_pad_ib,
.begin_use = amdgpu_vcn_ring_begin_use,
.end_use = amdgpu_vcn_ring_end_use,
.emit_wreg = vcn_v2_0_enc_ring_emit_wreg,
.emit_reg_wait = vcn_v2_0_enc_ring_emit_reg_wait,
.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
};
static void vcn_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
{
int i;
@ -1756,10 +1704,7 @@ static void vcn_v2_5_set_dec_ring_funcs(struct amdgpu_device *adev)
for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
if (adev->vcn.harvest_config & (1 << i))
continue;
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
adev->vcn.inst[i].ring_dec.funcs = &vcn_v2_5_dec_ring_vm_funcs;
else /* CHIP_ALDEBARAN */
adev->vcn.inst[i].ring_dec.funcs = &vcn_v2_6_dec_ring_vm_funcs;
adev->vcn.inst[i].ring_dec.funcs = &vcn_v2_5_dec_ring_vm_funcs;
adev->vcn.inst[i].ring_dec.me = i;
DRM_INFO("VCN(%d) decode is enabled in VM mode\n", i);
}
@ -1773,10 +1718,7 @@ static void vcn_v2_5_set_enc_ring_funcs(struct amdgpu_device *adev)
if (adev->vcn.harvest_config & (1 << j))
continue;
for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
adev->vcn.inst[j].ring_enc[i].funcs = &vcn_v2_5_enc_ring_vm_funcs;
else /* CHIP_ALDEBARAN */
adev->vcn.inst[j].ring_enc[i].funcs = &vcn_v2_6_enc_ring_vm_funcs;
adev->vcn.inst[j].ring_enc[i].funcs = &vcn_v2_5_enc_ring_vm_funcs;
adev->vcn.inst[j].ring_enc[i].me = j;
}
DRM_INFO("VCN(%d) encode is enabled in VM mode\n", j);

5
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
View file

@ -189,6 +189,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i;
}
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
AMDGPU_RING_PRIO_DEFAULT,
@ -212,6 +213,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + j + 8 * i;
}
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_enc_%d.%d", i, j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
hw_prio, &adev->vcn.inst[i].sched_score);
@ -1738,7 +1740,6 @@ static const struct amdgpu_ring_funcs vcn_v3_0_dec_sw_ring_vm_funcs = {
.align_mask = 0x3f,
.nop = VCN_DEC_SW_CMD_NO_OP,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v3_0_dec_ring_get_rptr,
.get_wptr = vcn_v3_0_dec_ring_get_wptr,
.set_wptr = vcn_v3_0_dec_ring_set_wptr,
@ -1899,7 +1900,6 @@ static const struct amdgpu_ring_funcs vcn_v3_0_dec_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_DEC,
.align_mask = 0xf,
.secure_submission_supported = true,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v3_0_dec_ring_get_rptr,
.get_wptr = vcn_v3_0_dec_ring_get_wptr,
.set_wptr = vcn_v3_0_dec_ring_set_wptr,
@ -2000,7 +2000,6 @@ static const struct amdgpu_ring_funcs vcn_v3_0_enc_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_ENC,
.align_mask = 0x3f,
.nop = VCN_ENC_CMD_NO_OP,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v3_0_enc_ring_get_rptr,
.get_wptr = vcn_v3_0_enc_ring_get_wptr,
.set_wptr = vcn_v3_0_enc_ring_set_wptr,

3
drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
View file

@ -149,7 +149,7 @@ static int vcn_v4_0_sw_init(void *handle)
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + i * (adev->vcn.num_enc_rings + 1) + 1;
else
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i;
ring->vm_hub = AMDGPU_MMHUB_0;
sprintf(ring->name, "vcn_unified_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
@ -1798,7 +1798,6 @@ static const struct amdgpu_ring_funcs vcn_v4_0_unified_ring_vm_funcs = {
.type = AMDGPU_RING_TYPE_VCN_ENC,
.align_mask = 0x3f,
.nop = VCN_ENC_CMD_NO_OP,
.vmhub = AMDGPU_MMHUB_0,
.get_rptr = vcn_v4_0_unified_ring_get_rptr,
.get_wptr = vcn_v4_0_unified_ring_get_wptr,
.set_wptr = vcn_v4_0_unified_ring_set_wptr,