summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c')
-rw-r--r--drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c120
1 files changed, 51 insertions, 69 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 0d8bf2cb1956..e992bf2ff66c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -2173,7 +2173,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev)
adev->gfx.config.num_tile_pipes = adev->gfx.config.max_tile_pipes;
adev->gfx.config.mem_max_burst_length_bytes = 256;
- if (adev->flags & AMDGPU_IS_APU) {
+ if (adev->flags & AMD_IS_APU) {
/* Get memory bank mapping mode. */
tmp = RREG32(mmMC_FUS_DRAM0_BANK_ADDR_MAPPING);
dimm00_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM0_BANK_ADDR_MAPPING, DIMM0ADDRMAP);
@@ -2648,6 +2648,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
struct amdgpu_ib ib;
+ struct fence *f = NULL;
uint32_t scratch;
uint32_t tmp = 0;
unsigned i;
@@ -2659,29 +2660,27 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
+ memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(ring, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%d).\n", r);
- amdgpu_gfx_scratch_free(adev, scratch);
- return r;
+ goto err1;
}
ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1);
ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START));
ib.ptr[2] = 0xDEADBEEF;
ib.length_dw = 3;
- r = amdgpu_ib_schedule(adev, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED);
- if (r) {
- amdgpu_gfx_scratch_free(adev, scratch);
- amdgpu_ib_free(adev, &ib);
- DRM_ERROR("amdgpu: failed to schedule ib (%d).\n", r);
- return r;
- }
- r = amdgpu_fence_wait(ib.fence, false);
+
+ r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, &ib, 1, NULL,
+ AMDGPU_FENCE_OWNER_UNDEFINED,
+ &f);
+ if (r)
+ goto err2;
+
+ r = fence_wait(f, false);
if (r) {
DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
- amdgpu_gfx_scratch_free(adev, scratch);
- amdgpu_ib_free(adev, &ib);
- return r;
+ goto err2;
}
for (i = 0; i < adev->usec_timeout; i++) {
tmp = RREG32(scratch);
@@ -2691,14 +2690,19 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
}
if (i < adev->usec_timeout) {
DRM_INFO("ib test on ring %d succeeded in %u usecs\n",
- ib.fence->ring->idx, i);
+ ring->idx, i);
+ goto err2;
} else {
DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n",
scratch, tmp);
r = -EINVAL;
}
- amdgpu_gfx_scratch_free(adev, scratch);
+
+err2:
+ fence_put(f);
amdgpu_ib_free(adev, &ib);
+err1:
+ amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
@@ -3202,7 +3206,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
r = amdgpu_bo_create(adev,
adev->gfx.mec.num_mec *adev->gfx.mec.num_pipe * MEC_HPD_SIZE * 2,
PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+ AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
&adev->gfx.mec.hpd_eop_obj);
if (r) {
dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
@@ -3369,7 +3373,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
r = amdgpu_bo_create(adev,
sizeof(struct bonaire_mqd),
PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+ AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
&ring->mqd_obj);
if (r) {
dev_warn(adev->dev, "(%d) create MQD bo failed\n", r);
@@ -3606,41 +3610,6 @@ static int gfx_v7_0_cp_resume(struct amdgpu_device *adev)
return 0;
}
-static void gfx_v7_0_ce_sync_me(struct amdgpu_ring *ring)
-{
- struct amdgpu_device *adev = ring->adev;
- u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
-
- /* instruct DE to set a magic number */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
- WRITE_DATA_DST_SEL(5)));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
-
- /* let CE wait till condition satisfied */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
- amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
- WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
- WAIT_REG_MEM_FUNCTION(3) | /* == */
- WAIT_REG_MEM_ENGINE(2))); /* ce */
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 1);
- amdgpu_ring_write(ring, 0xffffffff);
- amdgpu_ring_write(ring, 4); /* poll interval */
-
- /* instruct CE to reset wb of ce_sync to zero */
- amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
- amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
- WRITE_DATA_DST_SEL(5) |
- WR_CONFIRM));
- amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
- amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
- amdgpu_ring_write(ring, 0);
-}
-
/*
* vm
* VMID 0 is the physical GPU addresses as used by the kernel.
@@ -3659,6 +3628,13 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
unsigned vm_id, uint64_t pd_addr)
{
int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+ if (usepfp) {
+ /* synce CE with ME to prevent CE fetch CEIB before context switch done */
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ }
amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
@@ -3699,7 +3675,10 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, 0x0);
/* synce CE with ME to prevent CE fetch CEIB before context switch done */
- gfx_v7_0_ce_sync_me(ring);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
+ amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+ amdgpu_ring_write(ring, 0);
}
}
@@ -3758,7 +3737,7 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
int r;
/* allocate rlc buffers */
- if (adev->flags & AMDGPU_IS_APU) {
+ if (adev->flags & AMD_IS_APU) {
if (adev->asic_type == CHIP_KAVERI) {
adev->gfx.rlc.reg_list = spectre_rlc_save_restore_register_list;
adev->gfx.rlc.reg_list_size =
@@ -3782,7 +3761,10 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
/* save restore block */
if (adev->gfx.rlc.save_restore_obj == NULL) {
r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.save_restore_obj);
+ AMDGPU_GEM_DOMAIN_VRAM,
+ AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
+ NULL, NULL,
+ &adev->gfx.rlc.save_restore_obj);
if (r) {
dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r);
return r;
@@ -3823,7 +3805,10 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
if (adev->gfx.rlc.clear_state_obj == NULL) {
r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.clear_state_obj);
+ AMDGPU_GEM_DOMAIN_VRAM,
+ AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
+ NULL, NULL,
+ &adev->gfx.rlc.clear_state_obj);
if (r) {
dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
gfx_v7_0_rlc_fini(adev);
@@ -3860,7 +3845,10 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
if (adev->gfx.rlc.cp_table_size) {
if (adev->gfx.rlc.cp_table_obj == NULL) {
r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true,
- AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.cp_table_obj);
+ AMDGPU_GEM_DOMAIN_VRAM,
+ AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
+ NULL, NULL,
+ &adev->gfx.rlc.cp_table_obj);
if (r) {
dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r);
gfx_v7_0_rlc_fini(adev);
@@ -4792,12 +4780,6 @@ static int gfx_v7_0_sw_init(void *handle)
return r;
}
- r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
- if (r) {
- DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
- return r;
- }
-
for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
ring = &adev->gfx.gfx_ring[i];
ring->ring_obj = NULL;
@@ -4841,21 +4823,21 @@ static int gfx_v7_0_sw_init(void *handle)
r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size,
PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_GDS, 0,
- NULL, &adev->gds.gds_gfx_bo);
+ NULL, NULL, &adev->gds.gds_gfx_bo);
if (r)
return r;
r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size,
PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_GWS, 0,
- NULL, &adev->gds.gws_gfx_bo);
+ NULL, NULL, &adev->gds.gws_gfx_bo);
if (r)
return r;
r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size,
PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_OA, 0,
- NULL, &adev->gds.oa_gfx_bo);
+ NULL, NULL, &adev->gds.oa_gfx_bo);
if (r)
return r;
@@ -4876,8 +4858,6 @@ static int gfx_v7_0_sw_fini(void *handle)
for (i = 0; i < adev->gfx.num_compute_rings; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
- amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
-
gfx_v7_0_cp_compute_fini(adev);
gfx_v7_0_rlc_fini(adev);
gfx_v7_0_mec_fini(adev);
@@ -5594,6 +5574,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
.test_ring = gfx_v7_0_ring_test_ring,
.test_ib = gfx_v7_0_ring_test_ib,
.is_lockup = gfx_v7_0_ring_is_lockup,
+ .insert_nop = amdgpu_ring_insert_nop,
};
static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
@@ -5610,6 +5591,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
.test_ring = gfx_v7_0_ring_test_ring,
.test_ib = gfx_v7_0_ring_test_ib,
.is_lockup = gfx_v7_0_ring_is_lockup,
+ .insert_nop = amdgpu_ring_insert_nop,
};
static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev)
generated by cgit v1.2.3 (git 2.39.0) at 2025-03-02 08:20:34 +0300