diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu.h | 8 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 38 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 8 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 16 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 6 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 46 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 47 |
7 files changed, 116 insertions, 53 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index e9fde72cf038..31b00f91cfcd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1866,6 +1866,12 @@ typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t); typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t); typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t); +struct amdgpu_ip_block_status { + bool valid; + bool sw; + bool hw; +}; + struct amdgpu_device { struct device *dev; struct drm_device *ddev; @@ -2008,7 +2014,7 @@ struct amdgpu_device { const struct amdgpu_ip_block_version *ip_blocks; int num_ip_blocks; - bool *ip_block_enabled; + struct amdgpu_ip_block_status *ip_block_status; struct mutex mn_lock; DECLARE_HASHTABLE(mn_hash, 7); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index d79009b65867..99f158e1baff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1191,8 +1191,9 @@ static int amdgpu_early_init(struct amdgpu_device *adev) return -EINVAL; } - adev->ip_block_enabled = kcalloc(adev->num_ip_blocks, sizeof(bool), GFP_KERNEL); - if (adev->ip_block_enabled == NULL) + adev->ip_block_status = kcalloc(adev->num_ip_blocks, + sizeof(struct amdgpu_ip_block_status), GFP_KERNEL); + if (adev->ip_block_status == NULL) return -ENOMEM; if (adev->ip_blocks == NULL) { @@ -1203,18 +1204,18 @@ static int amdgpu_early_init(struct amdgpu_device *adev) for (i = 0; i < adev->num_ip_blocks; i++) { if ((amdgpu_ip_block_mask & (1 << i)) == 0) { DRM_ERROR("disabled ip block: %d\n", i); - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].valid = false; } else { if (adev->ip_blocks[i].funcs->early_init) { r = adev->ip_blocks[i].funcs->early_init((void *)adev); if (r == -ENOENT) - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].valid = false; else if (r) return r; else - adev->ip_block_enabled[i] = true; + adev->ip_block_status[i].valid = true; } else { - adev->ip_block_enabled[i] = true; + adev->ip_block_status[i].valid = true; } } } @@ -1227,11 +1228,12 @@ static int amdgpu_init(struct amdgpu_device *adev) int i, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; r = adev->ip_blocks[i].funcs->sw_init((void *)adev); if (r) return r; + adev->ip_block_status[i].sw = true; /* need to do gmc hw init early so we can allocate gpu mem */ if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) { r = amdgpu_vram_scratch_init(adev); @@ -1243,11 +1245,12 @@ static int amdgpu_init(struct amdgpu_device *adev) r = amdgpu_wb_init(adev); if (r) return r; + adev->ip_block_status[i].hw = true; } } for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].sw) continue; /* gmc hw init is done early */ if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) @@ -1255,6 +1258,7 @@ static int amdgpu_init(struct amdgpu_device *adev) r = adev->ip_blocks[i].funcs->hw_init((void *)adev); if (r) return r; + adev->ip_block_status[i].hw = true; } return 0; @@ -1265,7 +1269,7 @@ static int amdgpu_late_init(struct amdgpu_device *adev) int i = 0, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; /* enable clockgating to save power */ r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev, @@ -1287,7 +1291,7 @@ static int amdgpu_fini(struct amdgpu_device *adev) int i, r; for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].hw) continue; if (adev->ip_blocks[i].type == AMD_IP_BLOCK_TYPE_GMC) { amdgpu_wb_fini(adev); @@ -1300,14 +1304,16 @@ static int amdgpu_fini(struct amdgpu_device *adev) return r; r = adev->ip_blocks[i].funcs->hw_fini((void *)adev); /* XXX handle errors */ + adev->ip_block_status[i].hw = false; } for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].sw) continue; r = adev->ip_blocks[i].funcs->sw_fini((void *)adev); /* XXX handle errors */ - adev->ip_block_enabled[i] = false; + adev->ip_block_status[i].sw = false; + adev->ip_block_status[i].valid = false; } return 0; @@ -1318,7 +1324,7 @@ static int amdgpu_suspend(struct amdgpu_device *adev) int i, r; for (i = adev->num_ip_blocks - 1; i >= 0; i--) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; /* ungate blocks so that suspend can properly shut them down */ r = adev->ip_blocks[i].funcs->set_clockgating_state((void *)adev, @@ -1336,7 +1342,7 @@ static int amdgpu_resume(struct amdgpu_device *adev) int i, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_block_enabled[i]) + if (!adev->ip_block_status[i].valid) continue; r = adev->ip_blocks[i].funcs->resume(adev); if (r) @@ -1582,8 +1588,8 @@ void amdgpu_device_fini(struct amdgpu_device *adev) amdgpu_fence_driver_fini(adev); amdgpu_fbdev_fini(adev); r = amdgpu_fini(adev); - kfree(adev->ip_block_enabled); - adev->ip_block_enabled = NULL; + kfree(adev->ip_block_status); + adev->ip_block_status = NULL; adev->accel_working = false; /* free i2c buses */ amdgpu_i2c_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index ae43b58c9733..4afc507820c0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -449,7 +449,7 @@ out: * vital here, so they are not reported back to userspace. */ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, - struct amdgpu_bo_va *bo_va) + struct amdgpu_bo_va *bo_va, uint32_t operation) { struct ttm_validate_buffer tv, *entry; struct amdgpu_bo_list_entry *vm_bos; @@ -485,7 +485,9 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, if (r) goto error_unlock; - r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem); + + if (operation == AMDGPU_VA_OP_MAP) + r = amdgpu_vm_bo_update(adev, bo_va, &bo_va->bo->tbo.mem); error_unlock: mutex_unlock(&bo_va->vm->mutex); @@ -580,7 +582,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, } if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE)) - amdgpu_gem_va_update_vm(adev, bo_va); + amdgpu_gem_va_update_vm(adev, bo_va, args->operation); drm_gem_object_unreference_unlocked(gobj); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 52dff75aac6f..bc0fac618a3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -180,16 +180,16 @@ int amdgpu_ib_schedule(struct amdgpu_device *adev, unsigned num_ibs, if (vm) { /* do context switch */ amdgpu_vm_flush(ring, vm, ib->sync.last_vm_update); - } - if (vm && ring->funcs->emit_gds_switch) - amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id, - ib->gds_base, ib->gds_size, - ib->gws_base, ib->gws_size, - ib->oa_base, ib->oa_size); + if (ring->funcs->emit_gds_switch) + amdgpu_ring_emit_gds_switch(ring, ib->vm->ids[ring->idx].id, + ib->gds_base, ib->gds_size, + ib->gws_base, ib->gws_size, + ib->oa_base, ib->oa_size); - if (ring->funcs->emit_hdp_flush) - amdgpu_ring_emit_hdp_flush(ring); + if (ring->funcs->emit_hdp_flush) + amdgpu_ring_emit_hdp_flush(ring); + } old_ctx = ring->current_ctx; for (i = 0; i < num_ibs; ++i) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 31ad444c6386..9736892bcdf9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -235,7 +235,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file for (i = 0; i < adev->num_ip_blocks; i++) { if (adev->ip_blocks[i].type == type && - adev->ip_block_enabled[i]) { + adev->ip_block_status[i].valid) { ip.hw_ip_version_major = adev->ip_blocks[i].major; ip.hw_ip_version_minor = adev->ip_blocks[i].minor; ip.capabilities_flags = 0; @@ -274,7 +274,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file for (i = 0; i < adev->num_ip_blocks; i++) if (adev->ip_blocks[i].type == type && - adev->ip_block_enabled[i] && + adev->ip_block_status[i].valid && count < AMDGPU_HW_IP_INSTANCE_MAX_COUNT) count++; @@ -416,7 +416,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file return n ? -EFAULT : 0; } case AMDGPU_INFO_DEV_INFO: { - struct drm_amdgpu_info_device dev_info; + struct drm_amdgpu_info_device dev_info = {}; struct amdgpu_cu_info cu_info; dev_info.device_id = dev->pdev->device; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 2c188fb9fd22..2db6ab0a543d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2561,7 +2561,7 @@ static bool gfx_v7_0_ring_emit_semaphore(struct amdgpu_ring *ring, * sheduling on the ring. This function schedules the IB * on the gfx ring for execution by the GPU. */ -static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, +static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { bool need_ctx_switch = ring->current_ctx != ib->ctx; @@ -2569,15 +2569,10 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, u32 next_rptr = ring->wptr + 5; /* drop the CE preamble IB for the same context */ - if ((ring->type == AMDGPU_RING_TYPE_GFX) && - (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && - !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) return; - if (ring->type == AMDGPU_RING_TYPE_COMPUTE) - control |= INDIRECT_BUFFER_VALID; - - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) + if (need_ctx_switch) next_rptr += 2; next_rptr += 4; @@ -2588,7 +2583,7 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) { + if (need_ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -2611,6 +2606,35 @@ static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, control); } +static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_ib *ib) +{ + u32 header, control = 0; + u32 next_rptr = ring->wptr + 5; + + control |= INDIRECT_BUFFER_VALID; + next_rptr += 4; + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); + amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + amdgpu_ring_write(ring, next_rptr); + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + + control |= ib->length_dw | + (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0); + + amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); + amdgpu_ring_write(ring, control); +} + /** * gfx_v7_0_ring_test_ib - basic ring IB test * @@ -5555,7 +5579,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .get_wptr = gfx_v7_0_ring_get_wptr_gfx, .set_wptr = gfx_v7_0_ring_set_wptr_gfx, .parse_cs = NULL, - .emit_ib = gfx_v7_0_ring_emit_ib, + .emit_ib = gfx_v7_0_ring_emit_ib_gfx, .emit_fence = gfx_v7_0_ring_emit_fence_gfx, .emit_semaphore = gfx_v7_0_ring_emit_semaphore, .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, @@ -5571,7 +5595,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .get_wptr = gfx_v7_0_ring_get_wptr_compute, .set_wptr = gfx_v7_0_ring_set_wptr_compute, .parse_cs = NULL, - .emit_ib = gfx_v7_0_ring_emit_ib, + .emit_ib = gfx_v7_0_ring_emit_ib_compute, .emit_fence = gfx_v7_0_ring_emit_fence_compute, .emit_semaphore = gfx_v7_0_ring_emit_semaphore, .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 1c7c992dea37..9e1d4ddbf475 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -3753,7 +3753,7 @@ static void gfx_v8_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0x20); /* poll interval */ } -static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, +static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib) { bool need_ctx_switch = ring->current_ctx != ib->ctx; @@ -3761,15 +3761,10 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, u32 next_rptr = ring->wptr + 5; /* drop the CE preamble IB for the same context */ - if ((ring->type == AMDGPU_RING_TYPE_GFX) && - (ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && - !need_ctx_switch) + if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) && !need_ctx_switch) return; - if (ring->type == AMDGPU_RING_TYPE_COMPUTE) - control |= INDIRECT_BUFFER_VALID; - - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) + if (need_ctx_switch) next_rptr += 2; next_rptr += 4; @@ -3780,7 +3775,7 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, next_rptr); /* insert SWITCH_BUFFER packet before first IB in the ring frame */ - if (need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) { + if (need_ctx_switch) { amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); amdgpu_ring_write(ring, 0); } @@ -3803,6 +3798,36 @@ static void gfx_v8_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, control); } +static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring, + struct amdgpu_ib *ib) +{ + u32 header, control = 0; + u32 next_rptr = ring->wptr + 5; + + control |= INDIRECT_BUFFER_VALID; + + next_rptr += 4; + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM); + amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + amdgpu_ring_write(ring, next_rptr); + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + + control |= ib->length_dw | + (ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0); + + amdgpu_ring_write(ring, header); + amdgpu_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); + amdgpu_ring_write(ring, control); +} + static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags) { @@ -4224,7 +4249,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .get_wptr = gfx_v8_0_ring_get_wptr_gfx, .set_wptr = gfx_v8_0_ring_set_wptr_gfx, .parse_cs = NULL, - .emit_ib = gfx_v8_0_ring_emit_ib, + .emit_ib = gfx_v8_0_ring_emit_ib_gfx, .emit_fence = gfx_v8_0_ring_emit_fence_gfx, .emit_semaphore = gfx_v8_0_ring_emit_semaphore, .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, @@ -4240,7 +4265,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .get_wptr = gfx_v8_0_ring_get_wptr_compute, .set_wptr = gfx_v8_0_ring_set_wptr_compute, .parse_cs = NULL, - .emit_ib = gfx_v8_0_ring_emit_ib, + .emit_ib = gfx_v8_0_ring_emit_ib_compute, .emit_fence = gfx_v8_0_ring_emit_fence_compute, .emit_semaphore = gfx_v8_0_ring_emit_semaphore, .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, |