diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 44 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/psp_v11_0.c | 19 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 12 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 64 |
6 files changed, 75 insertions, 68 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 260165bbe373..b16cce7c22c3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -213,19 +213,35 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, spin_lock(&kfd_mem_limit.mem_limit_lock); if (kfd_mem_limit.system_mem_used + system_mem_needed > - kfd_mem_limit.max_system_mem_limit) + kfd_mem_limit.max_system_mem_limit) { pr_debug("Set no_system_mem_limit=1 if using shared memory\n"); + if (!no_system_mem_limit) { + ret = -ENOMEM; + goto release; + } + } - if ((kfd_mem_limit.system_mem_used + system_mem_needed > - kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) || - (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > - kfd_mem_limit.max_ttm_mem_limit) || - (adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] + vram_needed > - vram_size - reserved_for_pt - reserved_for_ras - atomic64_read(&adev->vram_pin_size))) { + if (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > + kfd_mem_limit.max_ttm_mem_limit) { ret = -ENOMEM; goto release; } + /*if is_app_apu is false and apu_prefer_gtt is true, it is an APU with + * carve out < gtt. In that case, VRAM allocation will go to gtt domain, skip + * VRAM check since ttm_mem_limit check already cover this allocation + */ + + if (adev && xcp_id >= 0 && (!adev->apu_prefer_gtt || adev->gmc.is_app_apu)) { + uint64_t vram_available = + vram_size - reserved_for_pt - reserved_for_ras - + atomic64_read(&adev->vram_pin_size); + if (adev->kfd.vram_used[xcp_id] + vram_needed > vram_available) { + ret = -ENOMEM; + goto release; + } + } + /* Update memory accounting by decreasing available system * memory, TTM memory and GPU memory as computed above */ @@ -1626,11 +1642,15 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev, uint64_t vram_available, system_mem_available, ttm_mem_available; spin_lock(&kfd_mem_limit.mem_limit_lock); - vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id) - - adev->kfd.vram_used_aligned[xcp_id] - - atomic64_read(&adev->vram_pin_size) - - reserved_for_pt - - reserved_for_ras; + if (adev->apu_prefer_gtt && !adev->gmc.is_app_apu) + vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id) + - adev->kfd.vram_used_aligned[xcp_id]; + else + vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id) + - adev->kfd.vram_used_aligned[xcp_id] + - atomic64_read(&adev->vram_pin_size) + - reserved_for_pt + - reserved_for_ras; if (adev->apu_prefer_gtt) { system_mem_available = no_system_mem_limit ? diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 6379bb25bf5c..486c3646710c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -421,8 +421,6 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) dma_fence_put(ring->vmid_wait); ring->vmid_wait = NULL; ring->me = 0; - - ring->adev->rings[ring->idx] = NULL; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c index a887df520414..4258d3e0b706 100644 --- a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c +++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c @@ -29,6 +29,8 @@ #include "amdgpu.h" #include "isp_v4_1_1.h" +MODULE_FIRMWARE("amdgpu/isp_4_1_1.bin"); + #define ISP_PERFORMANCE_STATE_LOW 0 #define ISP_PERFORMANCE_STATE_HIGH 1 diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c index 6cc05d36e359..64b240b51f1a 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -149,12 +149,12 @@ static int psp_v11_0_wait_for_bootloader(struct psp_context *psp) int ret; int retry_loop; - for (retry_loop = 0; retry_loop < 10; retry_loop++) { + for (retry_loop = 0; retry_loop < 20; retry_loop++) { /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ ret = psp_wait_for( psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), - 0x80000000, 0x80000000, PSP_WAITREG_NOVERBOSE); + 0x80000000, 0x8000FFFF, PSP_WAITREG_NOVERBOSE); if (ret == 0) return 0; @@ -397,18 +397,6 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp) msleep(500); - offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); - - ret = psp_wait_for(psp, offset, MBOX_TOS_RESP_FLAG, MBOX_TOS_RESP_MASK, - 0); - - if (ret) { - DRM_INFO("psp mode 1 reset failed!\n"); - return -EINVAL; - } - - DRM_INFO("psp mode1 reset succeed \n"); - return 0; } @@ -665,7 +653,8 @@ static const struct psp_funcs psp_v11_0_funcs = { .ring_get_wptr = psp_v11_0_ring_get_wptr, .ring_set_wptr = psp_v11_0_ring_set_wptr, .load_usbc_pd_fw = psp_v11_0_load_usbc_pd_fw, - .read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw + .read_usbc_pd_fw = psp_v11_0_read_usbc_pd_fw, + .wait_for_bootloader = psp_v11_0_wait_for_bootloader }; void psp_v11_0_set_psp_funcs(struct psp_context *psp) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c index 4b8f4407047f..2811226b0ea5 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c @@ -1888,15 +1888,19 @@ static int vcn_v3_0_limit_sched(struct amdgpu_cs_parser *p, struct amdgpu_job *job) { struct drm_gpu_scheduler **scheds; - - /* The create msg must be in the first IB submitted */ - if (atomic_read(&job->base.entity->fence_seq)) - return -EINVAL; + struct dma_fence *fence; /* if VCN0 is harvested, we can't support AV1 */ if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0) return -EINVAL; + /* wait for all jobs to finish before switching to instance 0 */ + fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull); + if (fence) { + dma_fence_wait(fence, false); + dma_fence_put(fence); + } + scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_DEC] [AMDGPU_RING_PRIO_DEFAULT].sched; drm_sched_entity_modify_sched(job->base.entity, scheds, 1); diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c index 1924e075b66f..706f3b2f484f 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c @@ -1808,15 +1808,19 @@ static int vcn_v4_0_limit_sched(struct amdgpu_cs_parser *p, struct amdgpu_job *job) { struct drm_gpu_scheduler **scheds; - - /* The create msg must be in the first IB submitted */ - if (atomic_read(&job->base.entity->fence_seq)) - return -EINVAL; + struct dma_fence *fence; /* if VCN0 is harvested, we can't support AV1 */ if (p->adev->vcn.harvest_config & AMDGPU_VCN_HARVEST_VCN0) return -EINVAL; + /* wait for all jobs to finish before switching to instance 0 */ + fence = amdgpu_ctx_get_fence(p->ctx, job->base.entity, ~0ull); + if (fence) { + dma_fence_wait(fence, false); + dma_fence_put(fence); + } + scheds = p->adev->gpu_sched[AMDGPU_HW_IP_VCN_ENC] [AMDGPU_RING_PRIO_0].sched; drm_sched_entity_modify_sched(job->base.entity, scheds, 1); @@ -1907,22 +1911,16 @@ out: #define RADEON_VCN_ENGINE_TYPE_ENCODE (0x00000002) #define RADEON_VCN_ENGINE_TYPE_DECODE (0x00000003) - #define RADEON_VCN_ENGINE_INFO (0x30000001) -#define RADEON_VCN_ENGINE_INFO_MAX_OFFSET 16 - #define RENCODE_ENCODE_STANDARD_AV1 2 #define RENCODE_IB_PARAM_SESSION_INIT 0x00000003 -#define RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET 64 -/* return the offset in ib if id is found, -1 otherwise - * to speed up the searching we only search upto max_offset - */ -static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int max_offset) +/* return the offset in ib if id is found, -1 otherwise */ +static int vcn_v4_0_enc_find_ib_param(struct amdgpu_ib *ib, uint32_t id, int start) { int i; - for (i = 0; i < ib->length_dw && i < max_offset && ib->ptr[i] >= 8; i += ib->ptr[i]/4) { + for (i = start; i < ib->length_dw && ib->ptr[i] >= 8; i += ib->ptr[i] / 4) { if (ib->ptr[i + 1] == id) return i; } @@ -1937,33 +1935,29 @@ static int vcn_v4_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, struct amdgpu_vcn_decode_buffer *decode_buffer; uint64_t addr; uint32_t val; - int idx; + int idx = 0, sidx; /* The first instance can decode anything */ if (!ring->me) return 0; - /* RADEON_VCN_ENGINE_INFO is at the top of ib block */ - idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO, - RADEON_VCN_ENGINE_INFO_MAX_OFFSET); - if (idx < 0) /* engine info is missing */ - return 0; - - val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */ - if (val == RADEON_VCN_ENGINE_TYPE_DECODE) { - decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6]; - - if (!(decode_buffer->valid_buf_flag & 0x1)) - return 0; - - addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 | - decode_buffer->msg_buffer_address_lo; - return vcn_v4_0_dec_msg(p, job, addr); - } else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) { - idx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT, - RENCODE_IB_PARAM_SESSION_INIT_MAX_OFFSET); - if (idx >= 0 && ib->ptr[idx + 2] == RENCODE_ENCODE_STANDARD_AV1) - return vcn_v4_0_limit_sched(p, job); + while ((idx = vcn_v4_0_enc_find_ib_param(ib, RADEON_VCN_ENGINE_INFO, idx)) >= 0) { + val = amdgpu_ib_get_value(ib, idx + 2); /* RADEON_VCN_ENGINE_TYPE */ + if (val == RADEON_VCN_ENGINE_TYPE_DECODE) { + decode_buffer = (struct amdgpu_vcn_decode_buffer *)&ib->ptr[idx + 6]; + + if (!(decode_buffer->valid_buf_flag & 0x1)) + return 0; + + addr = ((u64)decode_buffer->msg_buffer_address_hi) << 32 | + decode_buffer->msg_buffer_address_lo; + return vcn_v4_0_dec_msg(p, job, addr); + } else if (val == RADEON_VCN_ENGINE_TYPE_ENCODE) { + sidx = vcn_v4_0_enc_find_ib_param(ib, RENCODE_IB_PARAM_SESSION_INIT, idx); + if (sidx >= 0 && ib->ptr[sidx + 2] == RENCODE_ENCODE_STANDARD_AV1) + return vcn_v4_0_limit_sched(p, job); + } + idx += ib->ptr[idx] / 4; } return 0; } |