xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'drm-next-4.21' of git://people.freedesktop.org/~agd5f/linux into drm-next

Browse Source 
New features for 4.21:
amdgpu:
- Support for SDMA paging queue on vega
- Put compute EOP buffers into vram for better performance
- Share more code with amdkfd
- Support for scanout with DCC on gfx9
- Initial kerneldoc for DC
- Updated SMU firmware support for gfx8 chips
- Rework CSA handling for eventual support for preemption
- XGMI PSP support
- Clean up RLC handling
- Enable GPU reset by default on VI, SOC15 dGPUs
- Ring and IB test cleanups

amdkfd:
- Share more code with amdgpu

ttm:
- Move global init out of the drivers

scheduler:
- Track if schedulers are ready for work
- Timeout/fault handling changes to facilitate GPU recovery

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181114165113.3751-1-alexander.deucher@amd.com
This commit is contained in:
Dave Airlie
2018-11-19 11:07:52 +10:00
parent d7563c55ef 36b486bc68
commit 9235dd441a
207 changed files with 5633 additions and 5465 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
drivers/gpu/drm/amd/amdgpu/Makefile
View File

@@ -53,7 +53,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \
amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o amdgpu_ids.o \
amdgpu_gmc.o amdgpu_xgmi.o
amdgpu_gmc.o amdgpu_xgmi.o amdgpu_csa.o
# add asic specific block
amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \
@@ -105,6 +105,7 @@ amdgpu-y += \
# add GFX block
amdgpu-y += \
amdgpu_gfx.o \
amdgpu_rlc.o \
gfx_v8_0.o \
gfx_v9_0.o

4
drivers/gpu/drm/amd/amdgpu/amdgpu.h
View File

@@ -75,6 +75,7 @@
#include "amdgpu_sdma.h"
#include "amdgpu_dm.h"
#include "amdgpu_virt.h"
#include "amdgpu_csa.h"
#include "amdgpu_gart.h"
#include "amdgpu_debugfs.h"
#include "amdgpu_job.h"
@@ -433,7 +434,7 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
* default non-graphics QWORD index is 0xe0 - 0xFF inclusive
*/
/* sDMA engines reserved from 0xe0 -oxef */
/* sDMA engines reserved from 0xe0 -0xef */
AMDGPU_DOORBELL64_sDMA_ENGINE0 = 0xE0,
AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xE1,
AMDGPU_DOORBELL64_sDMA_ENGINE1 = 0xE8,
@@ -831,7 +832,6 @@ struct amdgpu_device {
bool need_dma32;
bool need_swiotlb;
bool accel_working;
struct work_struct reset_work;
struct notifier_block acpi_nb;
struct amdgpu_i2c_chan *i2c_bus[AMDGPU_MAX_I2C_BUS];
struct amdgpu_debugfs debugfs[AMDGPU_DEBUGFS_MAX_COMPONENTS];

20
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
View File

@@ -144,7 +144,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
KGD_MAX_QUEUES);
/* remove the KIQ bit as well */
if (adev->gfx.kiq.ring.ready)
if (adev->gfx.kiq.ring.sched.ready)
clear_bit(amdgpu_gfx_queue_to_bit(adev,
adev->gfx.kiq.ring.me - 1,
adev->gfx.kiq.ring.pipe,
@@ -268,9 +268,9 @@ void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd)
amdgpu_device_gpu_recover(adev, NULL);
}
int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
void **mem_obj, uint64_t *gpu_addr,
void **cpu_ptr, bool mqd_gfx9)
int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
void **mem_obj, uint64_t *gpu_addr,
void **cpu_ptr, bool mqd_gfx9)
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
struct amdgpu_bo *bo = NULL;
@@ -340,7 +340,7 @@ allocate_mem_reserve_bo_failed:
return r;
}
void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
{
struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj;
@@ -351,8 +351,8 @@ void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj)
amdgpu_bo_unref(&(bo));
}
void get_local_mem_info(struct kgd_dev *kgd,
struct kfd_local_mem_info *mem_info)
void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
struct kfd_local_mem_info *mem_info)
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
uint64_t address_mask = adev->dev->dma_mask ? ~*adev->dev->dma_mask :
@@ -383,7 +383,7 @@ void get_local_mem_info(struct kgd_dev *kgd,
mem_info->mem_clk_max = 100;
}
uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd)
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
@@ -392,7 +392,7 @@ uint64_t get_gpu_clock_counter(struct kgd_dev *kgd)
return 0;
}
uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
@@ -405,7 +405,7 @@ uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd)
return 100;
}
void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info)
{
struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
struct amdgpu_cu_info acu_info = adev->gfx.cu_info;

18
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
View File

@@ -134,16 +134,16 @@ int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);
void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd);
/* Shared API */
int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
void **mem_obj, uint64_t *gpu_addr,
void **cpu_ptr, bool mqd_gfx9);
void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
void get_local_mem_info(struct kgd_dev *kgd,
struct kfd_local_mem_info *mem_info);
uint64_t get_gpu_clock_counter(struct kgd_dev *kgd);
int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
void **mem_obj, uint64_t *gpu_addr,
void **cpu_ptr, bool mqd_gfx9);
void amdgpu_amdkfd_free_gtt_mem(struct kgd_dev *kgd, void *mem_obj);
void amdgpu_amdkfd_get_local_mem_info(struct kgd_dev *kgd,
struct kfd_local_mem_info *mem_info);
uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct kgd_dev *kgd);
uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info);
uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct kgd_dev *kgd);
void amdgpu_amdkfd_get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info);
uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd);
uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd);

25
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
View File

@@ -173,13 +173,6 @@ static int get_tile_config(struct kgd_dev *kgd,
}
static const struct kfd2kgd_calls kfd2kgd = {
.init_gtt_mem_allocation = alloc_gtt_mem,
.free_gtt_mem = free_gtt_mem,
.get_local_mem_info = get_local_mem_info,
.get_gpu_clock_counter = get_gpu_clock_counter,
.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
.alloc_pasid = amdgpu_pasid_alloc,
.free_pasid = amdgpu_pasid_free,
.program_sh_mem_settings = kgd_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
.init_interrupts = kgd_init_interrupts,
@@ -200,28 +193,10 @@ static const struct kfd2kgd_calls kfd2kgd = {
.get_fw_version = get_fw_version,
.set_scratch_backing_va = set_scratch_backing_va,
.get_tile_config = get_tile_config,
.get_cu_info = get_cu_info,
.get_vram_usage = amdgpu_amdkfd_get_vram_usage,
.create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm,
.acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm,
.destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm,
.release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm,
.get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir,
.set_vm_context_page_table_base = set_vm_context_page_table_base,
.alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu,
.free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu,
.map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu,
.unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu,
.sync_memory = amdgpu_amdkfd_gpuvm_sync_memory,
.map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel,
.restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos,
.invalidate_tlbs = invalidate_tlbs,
.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
.submit_ib = amdgpu_amdkfd_submit_ib,
.get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info,
.read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg,
.gpu_recover = amdgpu_amdkfd_gpu_reset,
.set_compute_idle = amdgpu_amdkfd_set_compute_idle
};
struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void)

25
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
View File

@@ -128,13 +128,6 @@ static int get_tile_config(struct kgd_dev *kgd,
}
static const struct kfd2kgd_calls kfd2kgd = {
.init_gtt_mem_allocation = alloc_gtt_mem,
.free_gtt_mem = free_gtt_mem,
.get_local_mem_info = get_local_mem_info,
.get_gpu_clock_counter = get_gpu_clock_counter,
.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
.alloc_pasid = amdgpu_pasid_alloc,
.free_pasid = amdgpu_pasid_free,
.program_sh_mem_settings = kgd_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
.init_interrupts = kgd_init_interrupts,
@@ -157,27 +150,9 @@ static const struct kfd2kgd_calls kfd2kgd = {
.get_fw_version = get_fw_version,
.set_scratch_backing_va = set_scratch_backing_va,
.get_tile_config = get_tile_config,
.get_cu_info = get_cu_info,
.get_vram_usage = amdgpu_amdkfd_get_vram_usage,
.create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm,
.acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm,
.destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm,
.release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm,
.get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir,
.set_vm_context_page_table_base = set_vm_context_page_table_base,
.alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu,
.free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu,
.map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu,
.unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu,
.sync_memory = amdgpu_amdkfd_gpuvm_sync_memory,
.map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel,
.restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos,
.invalidate_tlbs = invalidate_tlbs,
.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
.submit_ib = amdgpu_amdkfd_submit_ib,
.get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info,
.gpu_recover = amdgpu_amdkfd_gpu_reset,
.set_compute_idle = amdgpu_amdkfd_set_compute_idle
};
struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void)

120
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
View File

@@ -46,38 +46,9 @@
#include "v9_structs.h"
#include "soc15.h"
#include "soc15d.h"
#include "mmhub_v1_0.h"
#include "gfxhub_v1_0.h"
/* HACK: MMHUB and GC both have VM-related register with the same
* names but different offsets. Define the MMHUB register we need here
* with a prefix. A proper solution would be to move the functions
* programming these registers into gfx_v9_0.c and mmhub_v1_0.c
* respectively.
*/
#define mmMMHUB_VM_INVALIDATE_ENG16_REQ 0x06f3
#define mmMMHUB_VM_INVALIDATE_ENG16_REQ_BASE_IDX 0
#define mmMMHUB_VM_INVALIDATE_ENG16_ACK 0x0705
#define mmMMHUB_VM_INVALIDATE_ENG16_ACK_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 0x072b
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 0x072c
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 0x074b
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 0x074c
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 0x076b
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32_BASE_IDX 0
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 0x076c
#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32_BASE_IDX 0
#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32 0x0727
#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32_BASE_IDX 0
#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32 0x0728
#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32_BASE_IDX 0
#define V9_PIPE_PER_MEC (4)
#define V9_QUEUES_PER_PIPE_MEC (8)
@@ -167,13 +138,6 @@ static int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd,
}
static const struct kfd2kgd_calls kfd2kgd = {
.init_gtt_mem_allocation = alloc_gtt_mem,
.free_gtt_mem = free_gtt_mem,
.get_local_mem_info = get_local_mem_info,
.get_gpu_clock_counter = get_gpu_clock_counter,
.get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz,
.alloc_pasid = amdgpu_pasid_alloc,
.free_pasid = amdgpu_pasid_free,
.program_sh_mem_settings = kgd_program_sh_mem_settings,
.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
.init_interrupts = kgd_init_interrupts,
@@ -196,26 +160,9 @@ static const struct kfd2kgd_calls kfd2kgd = {
.get_fw_version = get_fw_version,
.set_scratch_backing_va = set_scratch_backing_va,
.get_tile_config = amdgpu_amdkfd_get_tile_config,
.get_cu_info = get_cu_info,
.get_vram_usage = amdgpu_amdkfd_get_vram_usage,
.create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm,
.acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm,
.destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm,
.release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm,
.get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir,
.set_vm_context_page_table_base = set_vm_context_page_table_base,
.alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu,
.free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu,
.map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu,
.unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu,
.sync_memory = amdgpu_amdkfd_gpuvm_sync_memory,
.map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel,
.restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos,
.invalidate_tlbs = invalidate_tlbs,
.invalidate_tlbs_vmid = invalidate_tlbs_vmid,
.submit_ib = amdgpu_amdkfd_submit_ib,
.gpu_recover = amdgpu_amdkfd_gpu_reset,
.set_compute_idle = amdgpu_amdkfd_set_compute_idle,
.get_hive_id = amdgpu_amdkfd_get_hive_id,
};
@@ -785,15 +732,6 @@ static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd,
static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
{
struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
uint32_t req = (1 << vmid) |
(0 << VM_INVALIDATE_ENG16_REQ__FLUSH_TYPE__SHIFT) | /* legacy */
VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PTES_MASK |
VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE0_MASK |
VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE1_MASK |
VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE2_MASK |
VM_INVALIDATE_ENG16_REQ__INVALIDATE_L1_PTES_MASK;
mutex_lock(&adev->srbm_mutex);
/* Use legacy mode tlb invalidation.
*
@@ -810,34 +748,7 @@ static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
* TODO 2: support range-based invalidation, requires kfg2kgd
* interface change
*/
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_LO32),
0xffffffff);
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_HI32),
0x0000001f);
WREG32(SOC15_REG_OFFSET(MMHUB, 0,
mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32),
0xffffffff);
WREG32(SOC15_REG_OFFSET(MMHUB, 0,
mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32),
0x0000001f);
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_REQ), req);
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_INVALIDATE_ENG16_REQ),
req);
while (!(RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ACK)) &
(1 << vmid)))
cpu_relax();
while (!(RREG32(SOC15_REG_OFFSET(MMHUB, 0,
mmMMHUB_VM_INVALIDATE_ENG16_ACK)) &
(1 << vmid)))
cpu_relax();
mutex_unlock(&adev->srbm_mutex);
amdgpu_gmc_flush_gpu_tlb(adev, vmid, 0);
}
static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid)
@@ -876,7 +787,7 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid)
if (adev->in_gpu_reset)
return -EIO;
if (ring->ready)
if (ring->sched.ready)
return invalidate_tlbs_with_kiq(adev, pasid);
for (vmid = 0; vmid < 16; vmid++) {
@@ -1016,7 +927,6 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
uint64_t page_table_base)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
uint64_t base = page_table_base | AMDGPU_PTE_VALID;
if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) {
pr_err("trying to set page table base for wrong VMID %u\n",
@@ -1028,25 +938,7 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid,
* now, all processes share the same address space size, like
* on GFX8 and older.
*/
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0);
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0);
mmhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base);
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2),
lower_32_bits(adev->vm_manager.max_pfn - 1));
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2),
upper_32_bits(adev->vm_manager.max_pfn - 1));
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base));
WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base));
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0);
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0);
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2),
lower_32_bits(adev->vm_manager.max_pfn - 1));
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2),
upper_32_bits(adev->vm_manager.max_pfn - 1));
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base));
WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base));
gfxhub_v1_0_setup_vm_pt_regs(adev, vmid, page_table_base);
}

33
drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
View File

@@ -330,7 +330,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
case CHIP_TOPAZ:
if (((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x81)) ||
((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x83)) ||
((adev->pdev->device == 0x6907) && (adev->pdev->revision == 0x87))) {
((adev->pdev->device == 0x6907) && (adev->pdev->revision == 0x87)) ||
((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD1)) ||
((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD3))) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/topaz_k_smc.bin");
} else
@@ -351,7 +353,6 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
if (type == CGS_UCODE_ID_SMU) {
if (((adev->pdev->device == 0x67ef) &&
((adev->pdev->revision == 0xe0) ||
(adev->pdev->revision == 0xe2) ||
(adev->pdev->revision == 0xe5))) ||
((adev->pdev->device == 0x67ff) &&
((adev->pdev->revision == 0xcf) ||
@@ -359,8 +360,13 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
(adev->pdev->revision == 0xff)))) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/polaris11_k_smc.bin");
} else
} else if ((adev->pdev->device == 0x67ef) &&
(adev->pdev->revision == 0xe2)) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/polaris11_k2_smc.bin");
} else {
strcpy(fw_name, "amdgpu/polaris11_smc.bin");
}
} else if (type == CGS_UCODE_ID_SMU_SK) {
strcpy(fw_name, "amdgpu/polaris11_smc_sk.bin");
}
@@ -378,14 +384,31 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
(adev->pdev->revision == 0xef))) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/polaris10_k_smc.bin");
} else
} else if ((adev->pdev->device == 0x67df) &&
((adev->pdev->revision == 0xe1) ||
(adev->pdev->revision == 0xf7))) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/polaris10_k2_smc.bin");
} else {
strcpy(fw_name, "amdgpu/polaris10_smc.bin");
}
} else if (type == CGS_UCODE_ID_SMU_SK) {
strcpy(fw_name, "amdgpu/polaris10_smc_sk.bin");
}
break;
case CHIP_POLARIS12:
strcpy(fw_name, "amdgpu/polaris12_smc.bin");
if (((adev->pdev->device == 0x6987) &&
((adev->pdev->revision == 0xc0) ||
(adev->pdev->revision == 0xc3))) ||
((adev->pdev->device == 0x6981) &&
((adev->pdev->revision == 0x00) ||
(adev->pdev->revision == 0x01) ||
(adev->pdev->revision == 0x10)))) {
info->is_kicker = true;
strcpy(fw_name, "amdgpu/polaris12_k_smc.bin");
} else {
strcpy(fw_name, "amdgpu/polaris12_smc.bin");
}
break;
case CHIP_VEGAM:
strcpy(fw_name, "amdgpu/vegam_smc.bin");

5
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
View File

@@ -1260,8 +1260,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
return 0;
error_abort:
dma_fence_put(&job->base.s_fence->finished);
job->base.s_fence = NULL;
drm_sched_job_cleanup(&job->base);
amdgpu_mn_unlock(p->mn);
error_unlock:
@@ -1285,7 +1284,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
r = amdgpu_cs_parser_init(&parser, data);
if (r) {
DRM_ERROR("Failed to initialize parser !\n");
DRM_ERROR("Failed to initialize parser %d!\n", r);
goto out;
}

117
drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c Normal file
View File

@@ -0,0 +1,117 @@
/*
* Copyright 2016 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
* * Author: Monk.liu@amd.com
*/
#include "amdgpu.h"
uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev)
{
uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
addr -= AMDGPU_VA_RESERVED_SIZE;
addr = amdgpu_gmc_sign_extend(addr);
return addr;
}
int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo,
u32 domain, uint32_t size)
{
int r;
void *ptr;
r = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
domain, bo,
NULL, &ptr);
if (!bo)
return -ENOMEM;
memset(ptr, 0, size);
return 0;
}
void amdgpu_free_static_csa(struct amdgpu_bo **bo)
{
amdgpu_bo_free_kernel(bo, NULL, NULL);
}
/*
* amdgpu_map_static_csa should be called during amdgpu_vm_init
* it maps virtual address amdgpu_csa_vaddr() to this VM, and each command
* submission of GFX should use this virtual address within META_DATA init
* package to support SRIOV gfx preemption.
*/
int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
uint64_t csa_addr, uint32_t size)
{
struct ww_acquire_ctx ticket;
struct list_head list;
struct amdgpu_bo_list_entry pd;
struct ttm_validate_buffer csa_tv;
int r;
INIT_LIST_HEAD(&list);
INIT_LIST_HEAD(&csa_tv.head);
csa_tv.bo = &bo->tbo;
csa_tv.shared = true;
list_add(&csa_tv.head, &list);
amdgpu_vm_get_pd_bo(vm, &list, &pd);
r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
if (r) {
DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
return r;
}
*bo_va = amdgpu_vm_bo_add(adev, vm, bo);
if (!*bo_va) {
ttm_eu_backoff_reservation(&ticket, &list);
DRM_ERROR("failed to create bo_va for static CSA\n");
return -ENOMEM;
}
r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, csa_addr,
size);
if (r) {
DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r);
amdgpu_vm_bo_rmv(adev, *bo_va);
ttm_eu_backoff_reservation(&ticket, &list);
return r;
}
r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, size,
AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
AMDGPU_PTE_EXECUTABLE);
if (r) {
DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r);
amdgpu_vm_bo_rmv(adev, *bo_va);
ttm_eu_backoff_reservation(&ticket, &list);
return r;
}
ttm_eu_backoff_reservation(&ticket, &list);
return 0;
}

39
drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h Normal file
View File

@@ -0,0 +1,39 @@
/*
* Copyright 2016 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
* Author: Monk.liu@amd.com
*/
#ifndef AMDGPU_CSA_MANAGER_H
#define AMDGPU_CSA_MANAGER_H
#define AMDGPU_CSA_SIZE (128 * 1024)
uint32_t amdgpu_get_total_csa_size(struct amdgpu_device *adev);
uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev);
int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo,
u32 domain, uint32_t size);
int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
uint64_t csa_addr, uint32_t size);
void amdgpu_free_static_csa(struct amdgpu_bo **bo);
#endif

39
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
View File

@@ -1656,7 +1656,9 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
/* right after GMC hw init, we create CSA */
if (amdgpu_sriov_vf(adev)) {
r = amdgpu_allocate_static_csa(adev);
r = amdgpu_allocate_static_csa(adev, &adev->virt.csa_obj,
AMDGPU_GEM_DOMAIN_VRAM,
AMDGPU_CSA_SIZE);
if (r) {
DRM_ERROR("allocate CSA failed %d\n", r);
return r;
@@ -1681,7 +1683,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
if (r)
return r;
amdgpu_xgmi_add_device(adev);
if (adev->gmc.xgmi.num_physical_nodes > 1)
amdgpu_xgmi_add_device(adev);
amdgpu_amdkfd_device_init(adev);
if (amdgpu_sriov_vf(adev))
@@ -1890,7 +1893,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
amdgpu_ucode_free_bo(adev);
amdgpu_free_static_csa(adev);
amdgpu_free_static_csa(&adev->virt.csa_obj);
amdgpu_device_wb_fini(adev);
amdgpu_device_vram_scratch_fini(adev);
}
@@ -3295,13 +3298,35 @@ bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev)
return false;
}
if (amdgpu_gpu_recovery == 0 || (amdgpu_gpu_recovery == -1 &&
!amdgpu_sriov_vf(adev))) {
DRM_INFO("GPU recovery disabled.\n");
return false;
if (amdgpu_gpu_recovery == 0)
goto disabled;
if (amdgpu_sriov_vf(adev))
return true;
if (amdgpu_gpu_recovery == -1) {
switch (adev->asic_type) {
case CHIP_TOPAZ:
case CHIP_TONGA:
case CHIP_FIJI:
case CHIP_POLARIS10:
case CHIP_POLARIS11:
case CHIP_POLARIS12:
case CHIP_VEGAM:
case CHIP_VEGA20:
case CHIP_VEGA10:
case CHIP_VEGA12:
break;
default:
goto disabled;
}
}
return true;
disabled:
DRM_INFO("GPU recovery disabled.\n");
return false;
}
/**

6
drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
View File

@@ -398,9 +398,9 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
ring->fence_drv.irq_type = irq_type;
ring->fence_drv.initialized = true;
dev_dbg(adev->dev, "fence driver on ring %d use gpu addr 0x%016llx, "
"cpu addr 0x%p\n", ring->idx,
ring->fence_drv.gpu_addr, ring->fence_drv.cpu_addr);
DRM_DEV_DEBUG(adev->dev, "fence driver on ring %s use gpu addr "
"0x%016llx, cpu addr 0x%p\n", ring->name,
ring->fence_drv.gpu_addr, ring->fence_drv.cpu_addr);
return 0;
}

6
drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
View File

@@ -248,7 +248,7 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
}
mb();
amdgpu_asic_flush_hdp(adev, NULL);
amdgpu_gmc_flush_gpu_tlb(adev, 0);
amdgpu_gmc_flush_gpu_tlb(adev, 0, 0);
return 0;
}
@@ -259,6 +259,8 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
* @offset: offset into the GPU's gart aperture
* @pages: number of pages to bind
* @dma_addr: DMA addresses of pages
* @flags: page table entry flags
* @dst: CPU address of the gart table
*
* Map the dma_addresses into GART entries (all asics).
* Returns 0 for success, -EINVAL for failure.
@@ -331,7 +333,7 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
mb();
amdgpu_asic_flush_hdp(adev, NULL);
amdgpu_gmc_flush_gpu_tlb(adev, 0);
amdgpu_gmc_flush_gpu_tlb(adev, 0, 0);
return 0;
}

1
drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
View File

@@ -41,6 +41,7 @@ struct amdgpu_bo;
struct amdgpu_gart {
struct amdgpu_bo *bo;
/* CPU kmapped address of gart table */
void *ptr;
unsigned num_gpu_pages;
unsigned num_cpu_pages;

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

@@ -25,6 +25,7 @@
#include <drm/drmP.h>
#include "amdgpu.h"
#include "amdgpu_gfx.h"
#include "amdgpu_rlc.h"
/* delay 0.1 second to enable gfx off feature */
#define GFX_OFF_DELAY_ENABLE msecs_to_jiffies(100)

54
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
View File

@@ -29,6 +29,7 @@
*/
#include "clearstate_defs.h"
#include "amdgpu_ring.h"
#include "amdgpu_rlc.h"
/* GFX current status */
#define AMDGPU_GFX_NORMAL_MODE 0x00000000L
@@ -37,59 +38,6 @@
#define AMDGPU_GFX_CG_DISABLED_MODE 0x00000004L
#define AMDGPU_GFX_LBPW_DISABLED_MODE 0x00000008L
struct amdgpu_rlc_funcs {
void (*enter_safe_mode)(struct amdgpu_device *adev);
void (*exit_safe_mode)(struct amdgpu_device *adev);
};
struct amdgpu_rlc {
/* for power gating */
struct amdgpu_bo *save_restore_obj;
uint64_t save_restore_gpu_addr;
volatile uint32_t *sr_ptr;
const u32 *reg_list;
u32 reg_list_size;
/* for clear state */
struct amdgpu_bo *clear_state_obj;
uint64_t clear_state_gpu_addr;
volatile uint32_t *cs_ptr;
const struct cs_section_def *cs_data;
u32 clear_state_size;
/* for cp tables */
struct amdgpu_bo *cp_table_obj;
uint64_t cp_table_gpu_addr;
volatile uint32_t *cp_table_ptr;
u32 cp_table_size;
/* safe mode for updating CG/PG state */
bool in_safe_mode;
const struct amdgpu_rlc_funcs *funcs;
/* for firmware data */
u32 save_and_restore_offset;
u32 clear_state_descriptor_offset;
u32 avail_scratch_ram_locations;
u32 reg_restore_list_size;
u32 reg_list_format_start;
u32 reg_list_format_separate_start;
u32 starting_offsets_start;
u32 reg_list_format_size_bytes;
u32 reg_list_size_bytes;
u32 reg_list_format_direct_reg_list_length;
u32 save_restore_list_cntl_size_bytes;
u32 save_restore_list_gpm_size_bytes;
u32 save_restore_list_srm_size_bytes;
u32 *register_list_format;
u32 *register_restore;
u8 *save_restore_list_cntl;
u8 *save_restore_list_gpm;
u8 *save_restore_list_srm;
bool is_rlc_v2_1;
};
#define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES
struct amdgpu_mec {

6
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
View File

@@ -64,7 +64,7 @@ struct amdgpu_vmhub {
struct amdgpu_gmc_funcs {
/* flush the vm tlb via mmio */
void (*flush_gpu_tlb)(struct amdgpu_device *adev,
uint32_t vmid);
uint32_t vmid, uint32_t flush_type);
/* flush the vm tlb via ring */
uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid,
uint64_t pd_addr);
@@ -89,7 +89,7 @@ struct amdgpu_gmc_funcs {
struct amdgpu_xgmi {
/* from psp */
u64 device_id;
u64 node_id;
u64 hive_id;
/* fixed per family */
u64 node_segment_size;
@@ -151,7 +151,7 @@ struct amdgpu_gmc {
struct amdgpu_xgmi xgmi;
};
#define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid))
#define amdgpu_gmc_flush_gpu_tlb(adev, vmid, type) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (type))
#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr))
#define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid))
#define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))

45
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
View File

@@ -146,7 +146,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
fence_ctx = 0;
}
if (!ring->ready) {
if (!ring->sched.ready) {
dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name);
return -EINVAL;
}
@@ -221,8 +221,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
!amdgpu_sriov_vf(adev)) /* for SRIOV preemption, Preamble CE ib must be inserted anyway */
continue;
amdgpu_ring_emit_ib(ring, ib, job ? job->vmid : 0,
need_ctx_switch);
amdgpu_ring_emit_ib(ring, job, ib, need_ctx_switch);
need_ctx_switch = false;
}
@@ -347,19 +346,14 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
tmo_gfx = 8 * AMDGPU_IB_TEST_TIMEOUT;
}
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
for (i = 0; i < adev->num_rings; ++i) {
struct amdgpu_ring *ring = adev->rings[i];
long tmo;
if (!ring || !ring->ready)
continue;
/* skip IB tests for KIQ in general for the below reasons:
* 1. We never submit IBs to the KIQ
* 2. KIQ doesn't use the EOP interrupts,
* we use some other CP interrupt.
/* KIQ rings don't have an IB test because we never submit IBs
* to them and they have no interrupt support.
*/
if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
if (!ring->sched.ready || !ring->funcs->test_ib)
continue;
/* MM engine need more time */
@@ -374,20 +368,23 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
tmo = tmo_gfx;
r = amdgpu_ring_test_ib(ring, tmo);
if (r) {
ring->ready = false;
if (!r) {
DRM_DEV_DEBUG(adev->dev, "ib test on %s succeeded\n",
ring->name);
continue;
}
if (ring == &adev->gfx.gfx_ring[0]) {
/* oh, oh, that's really bad */
DRM_ERROR("amdgpu: failed testing IB on GFX ring (%d).\n", r);
adev->accel_working = false;
return r;
ring->sched.ready = false;
DRM_DEV_ERROR(adev->dev, "IB test failed on %s (%d).\n",
ring->name, r);
} else {
/* still not good, but we can live with it */
DRM_ERROR("amdgpu: failed testing IB on ring %d (%d).\n", i, r);
ret = r;
}
if (ring == &adev->gfx.gfx_ring[0]) {
/* oh, oh, that's really bad */
adev->accel_working = false;
return r;
} else {
ret = r;
}
}
return ret;

21
drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
View File

@@ -93,23 +93,6 @@ static void amdgpu_hotplug_work_func(struct work_struct *work)
drm_helper_hpd_irq_event(dev);
}
/**
* amdgpu_irq_reset_work_func - execute GPU reset
*
* @work: work struct pointer
*
* Execute scheduled GPU reset (Cayman+).
* This function is called when the IRQ handler thinks we need a GPU reset.
*/
static void amdgpu_irq_reset_work_func(struct work_struct *work)
{
struct amdgpu_device *adev = container_of(work, struct amdgpu_device,
reset_work);
if (!amdgpu_sriov_vf(adev) && amdgpu_device_should_recover_gpu(adev))
amdgpu_device_gpu_recover(adev, NULL);
}
/**
* amdgpu_irq_disable_all - disable *all* interrupts
*
@@ -262,15 +245,12 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
amdgpu_hotplug_work_func);
}
INIT_WORK(&adev->reset_work, amdgpu_irq_reset_work_func);
adev->irq.installed = true;
r = drm_irq_install(adev->ddev, adev->ddev->pdev->irq);
if (r) {
adev->irq.installed = false;
if (!amdgpu_device_has_dc_support(adev))
flush_work(&adev->hotplug_work);
cancel_work_sync(&adev->reset_work);
return r;
}
adev->ddev->max_vblank_count = 0x00ffffff;
@@ -299,7 +279,6 @@ void amdgpu_irq_fini(struct amdgpu_device *adev)
pci_disable_msi(adev->pdev);
if (!amdgpu_device_has_dc_support(adev))
flush_work(&adev->hotplug_work);
cancel_work_sync(&adev->reset_work);
}
for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {

2
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
View File

@@ -112,6 +112,8 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
struct amdgpu_job *job = to_amdgpu_job(s_job);
drm_sched_job_cleanup(s_job);
amdgpu_ring_priority_put(ring, s_job->s_priority);
dma_fence_put(job->fence);
amdgpu_sync_free(&job->sync);

2
drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
View File

@@ -33,6 +33,8 @@
#define to_amdgpu_job(sched_job) \
container_of((sched_job), struct amdgpu_job, base)
#define AMDGPU_JOB_GET_VMID(job) ((job) ? (job)->vmid : 0)
struct amdgpu_fence;
struct amdgpu_job {

25
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
View File

@@ -336,7 +336,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
case AMDGPU_HW_IP_GFX:
type = AMD_IP_BLOCK_TYPE_GFX;
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
if (adev->gfx.gfx_ring[i].ready)
if (adev->gfx.gfx_ring[i].sched.ready)
++num_rings;
ib_start_alignment = 32;
ib_size_alignment = 32;
@@ -344,7 +344,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
case AMDGPU_HW_IP_COMPUTE:
type = AMD_IP_BLOCK_TYPE_GFX;
for (i = 0; i < adev->gfx.num_compute_rings; i++)
if (adev->gfx.compute_ring[i].ready)
if (adev->gfx.compute_ring[i].sched.ready)
++num_rings;
ib_start_alignment = 32;
ib_size_alignment = 32;
@@ -352,7 +352,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
case AMDGPU_HW_IP_DMA:
type = AMD_IP_BLOCK_TYPE_SDMA;
for (i = 0; i < adev->sdma.num_instances; i++)
if (adev->sdma.instance[i].ring.ready)
if (adev->sdma.instance[i].ring.sched.ready)
++num_rings;
ib_start_alignment = 256;
ib_size_alignment = 4;
@@ -363,7 +363,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
if (adev->uvd.harvest_config & (1 << i))
continue;
if (adev->uvd.inst[i].ring.ready)
if (adev->uvd.inst[i].ring.sched.ready)
++num_rings;
}
ib_start_alignment = 64;
@@ -372,7 +372,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
case AMDGPU_HW_IP_VCE:
type = AMD_IP_BLOCK_TYPE_VCE;
for (i = 0; i < adev->vce.num_rings; i++)
if (adev->vce.ring[i].ready)
if (adev->vce.ring[i].sched.ready)
++num_rings;
ib_start_alignment = 4;
ib_size_alignment = 1;
@@ -384,7 +384,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
continue;
for (j = 0; j < adev->uvd.num_enc_rings; j++)
if (adev->uvd.inst[i].ring_enc[j].ready)
if (adev->uvd.inst[i].ring_enc[j].sched.ready)
++num_rings;
}
ib_start_alignment = 64;
@@ -392,7 +392,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
break;
case AMDGPU_HW_IP_VCN_DEC:
type = AMD_IP_BLOCK_TYPE_VCN;
if (adev->vcn.ring_dec.ready)
if (adev->vcn.ring_dec.sched.ready)
++num_rings;
ib_start_alignment = 16;
ib_size_alignment = 16;
@@ -400,14 +400,14 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
case AMDGPU_HW_IP_VCN_ENC:
type = AMD_IP_BLOCK_TYPE_VCN;
for (i = 0; i < adev->vcn.num_enc_rings; i++)
if (adev->vcn.ring_enc[i].ready)
if (adev->vcn.ring_enc[i].sched.ready)
++num_rings;
ib_start_alignment = 64;
ib_size_alignment = 1;
break;
case AMDGPU_HW_IP_VCN_JPEG:
type = AMD_IP_BLOCK_TYPE_VCN;
if (adev->vcn.ring_jpeg.ready)
if (adev->vcn.ring_jpeg.sched.ready)
++num_rings;
ib_start_alignment = 16;
ib_size_alignment = 16;
@@ -978,7 +978,10 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
}
if (amdgpu_sriov_vf(adev)) {
r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va);
uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
r = amdgpu_map_static_csa(adev, &fpriv->vm, adev->virt.csa_obj,
&fpriv->csa_va, csa_addr, AMDGPU_CSA_SIZE);
if (r)
goto error_vm;
}
@@ -1048,8 +1051,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
pasid = fpriv->vm.pasid;
pd = amdgpu_bo_ref(fpriv->vm.root.base.bo);
amdgpu_vm_fini(adev, &fpriv->vm);
amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr);
amdgpu_vm_fini(adev, &fpriv->vm);
if (pasid)
amdgpu_pasid_free_delayed(pd->tbo.resv, pasid);

8
drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
View File

@@ -57,7 +57,6 @@ struct amdgpu_hpd;
#define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base)
#define to_amdgpu_encoder(x) container_of(x, struct amdgpu_encoder, base)
#define to_amdgpu_framebuffer(x) container_of(x, struct amdgpu_framebuffer, base)
#define to_amdgpu_plane(x) container_of(x, struct amdgpu_plane, base)
#define to_dm_plane_state(x) container_of(x, struct dm_plane_state, base);
@@ -325,7 +324,7 @@ struct amdgpu_mode_info {
struct card_info *atom_card_info;
bool mode_config_initialized;
struct amdgpu_crtc *crtcs[AMDGPU_MAX_CRTCS];
struct amdgpu_plane *planes[AMDGPU_MAX_PLANES];
struct drm_plane *planes[AMDGPU_MAX_PLANES];
struct amdgpu_afmt *afmt[AMDGPU_MAX_AFMT_BLOCKS];
/* DVI-I properties */
struct drm_property *coherent_mode_property;
@@ -434,11 +433,6 @@ struct amdgpu_crtc {
struct drm_pending_vblank_event *event;
};
struct amdgpu_plane {
struct drm_plane base;
enum drm_plane_type plane_type;
};
struct amdgpu_encoder_atom_dig {
bool linkb;
/* atom dig */

2
drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
View File

@@ -2129,7 +2129,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
struct amdgpu_ring *ring = adev->rings[i];
if (ring && ring->ready)
if (ring && ring->sched.ready)
amdgpu_fence_wait_empty(ring);
}

225
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
View File

@@ -90,6 +90,8 @@ static int psp_sw_fini(void *handle)
adev->psp.sos_fw = NULL;
release_firmware(adev->psp.asd_fw);
adev->psp.asd_fw = NULL;
release_firmware(adev->psp.ta_fw);
adev->psp.ta_fw = NULL;
return 0;
}
@@ -118,22 +120,26 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
static int
psp_cmd_submit_buf(struct psp_context *psp,
struct amdgpu_firmware_info *ucode,
struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr,
int index)
struct psp_gfx_cmd_resp *cmd, uint64_t fence_mc_addr)
{
int ret;
int index;
memset(psp->cmd_buf_mem, 0, PSP_CMD_BUFFER_SIZE);
memcpy(psp->cmd_buf_mem, cmd, sizeof(struct psp_gfx_cmd_resp));
index = atomic_inc_return(&psp->fence_value);
ret = psp_cmd_submit(psp, ucode, psp->cmd_buf_mc_addr,
fence_mc_addr, index);
while (*((unsigned int *)psp->fence_buf) != index) {
msleep(1);
if (ret) {
atomic_dec(&psp->fence_value);
return ret;
}
while (*((unsigned int *)psp->fence_buf) != index)
msleep(1);
/* the status field must be 0 after FW is loaded */
if (ucode && psp->cmd_buf_mem->resp.status) {
DRM_ERROR("failed loading with status (%d) and ucode id (%d)\n",
@@ -191,7 +197,7 @@ static int psp_tmr_load(struct psp_context *psp)
PSP_TMR_SIZE, psp->tmr_mc_addr);
ret = psp_cmd_submit_buf(psp, NULL, cmd,
psp->fence_buf_mc_addr, 1);
psp->fence_buf_mc_addr);
if (ret)
goto failed;
@@ -258,13 +264,194 @@ static int psp_asd_load(struct psp_context *psp)
psp->asd_ucode_size, PSP_ASD_SHARED_MEM_SIZE);
ret = psp_cmd_submit_buf(psp, NULL, cmd,
psp->fence_buf_mc_addr, 2);
psp->fence_buf_mc_addr);
kfree(cmd);
return ret;
}
static void psp_prep_xgmi_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd,
uint64_t xgmi_ta_mc, uint64_t xgmi_mc_shared,
uint32_t xgmi_ta_size, uint32_t shared_size)
{
cmd->cmd_id = GFX_CMD_ID_LOAD_TA;
cmd->cmd.cmd_load_ta.app_phy_addr_lo = lower_32_bits(xgmi_ta_mc);
cmd->cmd.cmd_load_ta.app_phy_addr_hi = upper_32_bits(xgmi_ta_mc);
cmd->cmd.cmd_load_ta.app_len = xgmi_ta_size;
cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_lo = lower_32_bits(xgmi_mc_shared);
cmd->cmd.cmd_load_ta.cmd_buf_phy_addr_hi = upper_32_bits(xgmi_mc_shared);
cmd->cmd.cmd_load_ta.cmd_buf_len = shared_size;
}
static int psp_xgmi_init_shared_buf(struct psp_context *psp)
{
int ret;
/*
* Allocate 16k memory aligned to 4k from Frame Buffer (local
* physical) for xgmi ta <-> Driver
*/
ret = amdgpu_bo_create_kernel(psp->adev, PSP_XGMI_SHARED_MEM_SIZE,
PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
&psp->xgmi_context.xgmi_shared_bo,
&psp->xgmi_context.xgmi_shared_mc_addr,
&psp->xgmi_context.xgmi_shared_buf);
return ret;
}
static int psp_xgmi_load(struct psp_context *psp)
{
int ret;
struct psp_gfx_cmd_resp *cmd;
/*
* TODO: bypass the loading in sriov for now
*/
if (amdgpu_sriov_vf(psp->adev))
return 0;
cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
if (!cmd)
return -ENOMEM;
memset(psp->fw_pri_buf, 0, PSP_1_MEG);
memcpy(psp->fw_pri_buf, psp->ta_xgmi_start_addr, psp->ta_xgmi_ucode_size);
psp_prep_xgmi_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr,
psp->xgmi_context.xgmi_shared_mc_addr,
psp->ta_xgmi_ucode_size, PSP_XGMI_SHARED_MEM_SIZE);
ret = psp_cmd_submit_buf(psp, NULL, cmd,
psp->fence_buf_mc_addr);
if (!ret) {
psp->xgmi_context.initialized = 1;
psp->xgmi_context.session_id = cmd->resp.session_id;
}
kfree(cmd);
return ret;
}
static void psp_prep_xgmi_ta_unload_cmd_buf(struct psp_gfx_cmd_resp *cmd,
uint32_t xgmi_session_id)
{
cmd->cmd_id = GFX_CMD_ID_UNLOAD_TA;
cmd->cmd.cmd_unload_ta.session_id = xgmi_session_id;
}
static int psp_xgmi_unload(struct psp_context *psp)
{
int ret;
struct psp_gfx_cmd_resp *cmd;
/*
* TODO: bypass the unloading in sriov for now
*/
if (amdgpu_sriov_vf(psp->adev))
return 0;
cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
if (!cmd)
return -ENOMEM;
psp_prep_xgmi_ta_unload_cmd_buf(cmd, psp->xgmi_context.session_id);
ret = psp_cmd_submit_buf(psp, NULL, cmd,
psp->fence_buf_mc_addr);
kfree(cmd);
return ret;
}
static void psp_prep_xgmi_ta_invoke_cmd_buf(struct psp_gfx_cmd_resp *cmd,
uint32_t ta_cmd_id,
uint32_t xgmi_session_id)
{
cmd->cmd_id = GFX_CMD_ID_INVOKE_CMD;
cmd->cmd.cmd_invoke_cmd.session_id = xgmi_session_id;
cmd->cmd.cmd_invoke_cmd.ta_cmd_id = ta_cmd_id;
/* Note: cmd_invoke_cmd.buf is not used for now */
}
int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
{
int ret;
struct psp_gfx_cmd_resp *cmd;
/*
* TODO: bypass the loading in sriov for now
*/
if (amdgpu_sriov_vf(psp->adev))
return 0;
cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
if (!cmd)
return -ENOMEM;
psp_prep_xgmi_ta_invoke_cmd_buf(cmd, ta_cmd_id,
psp->xgmi_context.session_id);
ret = psp_cmd_submit_buf(psp, NULL, cmd,
psp->fence_buf_mc_addr);
kfree(cmd);
return ret;
}
static int psp_xgmi_terminate(struct psp_context *psp)
{
int ret;
if (!psp->xgmi_context.initialized)
return 0;
ret = psp_xgmi_unload(psp);
if (ret)
return ret;
psp->xgmi_context.initialized = 0;
/* free xgmi shared memory */
amdgpu_bo_free_kernel(&psp->xgmi_context.xgmi_shared_bo,
&psp->xgmi_context.xgmi_shared_mc_addr,
&psp->xgmi_context.xgmi_shared_buf);
return 0;
}
static int psp_xgmi_initialize(struct psp_context *psp)
{
struct ta_xgmi_shared_memory *xgmi_cmd;
int ret;
if (!psp->xgmi_context.initialized) {
ret = psp_xgmi_init_shared_buf(psp);
if (ret)
return ret;
}
/* Load XGMI TA */
ret = psp_xgmi_load(psp);
if (ret)
return ret;
/* Initialize XGMI session */
xgmi_cmd = (struct ta_xgmi_shared_memory *)(psp->xgmi_context.xgmi_shared_buf);
memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
xgmi_cmd->cmd_id = TA_COMMAND_XGMI__INITIALIZE;
ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
return ret;
}
static int psp_hw_start(struct psp_context *psp)
{
struct amdgpu_device *adev = psp->adev;
@@ -292,6 +479,15 @@ static int psp_hw_start(struct psp_context *psp)
if (ret)
return ret;
if (adev->gmc.xgmi.num_physical_nodes > 1) {
ret = psp_xgmi_initialize(psp);
/* Warning the XGMI seesion initialize failure
* Instead of stop driver initialization
*/
if (ret)
dev_err(psp->adev->dev,
"XGMI: Failed to initialize XGMI session\n");
}
return 0;
}
@@ -321,7 +517,7 @@ static int psp_np_fw_load(struct psp_context *psp)
return ret;
ret = psp_cmd_submit_buf(psp, ucode, psp->cmd,
psp->fence_buf_mc_addr, i + 3);
psp->fence_buf_mc_addr);
if (ret)
return ret;
@@ -452,6 +648,10 @@ static int psp_hw_fini(void *handle)
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
return 0;
if (adev->gmc.xgmi.num_physical_nodes > 1 &&
psp->xgmi_context.initialized == 1)
psp_xgmi_terminate(psp);
psp_ring_destroy(psp, PSP_RING_TYPE__KM);
amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);
@@ -479,6 +679,15 @@ static int psp_suspend(void *handle)
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
return 0;
if (adev->gmc.xgmi.num_physical_nodes > 1 &&
psp->xgmi_context.initialized == 1) {
ret = psp_xgmi_terminate(psp);
if (ret) {
DRM_ERROR("Failed to terminate xgmi ta\n");
return ret;
}
}
ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
if (ret) {
DRM_ERROR("PSP ring stop failed\n");

61
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
View File

@@ -27,14 +27,17 @@
#include "amdgpu.h"
#include "psp_gfx_if.h"
#include "ta_xgmi_if.h"
#define PSP_FENCE_BUFFER_SIZE 0x1000
#define PSP_CMD_BUFFER_SIZE 0x1000
#define PSP_ASD_SHARED_MEM_SIZE 0x4000
#define PSP_ASD_SHARED_MEM_SIZE 0x4000
#define PSP_XGMI_SHARED_MEM_SIZE 0x4000
#define PSP_1_MEG 0x100000
#define PSP_TMR_SIZE 0x400000
struct psp_context;
struct psp_xgmi_node_info;
struct psp_xgmi_topology_info;
enum psp_ring_type
@@ -80,12 +83,20 @@ struct psp_funcs
enum AMDGPU_UCODE_ID ucode_type);
bool (*smu_reload_quirk)(struct psp_context *psp);
int (*mode1_reset)(struct psp_context *psp);
uint64_t (*xgmi_get_device_id)(struct psp_context *psp);
uint64_t (*xgmi_get_node_id)(struct psp_context *psp);
uint64_t (*xgmi_get_hive_id)(struct psp_context *psp);
int (*xgmi_get_topology_info)(struct psp_context *psp, int number_devices,
struct psp_xgmi_topology_info *topology);
struct psp_xgmi_topology_info *topology);
int (*xgmi_set_topology_info)(struct psp_context *psp, int number_devices,
struct psp_xgmi_topology_info *topology);
struct psp_xgmi_topology_info *topology);
};
struct psp_xgmi_context {
uint8_t initialized;
uint32_t session_id;
struct amdgpu_bo *xgmi_shared_bo;
uint64_t xgmi_shared_mc_addr;
void *xgmi_shared_buf;
};
struct psp_context
@@ -96,7 +107,7 @@ struct psp_context
const struct psp_funcs *funcs;
/* fence buffer */
/* firmware buffer */
struct amdgpu_bo *fw_pri_bo;
uint64_t fw_pri_mc_addr;
void *fw_pri_buf;
@@ -134,6 +145,16 @@ struct psp_context
struct amdgpu_bo *cmd_buf_bo;
uint64_t cmd_buf_mc_addr;
struct psp_gfx_cmd_resp *cmd_buf_mem;
/* fence value associated with cmd buffer */
atomic_t fence_value;
/* xgmi ta firmware and buffer */
const struct firmware *ta_fw;
uint32_t ta_xgmi_ucode_version;
uint32_t ta_xgmi_ucode_size;
uint8_t *ta_xgmi_start_addr;
struct psp_xgmi_context xgmi_context;
};
struct amdgpu_psp_funcs {
@@ -141,21 +162,17 @@ struct amdgpu_psp_funcs {
enum AMDGPU_UCODE_ID);
};
#define AMDGPU_XGMI_MAX_CONNECTED_NODES 64
struct psp_xgmi_node_info {
uint64_t node_id;
uint8_t num_hops;
uint8_t is_sharing_enabled;
enum ta_xgmi_assigned_sdma_engine sdma_engine;
};
struct psp_xgmi_topology_info {
/* Generated by PSP to identify the GPU instance within xgmi connection */
uint64_t device_id;
/*
* If all bits set to 0 , driver indicates it wants to retrieve the xgmi
* connection vector topology, but not access enable the connections
* if some or all bits are set to 1, driver indicates it want to retrieve the
* current xgmi topology and access enable the link to GPU[i] associated
* with the bit position in the vector.
* On return,: bits indicated which xgmi links are present/active depending
* on the value passed in. The relative bit offset for the relative GPU index
* within the hive is always marked active.
*/
uint32_t connection_mask;
uint32_t reserved; /* must be 0 */
uint32_t num_nodes;
struct psp_xgmi_node_info nodes[AMDGPU_XGMI_MAX_CONNECTED_NODES];
};
#define psp_prep_cmd_buf(ucode, type) (psp)->funcs->prep_cmd_buf((ucode), (type))
@@ -177,8 +194,8 @@ struct psp_xgmi_topology_info {
((psp)->funcs->smu_reload_quirk ? (psp)->funcs->smu_reload_quirk((psp)) : false)
#define psp_mode1_reset(psp) \
((psp)->funcs->mode1_reset ? (psp)->funcs->mode1_reset((psp)) : false)
#define psp_xgmi_get_device_id(psp) \
((psp)->funcs->xgmi_get_device_id ? (psp)->funcs->xgmi_get_device_id((psp)) : 0)
#define psp_xgmi_get_node_id(psp) \
((psp)->funcs->xgmi_get_node_id ? (psp)->funcs->xgmi_get_node_id((psp)) : 0)
#define psp_xgmi_get_hive_id(psp) \
((psp)->funcs->xgmi_get_hive_id ? (psp)->funcs->xgmi_get_hive_id((psp)) : 0)
#define psp_xgmi_get_topology_info(psp, num_device, topology) \
@@ -199,6 +216,8 @@ extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
extern const struct amdgpu_ip_block_version psp_v10_0_ip_block;
int psp_gpu_reset(struct amdgpu_device *adev);
int psp_xgmi_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
extern const struct amdgpu_ip_block_version psp_v11_0_ip_block;
#endif

28
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
View File

@@ -338,7 +338,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
*/
void amdgpu_ring_fini(struct amdgpu_ring *ring)
{
ring->ready = false;
ring->sched.ready = false;
/* Not to finish a ring which is not initialized */
if (!(ring->adev) || !(ring->adev->rings[ring->idx]))
@@ -500,3 +500,29 @@ static void amdgpu_debugfs_ring_fini(struct amdgpu_ring *ring)
debugfs_remove(ring->ent);
#endif
}
/**
* amdgpu_ring_test_helper - tests ring and set sched readiness status
*
* @ring: ring to try the recovery on
*
* Tests ring and set sched readiness status
*
* Returns 0 on success, error on failure.
*/
int amdgpu_ring_test_helper(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
int r;
r = amdgpu_ring_test_ring(ring);
if (r)
DRM_DEV_ERROR(adev->dev, "ring %s test failed (%d)\n",
ring->name, r);
else
DRM_DEV_DEBUG(adev->dev, "ring test on %s succeeded\n",
ring->name);
ring->sched.ready = !r;
return r;
}

8
drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
View File

@@ -129,8 +129,9 @@ struct amdgpu_ring_funcs {
unsigned emit_ib_size;
/* command emit functions */
void (*emit_ib)(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch);
bool ctx_switch);
void (*emit_fence)(struct amdgpu_ring *ring, uint64_t addr,
uint64_t seq, unsigned flags);
void (*emit_pipeline_sync)(struct amdgpu_ring *ring);
@@ -189,7 +190,6 @@ struct amdgpu_ring {
uint64_t gpu_addr;
uint64_t ptr_mask;
uint32_t buf_mask;
bool ready;
u32 idx;
u32 me;
u32 pipe;
@@ -229,7 +229,7 @@ struct amdgpu_ring {
#define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
#define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
#define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))
#define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c))
#define amdgpu_ring_emit_ib(r, job, ib, c) ((r)->funcs->emit_ib((r), (job), (ib), (c)))
#define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r))
#define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
#define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))
@@ -313,4 +313,6 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
ring->count_dw -= count_dw;
}
int amdgpu_ring_test_helper(struct amdgpu_ring *ring);
#endif

282
drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c Normal file
View File

@@ -0,0 +1,282 @@
/*
* Copyright 2014 Advanced Micro Devices, Inc.
* Copyright 2008 Red Hat Inc.
* Copyright 2009 Jerome Glisse.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
#include <linux/firmware.h>
#include "amdgpu.h"
#include "amdgpu_gfx.h"
#include "amdgpu_rlc.h"
/**
* amdgpu_gfx_rlc_enter_safe_mode - Set RLC into safe mode
*
* @adev: amdgpu_device pointer
*
* Set RLC enter into safe mode if RLC is enabled and haven't in safe mode.
*/
void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
{
if (adev->gfx.rlc.in_safe_mode)
return;
/* if RLC is not enabled, do nothing */
if (!adev->gfx.rlc.funcs->is_rlc_enabled(adev))
return;
if (adev->cg_flags &
(AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
AMD_CG_SUPPORT_GFX_3D_CGCG)) {
adev->gfx.rlc.funcs->set_safe_mode(adev);
adev->gfx.rlc.in_safe_mode = true;
}
}
/**
* amdgpu_gfx_rlc_exit_safe_mode - Set RLC out of safe mode
*
* @adev: amdgpu_device pointer
*
* Set RLC exit safe mode if RLC is enabled and have entered into safe mode.
*/
void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
{
if (!(adev->gfx.rlc.in_safe_mode))
return;
/* if RLC is not enabled, do nothing */
if (!adev->gfx.rlc.funcs->is_rlc_enabled(adev))
return;
if (adev->cg_flags &
(AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
AMD_CG_SUPPORT_GFX_3D_CGCG)) {
adev->gfx.rlc.funcs->unset_safe_mode(adev);
adev->gfx.rlc.in_safe_mode = false;
}
}
/**
* amdgpu_gfx_rlc_init_sr - Init save restore block
*
* @adev: amdgpu_device pointer
* @dws: the size of save restore block
*
* Allocate and setup value to save restore block of rlc.
* Returns 0 on succeess or negative error code if allocate failed.
*/
int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws)
{
const u32 *src_ptr;
volatile u32 *dst_ptr;
u32 i;
int r;
/* allocate save restore block */
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.save_restore_obj,
&adev->gfx.rlc.save_restore_gpu_addr,
(void **)&adev->gfx.rlc.sr_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r);
amdgpu_gfx_rlc_fini(adev);
return r;
}
/* write the sr buffer */
src_ptr = adev->gfx.rlc.reg_list;
dst_ptr = adev->gfx.rlc.sr_ptr;
for (i = 0; i < adev->gfx.rlc.reg_list_size; i++)
dst_ptr[i] = cpu_to_le32(src_ptr[i]);
amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
return 0;
}
/**
* amdgpu_gfx_rlc_init_csb - Init clear state block
*
* @adev: amdgpu_device pointer
*
* Allocate and setup value to clear state block of rlc.
* Returns 0 on succeess or negative error code if allocate failed.
*/
int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev)
{
volatile u32 *dst_ptr;
u32 dws;
int r;
/* allocate clear state block */
adev->gfx.rlc.clear_state_size = dws = adev->gfx.rlc.funcs->get_csb_size(adev);
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
if (r) {
dev_err(adev->dev, "(%d) failed to create rlc csb bo\n", r);
amdgpu_gfx_rlc_fini(adev);
return r;
}
/* set up the cs buffer */
dst_ptr = adev->gfx.rlc.cs_ptr;
adev->gfx.rlc.funcs->get_csb_buffer(adev, dst_ptr);
amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
return 0;
}
/**
* amdgpu_gfx_rlc_init_cpt - Init cp table
*
* @adev: amdgpu_device pointer
*
* Allocate and setup value to cp table of rlc.
* Returns 0 on succeess or negative error code if allocate failed.
*/
int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev)
{
int r;
r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size,
PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
if (r) {
dev_err(adev->dev, "(%d) failed to create cp table bo\n", r);
amdgpu_gfx_rlc_fini(adev);
return r;
}
/* set up the cp table */
amdgpu_gfx_rlc_setup_cp_table(adev);
amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
return 0;
}
/**
* amdgpu_gfx_rlc_setup_cp_table - setup cp the buffer of cp table
*
* @adev: amdgpu_device pointer
*
* Write cp firmware data into cp table.
*/
void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev)
{
const __le32 *fw_data;
volatile u32 *dst_ptr;
int me, i, max_me;
u32 bo_offset = 0;
u32 table_offset, table_size;
max_me = adev->gfx.rlc.funcs->get_cp_table_num(adev);
/* write the cp table buffer */
dst_ptr = adev->gfx.rlc.cp_table_ptr;
for (me = 0; me < max_me; me++) {
if (me == 0) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
fw_data = (const __le32 *)
(adev->gfx.ce_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 1) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
fw_data = (const __le32 *)
(adev->gfx.pfp_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 2) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
fw_data = (const __le32 *)
(adev->gfx.me_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 3) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 4) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec2_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
}
for (i = 0; i < table_size; i ++) {
dst_ptr[bo_offset + i] =
cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
}
bo_offset += table_size;
}
}
/**
* amdgpu_gfx_rlc_fini - Free BO which used for RLC
*
* @adev: amdgpu_device pointer
*
* Free three BO which is used for rlc_save_restore_block, rlc_clear_state_block
* and rlc_jump_table_block.
*/
void amdgpu_gfx_rlc_fini(struct amdgpu_device *adev)
{
/* save restore block */
if (adev->gfx.rlc.save_restore_obj) {
amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj,
&adev->gfx.rlc.save_restore_gpu_addr,
(void **)&adev->gfx.rlc.sr_ptr);
}
/* clear state block */
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
/* jump table block */
amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
}

98
drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h Normal file
View File

@@ -0,0 +1,98 @@
/*
* Copyright 2014 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
#ifndef __AMDGPU_RLC_H__
#define __AMDGPU_RLC_H__
#include "clearstate_defs.h"
struct amdgpu_rlc_funcs {
bool (*is_rlc_enabled)(struct amdgpu_device *adev);
void (*set_safe_mode)(struct amdgpu_device *adev);
void (*unset_safe_mode)(struct amdgpu_device *adev);
int (*init)(struct amdgpu_device *adev);
u32 (*get_csb_size)(struct amdgpu_device *adev);
void (*get_csb_buffer)(struct amdgpu_device *adev, volatile u32 *buffer);
int (*get_cp_table_num)(struct amdgpu_device *adev);
int (*resume)(struct amdgpu_device *adev);
void (*stop)(struct amdgpu_device *adev);
void (*reset)(struct amdgpu_device *adev);
void (*start)(struct amdgpu_device *adev);
};
struct amdgpu_rlc {
/* for power gating */
struct amdgpu_bo *save_restore_obj;
uint64_t save_restore_gpu_addr;
volatile uint32_t *sr_ptr;
const u32 *reg_list;
u32 reg_list_size;
/* for clear state */
struct amdgpu_bo *clear_state_obj;
uint64_t clear_state_gpu_addr;
volatile uint32_t *cs_ptr;
const struct cs_section_def *cs_data;
u32 clear_state_size;
/* for cp tables */
struct amdgpu_bo *cp_table_obj;
uint64_t cp_table_gpu_addr;
volatile uint32_t *cp_table_ptr;
u32 cp_table_size;
/* safe mode for updating CG/PG state */
bool in_safe_mode;
const struct amdgpu_rlc_funcs *funcs;
/* for firmware data */
u32 save_and_restore_offset;
u32 clear_state_descriptor_offset;
u32 avail_scratch_ram_locations;
u32 reg_restore_list_size;
u32 reg_list_format_start;
u32 reg_list_format_separate_start;
u32 starting_offsets_start;
u32 reg_list_format_size_bytes;
u32 reg_list_size_bytes;
u32 reg_list_format_direct_reg_list_length;
u32 save_restore_list_cntl_size_bytes;
u32 save_restore_list_gpm_size_bytes;
u32 save_restore_list_srm_size_bytes;
u32 *register_list_format;
u32 *register_restore;
u8 *save_restore_list_cntl;
u8 *save_restore_list_gpm;
u8 *save_restore_list_srm;
bool is_rlc_v2_1;
};
void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev);
void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev);
int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws);
int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev);
int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev);
void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev);
void amdgpu_gfx_rlc_fini(struct amdgpu_device *adev);
#endif

28
drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
View File

@@ -28,17 +28,31 @@
* GPU SDMA IP block helpers function.
*/
struct amdgpu_sdma_instance * amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
struct amdgpu_sdma_instance *amdgpu_sdma_get_instance_from_ring(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
int i;
for (i = 0; i < adev->sdma.num_instances; i++)
if (&adev->sdma.instance[i].ring == ring)
break;
if (ring == &adev->sdma.instance[i].ring ||
ring == &adev->sdma.instance[i].page)
return &adev->sdma.instance[i];
if (i < AMDGPU_MAX_SDMA_INSTANCES)
return &adev->sdma.instance[i];
else
return NULL;
return NULL;
}
int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index)
{
struct amdgpu_device *adev = ring->adev;
int i;
for (i = 0; i < adev->sdma.num_instances; i++) {
if (ring == &adev->sdma.instance[i].ring ||
ring == &adev->sdma.instance[i].page) {
*index = i;
return 0;
}
}
return -EINVAL;
}

5
drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
View File

@@ -41,6 +41,7 @@ struct amdgpu_sdma_instance {
uint32_t feature_version;
struct amdgpu_ring ring;
struct amdgpu_ring page;
bool burst_nop;
};
@@ -50,6 +51,7 @@ struct amdgpu_sdma {
struct amdgpu_irq_src illegal_inst_irq;
int num_instances;
uint32_t srbm_soft_reset;
bool has_page_queue;
};
/*
@@ -92,6 +94,7 @@ struct amdgpu_buffer_funcs {
#define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b))
struct amdgpu_sdma_instance *
amdgpu_get_sdma_instance(struct amdgpu_ring *ring);
amdgpu_sdma_get_instance_from_ring(struct amdgpu_ring *ring);
int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index);
#endif

15
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
View File

@@ -218,6 +218,7 @@ TRACE_EVENT(amdgpu_vm_grab_id,
TP_ARGS(vm, ring, job),
TP_STRUCT__entry(
__field(u32, pasid)
__string(ring, ring->name)
__field(u32, ring)
__field(u32, vmid)
__field(u32, vm_hub)
@@ -227,14 +228,14 @@ TRACE_EVENT(amdgpu_vm_grab_id,
TP_fast_assign(
__entry->pasid = vm->pasid;
__entry->ring = ring->idx;
__assign_str(ring, ring->name)
__entry->vmid = job->vmid;
__entry->vm_hub = ring->funcs->vmhub,
__entry->pd_addr = job->vm_pd_addr;
__entry->needs_flush = job->vm_needs_flush;
),
TP_printk("pasid=%d, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
__entry->pasid, __entry->ring, __entry->vmid,
TP_printk("pasid=%d, ring=%s, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
__entry->pasid, __get_str(ring), __entry->vmid,
__entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
);
@@ -366,20 +367,20 @@ TRACE_EVENT(amdgpu_vm_flush,
uint64_t pd_addr),
TP_ARGS(ring, vmid, pd_addr),
TP_STRUCT__entry(
__field(u32, ring)
__string(ring, ring->name)
__field(u32, vmid)
__field(u32, vm_hub)
__field(u64, pd_addr)
),
TP_fast_assign(
__entry->ring = ring->idx;
__assign_str(ring, ring->name)
__entry->vmid = vmid;
__entry->vm_hub = ring->funcs->vmhub;
__entry->pd_addr = pd_addr;
),
TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx",
__entry->ring, __entry->vmid,
TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",
__get_str(ring), __entry->vmid,
__entry->vm_hub,__entry->pd_addr)
);

105
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
View File

@@ -61,100 +61,6 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo,
static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev);
static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev);
/*
* Global memory.
*/
/**
* amdgpu_ttm_mem_global_init - Initialize and acquire reference to
* memory object
*
* @ref: Object for initialization.
*
* This is called by drm_global_item_ref() when an object is being
* initialized.
*/
static int amdgpu_ttm_mem_global_init(struct drm_global_reference *ref)
{
return ttm_mem_global_init(ref->object);
}
/**
* amdgpu_ttm_mem_global_release - Drop reference to a memory object
*
* @ref: Object being removed
*
* This is called by drm_global_item_unref() when an object is being
* released.
*/
static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref)
{
ttm_mem_global_release(ref->object);
}
/**
* amdgpu_ttm_global_init - Initialize global TTM memory reference structures.
*
* @adev: AMDGPU device for which the global structures need to be registered.
*
* This is called as part of the AMDGPU ttm init from amdgpu_ttm_init()
* during bring up.
*/
static int amdgpu_ttm_global_init(struct amdgpu_device *adev)
{
struct drm_global_reference *global_ref;
int r;
/* ensure reference is false in case init fails */
adev->mman.mem_global_referenced = false;
global_ref = &adev->mman.mem_global_ref;
global_ref->global_type = DRM_GLOBAL_TTM_MEM;
global_ref->size = sizeof(struct ttm_mem_global);
global_ref->init = &amdgpu_ttm_mem_global_init;
global_ref->release = &amdgpu_ttm_mem_global_release;
r = drm_global_item_ref(global_ref);
if (r) {
DRM_ERROR("Failed setting up TTM memory accounting "
"subsystem.\n");
goto error_mem;
}
adev->mman.bo_global_ref.mem_glob =
adev->mman.mem_global_ref.object;
global_ref = &adev->mman.bo_global_ref.ref;
global_ref->global_type = DRM_GLOBAL_TTM_BO;
global_ref->size = sizeof(struct ttm_bo_global);
global_ref->init = &ttm_bo_global_init;
global_ref->release = &ttm_bo_global_release;
r = drm_global_item_ref(global_ref);
if (r) {
DRM_ERROR("Failed setting up TTM BO subsystem.\n");
goto error_bo;
}
mutex_init(&adev->mman.gtt_window_lock);
adev->mman.mem_global_referenced = true;
return 0;
error_bo:
drm_global_item_unref(&adev->mman.mem_global_ref);
error_mem:
return r;
}
static void amdgpu_ttm_global_fini(struct amdgpu_device *adev)
{
if (adev->mman.mem_global_referenced) {
mutex_destroy(&adev->mman.gtt_window_lock);
drm_global_item_unref(&adev->mman.bo_global_ref.ref);
drm_global_item_unref(&adev->mman.mem_global_ref);
adev->mman.mem_global_referenced = false;
}
}
static int amdgpu_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags)
{
return 0;
@@ -1758,14 +1664,10 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
int r;
u64 vis_vram_limit;
/* initialize global references for vram/gtt */
r = amdgpu_ttm_global_init(adev);
if (r) {
return r;
}
mutex_init(&adev->mman.gtt_window_lock);
/* No others user of address space so set it to 0 */
r = ttm_bo_device_init(&adev->mman.bdev,
adev->mman.bo_global_ref.ref.object,
&amdgpu_bo_driver,
adev->ddev->anon_inode->i_mapping,
DRM_FILE_PAGE_OFFSET,
@@ -1922,7 +1824,6 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GWS);
ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_OA);
ttm_bo_device_release(&adev->mman.bdev);
amdgpu_ttm_global_fini(adev);
adev->mman.initialized = false;
DRM_INFO("amdgpu: ttm finalized\n");
}
@@ -2069,7 +1970,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
unsigned i;
int r;
if (direct_submit && !ring->ready) {
if (direct_submit && !ring->sched.ready) {
DRM_ERROR("Trying to move memory with ring turned off.\n");
return -EINVAL;
}

2
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
View File

@@ -39,8 +39,6 @@
#define AMDGPU_GTT_NUM_TRANSFER_WINDOWS 2
struct amdgpu_mman {
struct ttm_bo_global_ref bo_global_ref;
struct drm_global_reference mem_global_ref;
struct ttm_bo_device bdev;
bool mem_global_referenced;
bool initialized;

12
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
View File

@@ -57,6 +57,17 @@ struct psp_firmware_header_v1_0 {
uint32_t sos_size_bytes;
};
/* version_major=1, version_minor=0 */
struct ta_firmware_header_v1_0 {
struct common_firmware_header header;
uint32_t ta_xgmi_ucode_version;
uint32_t ta_xgmi_offset_bytes;
uint32_t ta_xgmi_size_bytes;
uint32_t ta_ras_ucode_version;
uint32_t ta_ras_offset_bytes;
uint32_t ta_ras_size_bytes;
};
/* version_major=1, version_minor=0 */
struct gfx_firmware_header_v1_0 {
struct common_firmware_header header;
@@ -170,6 +181,7 @@ union amdgpu_firmware_header {
struct mc_firmware_header_v1_0 mc;
struct smc_firmware_header_v1_0 smc;
struct psp_firmware_header_v1_0 psp;
struct ta_firmware_header_v1_0 ta;
struct gfx_firmware_header_v1_0 gfx;
struct rlc_firmware_header_v1_0 rlc;
struct rlc_firmware_header_v2_0 rlc_v2_0;

18
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
View File

@@ -1243,30 +1243,20 @@ int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout)
{
struct dma_fence *fence;
long r;
uint32_t ip_instance = ring->me;
r = amdgpu_uvd_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ip_instance, r);
if (r)
goto error;
}
r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence);
if (r) {
DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ip_instance, r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ip_instance);
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ip_instance, r);
} else {
DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ip_instance, ring->idx);
else if (r > 0)
r = 0;
}
dma_fence_put(fence);

38
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
View File

@@ -1032,8 +1032,10 @@ out:
* @ib: the IB to execute
*
*/
void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
amdgpu_ring_write(ring, VCE_CMD_IB);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
@@ -1079,11 +1081,9 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring)
return 0;
r = amdgpu_ring_alloc(ring, 16);
if (r) {
DRM_ERROR("amdgpu: vce failed to lock ring %d (%d).\n",
ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring, VCE_CMD_END);
amdgpu_ring_commit(ring);
@@ -1093,14 +1093,8 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed\n",
ring->idx);
if (i >= timeout)
r = -ETIMEDOUT;
}
return r;
}
@@ -1121,27 +1115,19 @@ int amdgpu_vce_ring_test_ib(struct amdgpu_ring *ring, long timeout)
return 0;
r = amdgpu_vce_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r);
if (r)
goto error;
}
r = amdgpu_vce_get_destroy_msg(ring, 1, true, &fence);
if (r) {
DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
} else {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
else if (r > 0)
r = 0;
}
error:
dma_fence_put(fence);
return r;

4
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
View File

@@ -65,8 +65,8 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp);
int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx);
int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p, uint32_t ib_idx);
void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch);
void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_job *job,
struct amdgpu_ib *ib, bool ctx_switch);
void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
unsigned flags);
int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring);

100
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
View File

@@ -425,11 +425,9 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring,
PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -441,14 +439,9 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -570,30 +563,20 @@ int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_vcn_dec_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r);
if (r)
goto error;
}
r = amdgpu_vcn_dec_get_destroy_msg(ring, 1, &fence);
if (r) {
DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
} else {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
else if (r > 0)
r = 0;
}
dma_fence_put(fence);
error:
return r;
}
@@ -606,11 +589,9 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_ring_alloc(ring, 16);
if (r) {
DRM_ERROR("amdgpu: vcn enc failed to lock ring %d (%d).\n",
ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring, VCN_ENC_CMD_END);
amdgpu_ring_commit(ring);
@@ -620,14 +601,8 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed\n",
ring->idx);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
}
return r;
}
@@ -742,27 +717,19 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_vcn_enc_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r);
if (r)
goto error;
}
r = amdgpu_vcn_enc_get_destroy_msg(ring, 1, &fence);
if (r) {
DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
} else {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
else if (r > 0)
r = 0;
}
error:
dma_fence_put(fence);
return r;
@@ -778,11 +745,8 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring)
WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring,
PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, 0));
@@ -796,14 +760,8 @@ int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -856,21 +814,18 @@ int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r = 0;
r = amdgpu_vcn_jpeg_set_reg(ring, 1, &fence);
if (r) {
DRM_ERROR("amdgpu: failed to set jpeg register (%ld).\n", r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
r = -ETIMEDOUT;
goto error;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto error;
} else
} else {
r = 0;
}
for (i = 0; i < adev->usec_timeout; i++) {
tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9));
@@ -879,15 +834,10 @@ int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout)
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
else {
DRM_ERROR("ib test failed (0x%08X)\n", tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
dma_fence_put(fence);
error:
return r;
}

140
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
View File

@@ -23,16 +23,6 @@
#include "amdgpu.h"
uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev)
{
uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
addr -= AMDGPU_VA_RESERVED_SIZE;
addr = amdgpu_gmc_sign_extend(addr);
return addr;
}
bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev)
{
/* By now all MMIO pages except mailbox are blocked */
@@ -41,88 +31,6 @@ bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev)
return RREG32_NO_KIQ(0xc040) == 0xffffffff;
}
int amdgpu_allocate_static_csa(struct amdgpu_device *adev)
{
int r;
void *ptr;
r = amdgpu_bo_create_kernel(adev, AMDGPU_CSA_SIZE, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM, &adev->virt.csa_obj,
&adev->virt.csa_vmid0_addr, &ptr);
if (r)
return r;
memset(ptr, 0, AMDGPU_CSA_SIZE);
return 0;
}
void amdgpu_free_static_csa(struct amdgpu_device *adev) {
amdgpu_bo_free_kernel(&adev->virt.csa_obj,
&adev->virt.csa_vmid0_addr,
NULL);
}
/*
* amdgpu_map_static_csa should be called during amdgpu_vm_init
* it maps virtual address amdgpu_csa_vaddr() to this VM, and each command
* submission of GFX should use this virtual address within META_DATA init
* package to support SRIOV gfx preemption.
*/
int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
struct amdgpu_bo_va **bo_va)
{
uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
struct ww_acquire_ctx ticket;
struct list_head list;
struct amdgpu_bo_list_entry pd;
struct ttm_validate_buffer csa_tv;
int r;
INIT_LIST_HEAD(&list);
INIT_LIST_HEAD(&csa_tv.head);
csa_tv.bo = &adev->virt.csa_obj->tbo;
csa_tv.shared = true;
list_add(&csa_tv.head, &list);
amdgpu_vm_get_pd_bo(vm, &list, &pd);
r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
if (r) {
DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
return r;
}
*bo_va = amdgpu_vm_bo_add(adev, vm, adev->virt.csa_obj);
if (!*bo_va) {
ttm_eu_backoff_reservation(&ticket, &list);
DRM_ERROR("failed to create bo_va for static CSA\n");
return -ENOMEM;
}
r = amdgpu_vm_alloc_pts(adev, (*bo_va)->base.vm, csa_addr,
AMDGPU_CSA_SIZE);
if (r) {
DRM_ERROR("failed to allocate pts for static CSA, err=%d\n", r);
amdgpu_vm_bo_rmv(adev, *bo_va);
ttm_eu_backoff_reservation(&ticket, &list);
return r;
}
r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, AMDGPU_CSA_SIZE,
AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
AMDGPU_PTE_EXECUTABLE);
if (r) {
DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r);
amdgpu_vm_bo_rmv(adev, *bo_va);
ttm_eu_backoff_reservation(&ticket, &list);
return r;
}
ttm_eu_backoff_reservation(&ticket, &list);
return 0;
}
void amdgpu_virt_init_setting(struct amdgpu_device *adev)
{
/* enable virtual display */
@@ -162,9 +70,7 @@ uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_read;
if (in_interrupt())
might_sleep();
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);
@@ -210,9 +116,7 @@ void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
if (r < 1 && (adev->in_gpu_reset || in_interrupt()))
goto failed_kiq_write;
if (in_interrupt())
might_sleep();
might_sleep();
while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
@@ -228,6 +132,46 @@ failed_kiq_write:
pr_err("failed to write reg:%x\n", reg);
}
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);
}
/**
* amdgpu_virt_request_full_gpu() - request full gpu access
* @amdgpu: amdgpu device.

13
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
View File

@@ -238,7 +238,6 @@ typedef struct amdgim_vf2pf_info_v2 amdgim_vf2pf_info ;
struct amdgpu_virt {
uint32_t caps;
struct amdgpu_bo *csa_obj;
uint64_t csa_vmid0_addr;
bool chained_ib_support;
uint32_t reg_val_offs;
struct amdgpu_irq_src ack_irq;
@@ -251,8 +250,6 @@ struct amdgpu_virt {
uint32_t gim_feature;
};
#define AMDGPU_CSA_SIZE (8 * 1024)
#define amdgpu_sriov_enabled(adev) \
((adev)->virt.caps & AMDGPU_SRIOV_CAPS_ENABLE_IOV)
@@ -277,17 +274,13 @@ static inline bool is_virtual_machine(void)
#endif
}
struct amdgpu_vm;
uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev);
bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev);
int amdgpu_allocate_static_csa(struct amdgpu_device *adev);
int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
struct amdgpu_bo_va **bo_va);
void amdgpu_free_static_csa(struct amdgpu_device *adev);
void amdgpu_virt_init_setting(struct amdgpu_device *adev);
uint32_t amdgpu_virt_kiq_rreg(struct amdgpu_device *adev, uint32_t reg);
void amdgpu_virt_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v);
void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
uint32_t reg0, uint32_t rreg1,
uint32_t ref, uint32_t mask);
int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init);
int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init);
int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);

34
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
View File

@@ -63,7 +63,7 @@ static struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
{
struct psp_xgmi_topology_info tmp_topology[AMDGPU_MAX_XGMI_DEVICE_PER_HIVE];
struct psp_xgmi_topology_info *tmp_topology;
struct amdgpu_hive_info *hive;
struct amdgpu_xgmi *entry;
struct amdgpu_device *tmp_adev;
@@ -73,10 +73,12 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
if ((adev->asic_type < CHIP_VEGA20) ||
(adev->flags & AMD_IS_APU) )
return 0;
adev->gmc.xgmi.device_id = psp_xgmi_get_device_id(&adev->psp);
adev->gmc.xgmi.node_id = psp_xgmi_get_node_id(&adev->psp);
adev->gmc.xgmi.hive_id = psp_xgmi_get_hive_id(&adev->psp);
memset(&tmp_topology[0], 0, sizeof(tmp_topology));
tmp_topology = kzalloc(sizeof(struct psp_xgmi_topology_info), GFP_KERNEL);
if (!tmp_topology)
return -ENOMEM;
mutex_lock(&xgmi_mutex);
hive = amdgpu_get_xgmi_hive(adev);
if (!hive)
@@ -84,23 +86,28 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
list_add_tail(&adev->gmc.xgmi.head, &hive->device_list);
list_for_each_entry(entry, &hive->device_list, head)
tmp_topology[count++].device_id = entry->device_id;
tmp_topology->nodes[count++].node_id = entry->node_id;
ret = psp_xgmi_get_topology_info(&adev->psp, count, tmp_topology);
if (ret) {
dev_err(adev->dev,
"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
adev->gmc.xgmi.device_id,
adev->gmc.xgmi.hive_id, ret);
goto exit;
/* Each psp need to get the latest topology */
list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count, tmp_topology);
if (ret) {
dev_err(tmp_adev->dev,
"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
tmp_adev->gmc.xgmi.node_id,
tmp_adev->gmc.xgmi.hive_id, ret);
/* To do : continue with some node failed or disable the whole hive */
break;
}
}
/* Each psp need to set the latest topology */
list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
ret = psp_xgmi_set_topology_info(&tmp_adev->psp, count, tmp_topology);
if (ret) {
dev_err(tmp_adev->dev,
"XGMI: Set topology failure on device %llx, hive %llx, ret %d",
tmp_adev->gmc.xgmi.device_id,
tmp_adev->gmc.xgmi.node_id,
tmp_adev->gmc.xgmi.hive_id, ret);
/* To do : continue with some node failed or disable the whole hive */
break;
@@ -113,7 +120,6 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
exit:
mutex_unlock(&xgmi_mutex);
kfree(tmp_topology);
return ret;
}

6
drivers/gpu/drm/amd/amdgpu/ci_dpm.c
View File

@@ -743,19 +743,19 @@ static int ci_enable_didt(struct amdgpu_device *adev, bool enable)
if (pi->caps_sq_ramping || pi->caps_db_ramping ||
pi->caps_td_ramping || pi->caps_tcp_ramping) {
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if (enable) {
ret = ci_program_pt_config_registers(adev, didt_config_ci);
if (ret) {
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
return ret;
}
}
ci_do_enable_didt(adev, enable);
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
return 0;

66
drivers/gpu/drm/amd/amdgpu/cik_sdma.c
View File

@@ -198,7 +198,7 @@ static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring)
static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
int i;
for (i = 0; i < count; i++)
@@ -218,9 +218,11 @@ static void cik_sdma_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
* Schedule an IB in the DMA ring (CIK).
*/
static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 extra_bits = vmid & 0xf;
/* IB packet must end on a 8 DW boundary */
@@ -316,8 +318,8 @@ static void cik_sdma_gfx_stop(struct amdgpu_device *adev)
WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], 0);
}
sdma0->ready = false;
sdma1->ready = false;
sdma0->sched.ready = false;
sdma1->sched.ready = false;
}
/**
@@ -494,18 +496,16 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
/* enable DMA IBs */
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
ring->ready = true;
ring->sched.ready = true;
}
cik_sdma_enable(adev, true);
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
@@ -618,21 +618,17 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
if (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_device_wb_free(adev, index);
return r;
}
if (r)
goto error_free_wb;
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));
@@ -647,15 +643,11 @@ static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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_device_wb_free(adev, index);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_wb:
amdgpu_device_wb_free(adev, index);
return r;
}
@@ -678,20 +670,16 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err0;
}
ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE,
SDMA_WRITE_SUB_OPCODE_LINEAR, 0);
@@ -706,21 +694,16 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
else
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
@@ -822,7 +805,7 @@ static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe,
*/
static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
u32 pad_count;
int i;
@@ -1214,8 +1197,11 @@ static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
u8 instance_id;
DRM_ERROR("Illegal instruction in SDMA command stream\n");
schedule_work(&adev->reset_work);
instance_id = (entry->ring_id & 0x3) >> 0;
drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched);
return 0;
}

144
drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
View File

@@ -1775,18 +1775,15 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", ring->idx, r);
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
if (r)
goto error_free_scratch;
amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
amdgpu_ring_write(ring, (scratch - PACKET3_SET_CONFIG_REG_START));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -1798,13 +1795,11 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
ring->idx, scratch, tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_scratch:
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
@@ -1845,9 +1840,11 @@ static void gfx_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
}
static void gfx_v6_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 header, control = 0;
/* insert SWITCH_BUFFER packet before first IB in the ring frame */
@@ -1892,17 +1889,15 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err1;
}
ib.ptr[0] = PACKET3(PACKET3_SET_CONFIG_REG, 1);
ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_START));
ib.ptr[2] = 0xDEADBEEF;
@@ -1914,22 +1909,16 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err2;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err2;
}
tmp = RREG32(scratch);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n",
scratch, tmp);
else
r = -EINVAL;
}
err2:
amdgpu_ib_free(adev, &ib, NULL);
@@ -1950,9 +1939,9 @@ static void gfx_v6_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
CP_ME_CNTL__CE_HALT_MASK));
WREG32(mmSCRATCH_UMSK, 0);
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
adev->gfx.gfx_ring[i].ready = false;
adev->gfx.gfx_ring[i].sched.ready = false;
for (i = 0; i < adev->gfx.num_compute_rings; i++)
adev->gfx.compute_ring[i].ready = false;
adev->gfx.compute_ring[i].sched.ready = false;
}
udelay(50);
}
@@ -2124,12 +2113,9 @@ static int gfx_v6_0_cp_gfx_resume(struct amdgpu_device *adev)
/* start the rings */
gfx_v6_0_cp_gfx_start(adev);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
return 0;
}
@@ -2227,14 +2213,11 @@ static int gfx_v6_0_cp_compute_resume(struct amdgpu_device *adev)
WREG32(mmCP_RB2_CNTL, tmp);
WREG32(mmCP_RB2_BASE, ring->gpu_addr >> 8);
adev->gfx.compute_ring[0].ready = false;
adev->gfx.compute_ring[1].ready = false;
for (i = 0; i < 2; i++) {
r = amdgpu_ring_test_ring(&adev->gfx.compute_ring[i]);
r = amdgpu_ring_test_helper(&adev->gfx.compute_ring[i]);
if (r)
return r;
adev->gfx.compute_ring[i].ready = true;
}
return 0;
@@ -2368,18 +2351,11 @@ static void gfx_v6_0_ring_emit_wreg(struct amdgpu_ring *ring,
amdgpu_ring_write(ring, val);
}
static void gfx_v6_0_rlc_fini(struct amdgpu_device *adev)
{
amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL);
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL);
amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL);
}
static int gfx_v6_0_rlc_init(struct amdgpu_device *adev)
{
const u32 *src_ptr;
volatile u32 *dst_ptr;
u32 dws, i;
u32 dws;
u64 reg_list_mc_addr;
const struct cs_section_def *cs_data;
int r;
@@ -2394,26 +2370,10 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev)
cs_data = adev->gfx.rlc.cs_data;
if (src_ptr) {
/* save restore block */
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.save_restore_obj,
&adev->gfx.rlc.save_restore_gpu_addr,
(void **)&adev->gfx.rlc.sr_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC sr bo failed\n",
r);
gfx_v6_0_rlc_fini(adev);
/* init save restore block */
r = amdgpu_gfx_rlc_init_sr(adev, dws);
if (r)
return r;
}
/* write the sr buffer */
dst_ptr = adev->gfx.rlc.sr_ptr;
for (i = 0; i < adev->gfx.rlc.reg_list_size; i++)
dst_ptr[i] = cpu_to_le32(src_ptr[i]);
amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
}
if (cs_data) {
@@ -2428,7 +2388,7 @@ static int gfx_v6_0_rlc_init(struct amdgpu_device *adev)
(void **)&adev->gfx.rlc.cs_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
gfx_v6_0_rlc_fini(adev);
amdgpu_gfx_rlc_fini(adev);
return r;
}
@@ -2549,8 +2509,8 @@ static int gfx_v6_0_rlc_resume(struct amdgpu_device *adev)
if (!adev->gfx.rlc_fw)
return -EINVAL;
gfx_v6_0_rlc_stop(adev);
gfx_v6_0_rlc_reset(adev);
adev->gfx.rlc.funcs->stop(adev);
adev->gfx.rlc.funcs->reset(adev);
gfx_v6_0_init_pg(adev);
gfx_v6_0_init_cg(adev);
@@ -2578,7 +2538,7 @@ static int gfx_v6_0_rlc_resume(struct amdgpu_device *adev)
WREG32(mmRLC_UCODE_ADDR, 0);
gfx_v6_0_enable_lbpw(adev, gfx_v6_0_lbpw_supported(adev));
gfx_v6_0_rlc_start(adev);
adev->gfx.rlc.funcs->start(adev);
return 0;
}
@@ -3075,6 +3035,14 @@ static const struct amdgpu_gfx_funcs gfx_v6_0_gfx_funcs = {
.select_me_pipe_q = &gfx_v6_0_select_me_pipe_q
};
static const struct amdgpu_rlc_funcs gfx_v6_0_rlc_funcs = {
.init = gfx_v6_0_rlc_init,
.resume = gfx_v6_0_rlc_resume,
.stop = gfx_v6_0_rlc_stop,
.reset = gfx_v6_0_rlc_reset,
.start = gfx_v6_0_rlc_start
};
static int gfx_v6_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -3082,6 +3050,7 @@ static int gfx_v6_0_early_init(void *handle)
adev->gfx.num_gfx_rings = GFX6_NUM_GFX_RINGS;
adev->gfx.num_compute_rings = GFX6_NUM_COMPUTE_RINGS;
adev->gfx.funcs = &gfx_v6_0_gfx_funcs;
adev->gfx.rlc.funcs = &gfx_v6_0_rlc_funcs;
gfx_v6_0_set_ring_funcs(adev);
gfx_v6_0_set_irq_funcs(adev);
@@ -3114,7 +3083,7 @@ static int gfx_v6_0_sw_init(void *handle)
return r;
}
r = gfx_v6_0_rlc_init(adev);
r = adev->gfx.rlc.funcs->init(adev);
if (r) {
DRM_ERROR("Failed to init rlc BOs!\n");
return r;
@@ -3165,7 +3134,7 @@ static int gfx_v6_0_sw_fini(void *handle)
for (i = 0; i < adev->gfx.num_compute_rings; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
gfx_v6_0_rlc_fini(adev);
amdgpu_gfx_rlc_fini(adev);
return 0;
}
@@ -3177,7 +3146,7 @@ static int gfx_v6_0_hw_init(void *handle)
gfx_v6_0_constants_init(adev);
r = gfx_v6_0_rlc_resume(adev);
r = adev->gfx.rlc.funcs->resume(adev);
if (r)
return r;
@@ -3195,7 +3164,7 @@ static int gfx_v6_0_hw_fini(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
gfx_v6_0_cp_enable(adev, false);
gfx_v6_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
gfx_v6_0_fini_pg(adev);
return 0;
@@ -3393,12 +3362,31 @@ static int gfx_v6_0_eop_irq(struct amdgpu_device *adev,
return 0;
}
static void gfx_v6_0_fault(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
struct amdgpu_ring *ring;
switch (entry->ring_id) {
case 0:
ring = &adev->gfx.gfx_ring[0];
break;
case 1:
case 2:
ring = &adev->gfx.compute_ring[entry->ring_id - 1];
break;
default:
return;
}
drm_sched_fault(&ring->sched);
}
static int gfx_v6_0_priv_reg_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal register access in command stream\n");
schedule_work(&adev->reset_work);
gfx_v6_0_fault(adev, entry);
return 0;
}
@@ -3407,7 +3395,7 @@ static int gfx_v6_0_priv_inst_irq(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal instruction in command stream\n");
schedule_work(&adev->reset_work);
gfx_v6_0_fault(adev, entry);
return 0;
}

276
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
View File

@@ -882,7 +882,6 @@ static const u32 kalindi_rlc_save_restore_register_list[] =
static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev);
static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev);
static void gfx_v7_0_init_pg(struct amdgpu_device *adev);
static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev);
@@ -2064,17 +2063,14 @@ static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", ring->idx, r);
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
if (r)
goto error_free_scratch;
amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -2086,13 +2082,10 @@ static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n", ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
ring->idx, scratch, tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_scratch:
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
@@ -2233,9 +2226,11 @@ static void gfx_v7_0_ring_emit_fence_compute(struct amdgpu_ring *ring,
* on the gfx ring for execution by the GPU.
*/
static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 header, control = 0;
/* insert SWITCH_BUFFER packet before first IB in the ring frame */
@@ -2262,9 +2257,11 @@ static void gfx_v7_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
}
static void gfx_v7_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
@@ -2316,17 +2313,15 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (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;
@@ -2338,22 +2333,16 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err2;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err2;
}
tmp = RREG32(scratch);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n",
scratch, tmp);
else
r = -EINVAL;
}
err2:
amdgpu_ib_free(adev, &ib, NULL);
@@ -2403,7 +2392,7 @@ static void gfx_v7_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
} else {
WREG32(mmCP_ME_CNTL, (CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK));
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
adev->gfx.gfx_ring[i].ready = false;
adev->gfx.gfx_ring[i].sched.ready = false;
}
udelay(50);
}
@@ -2613,12 +2602,9 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev)
/* start the ring */
gfx_v7_0_cp_gfx_start(adev);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
return 0;
}
@@ -2675,7 +2661,7 @@ static void gfx_v7_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
} else {
WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
for (i = 0; i < adev->gfx.num_compute_rings; i++)
adev->gfx.compute_ring[i].ready = false;
adev->gfx.compute_ring[i].sched.ready = false;
}
udelay(50);
}
@@ -2781,7 +2767,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
* GFX7_MEC_HPD_SIZE * 2;
r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.mec.hpd_eop_obj,
&adev->gfx.mec.hpd_eop_gpu_addr,
(void **)&hpd);
@@ -3106,10 +3092,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
ring = &adev->gfx.compute_ring[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r)
ring->ready = false;
amdgpu_ring_test_helper(ring);
}
return 0;
@@ -3268,18 +3251,10 @@ static void gfx_v7_0_ring_emit_wreg(struct amdgpu_ring *ring,
* The RLC is a multi-purpose microengine that handles a
* variety of functions.
*/
static void gfx_v7_0_rlc_fini(struct amdgpu_device *adev)
{
amdgpu_bo_free_kernel(&adev->gfx.rlc.save_restore_obj, NULL, NULL);
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL);
amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL);
}
static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
{
const u32 *src_ptr;
volatile u32 *dst_ptr;
u32 dws, i;
u32 dws;
const struct cs_section_def *cs_data;
int r;
@@ -3306,66 +3281,23 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
cs_data = adev->gfx.rlc.cs_data;
if (src_ptr) {
/* save restore block */
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.save_restore_obj,
&adev->gfx.rlc.save_restore_gpu_addr,
(void **)&adev->gfx.rlc.sr_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create, pin or map of RLC sr bo failed\n", r);
gfx_v7_0_rlc_fini(adev);
/* init save restore block */
r = amdgpu_gfx_rlc_init_sr(adev, dws);
if (r)
return r;
}
/* write the sr buffer */
dst_ptr = adev->gfx.rlc.sr_ptr;
for (i = 0; i < adev->gfx.rlc.reg_list_size; i++)
dst_ptr[i] = cpu_to_le32(src_ptr[i]);
amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
}
if (cs_data) {
/* clear state block */
adev->gfx.rlc.clear_state_size = dws = gfx_v7_0_get_csb_size(adev);
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
gfx_v7_0_rlc_fini(adev);
/* init clear state block */
r = amdgpu_gfx_rlc_init_csb(adev);
if (r)
return r;
}
/* set up the cs buffer */
dst_ptr = adev->gfx.rlc.cs_ptr;
gfx_v7_0_get_csb_buffer(adev, dst_ptr);
amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
}
if (adev->gfx.rlc.cp_table_size) {
r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size,
PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r);
gfx_v7_0_rlc_fini(adev);
r = amdgpu_gfx_rlc_init_cpt(adev);
if (r)
return r;
}
gfx_v7_0_init_cp_pg_table(adev);
amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
}
return 0;
@@ -3446,7 +3378,12 @@ static u32 gfx_v7_0_halt_rlc(struct amdgpu_device *adev)
return orig;
}
static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
static bool gfx_v7_0_is_rlc_enabled(struct amdgpu_device *adev)
{
return true;
}
static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev)
{
u32 tmp, i, mask;
@@ -3468,7 +3405,7 @@ static void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
}
}
static void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev)
{
u32 tmp;
@@ -3545,13 +3482,13 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
adev->gfx.rlc_feature_version = le32_to_cpu(
hdr->ucode_feature_version);
gfx_v7_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
/* disable CG */
tmp = RREG32(mmRLC_CGCG_CGLS_CTRL) & 0xfffffffc;
WREG32(mmRLC_CGCG_CGLS_CTRL, tmp);
gfx_v7_0_rlc_reset(adev);
adev->gfx.rlc.funcs->reset(adev);
gfx_v7_0_init_pg(adev);
@@ -3582,7 +3519,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
if (adev->asic_type == CHIP_BONAIRE)
WREG32(mmRLC_DRIVER_CPDMA_STATUS, 0);
gfx_v7_0_rlc_start(adev);
adev->gfx.rlc.funcs->start(adev);
return 0;
}
@@ -3784,72 +3721,12 @@ static void gfx_v7_0_enable_gds_pg(struct amdgpu_device *adev, bool enable)
WREG32(mmRLC_PG_CNTL, data);
}
static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev)
static int gfx_v7_0_cp_pg_table_num(struct amdgpu_device *adev)
{
const __le32 *fw_data;
volatile u32 *dst_ptr;
int me, i, max_me = 4;
u32 bo_offset = 0;
u32 table_offset, table_size;
if (adev->asic_type == CHIP_KAVERI)
max_me = 5;
if (adev->gfx.rlc.cp_table_ptr == NULL)
return;
/* write the cp table buffer */
dst_ptr = adev->gfx.rlc.cp_table_ptr;
for (me = 0; me < max_me; me++) {
if (me == 0) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
fw_data = (const __le32 *)
(adev->gfx.ce_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 1) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
fw_data = (const __le32 *)
(adev->gfx.pfp_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 2) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
fw_data = (const __le32 *)
(adev->gfx.me_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 3) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec2_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
}
for (i = 0; i < table_size; i ++) {
dst_ptr[bo_offset + i] =
cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
}
bo_offset += table_size;
}
return 5;
else
return 4;
}
static void gfx_v7_0_enable_gfx_cgpg(struct amdgpu_device *adev,
@@ -4288,8 +4165,17 @@ static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = {
};
static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = {
.enter_safe_mode = gfx_v7_0_enter_rlc_safe_mode,
.exit_safe_mode = gfx_v7_0_exit_rlc_safe_mode
.is_rlc_enabled = gfx_v7_0_is_rlc_enabled,
.set_safe_mode = gfx_v7_0_set_safe_mode,
.unset_safe_mode = gfx_v7_0_unset_safe_mode,
.init = gfx_v7_0_rlc_init,
.get_csb_size = gfx_v7_0_get_csb_size,
.get_csb_buffer = gfx_v7_0_get_csb_buffer,
.get_cp_table_num = gfx_v7_0_cp_pg_table_num,
.resume = gfx_v7_0_rlc_resume,
.stop = gfx_v7_0_rlc_stop,
.reset = gfx_v7_0_rlc_reset,
.start = gfx_v7_0_rlc_start
};
static int gfx_v7_0_early_init(void *handle)
@@ -4540,7 +4426,7 @@ static int gfx_v7_0_sw_init(void *handle)
return r;
}
r = gfx_v7_0_rlc_init(adev);
r = adev->gfx.rlc.funcs->init(adev);
if (r) {
DRM_ERROR("Failed to init rlc BOs!\n");
return r;
@@ -4604,7 +4490,7 @@ static int gfx_v7_0_sw_fini(void *handle)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
gfx_v7_0_cp_compute_fini(adev);
gfx_v7_0_rlc_fini(adev);
amdgpu_gfx_rlc_fini(adev);
gfx_v7_0_mec_fini(adev);
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
@@ -4627,7 +4513,7 @@ static int gfx_v7_0_hw_init(void *handle)
gfx_v7_0_constants_init(adev);
/* init rlc */
r = gfx_v7_0_rlc_resume(adev);
r = adev->gfx.rlc.funcs->resume(adev);
if (r)
return r;
@@ -4645,7 +4531,7 @@ static int gfx_v7_0_hw_fini(void *handle)
amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
gfx_v7_0_cp_enable(adev, false);
gfx_v7_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
gfx_v7_0_fini_pg(adev);
return 0;
@@ -4730,7 +4616,7 @@ static int gfx_v7_0_soft_reset(void *handle)
gfx_v7_0_update_cg(adev, false);
/* stop the rlc */
gfx_v7_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
/* Disable GFX parsing/prefetching */
WREG32(mmCP_ME_CNTL, CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK);
@@ -4959,12 +4845,36 @@ static int gfx_v7_0_eop_irq(struct amdgpu_device *adev,
return 0;
}
static void gfx_v7_0_fault(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
struct amdgpu_ring *ring;
u8 me_id, pipe_id;
int i;
me_id = (entry->ring_id & 0x0c) >> 2;
pipe_id = (entry->ring_id & 0x03) >> 0;
switch (me_id) {
case 0:
drm_sched_fault(&adev->gfx.gfx_ring[0].sched);
break;
case 1:
case 2:
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
ring = &adev->gfx.compute_ring[i];
if ((ring->me == me_id) && (ring->pipe == pipe_id))
drm_sched_fault(&ring->sched);
}
break;
}
}
static int gfx_v7_0_priv_reg_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal register access in command stream\n");
schedule_work(&adev->reset_work);
gfx_v7_0_fault(adev, entry);
return 0;
}
@@ -4974,7 +4884,7 @@ static int gfx_v7_0_priv_inst_irq(struct amdgpu_device *adev,
{
DRM_ERROR("Illegal instruction in command stream\n");
// XXX soft reset the gfx block only
schedule_work(&adev->reset_work);
gfx_v7_0_fault(adev, entry);
return 0;
}

353
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
View File

@@ -54,7 +54,7 @@
#include "ivsrcid/ivsrcid_vislands30.h"
#define GFX8_NUM_GFX_RINGS 1
#define GFX8_MEC_HPD_SIZE 2048
#define GFX8_MEC_HPD_SIZE 4096
#define TOPAZ_GB_ADDR_CONFIG_GOLDEN 0x22010001
#define CARRIZO_GB_ADDR_CONFIG_GOLDEN 0x22010001
@@ -839,18 +839,14 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
ring->idx, r);
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
if (r)
goto error_free_scratch;
amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -862,14 +858,11 @@ static int gfx_v8_0_ring_test_ring(struct amdgpu_ring *ring)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
ring->idx, scratch, tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_scratch:
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
@@ -886,19 +879,16 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 16, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err1;
}
ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
ib.ptr[2] = lower_32_bits(gpu_addr);
@@ -912,22 +902,17 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
r = -ETIMEDOUT;
goto err2;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err2;
}
tmp = adev->wb.wb[index];
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("ib test on ring %d failed\n", ring->idx);
else
r = -EINVAL;
}
err2:
amdgpu_ib_free(adev, &ib, NULL);
@@ -1298,81 +1283,16 @@ static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev,
buffer[count++] = cpu_to_le32(0);
}
static void cz_init_cp_jump_table(struct amdgpu_device *adev)
static int gfx_v8_0_cp_jump_table_num(struct amdgpu_device *adev)
{
const __le32 *fw_data;
volatile u32 *dst_ptr;
int me, i, max_me = 4;
u32 bo_offset = 0;
u32 table_offset, table_size;
if (adev->asic_type == CHIP_CARRIZO)
max_me = 5;
/* write the cp table buffer */
dst_ptr = adev->gfx.rlc.cp_table_ptr;
for (me = 0; me < max_me; me++) {
if (me == 0) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
fw_data = (const __le32 *)
(adev->gfx.ce_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 1) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
fw_data = (const __le32 *)
(adev->gfx.pfp_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 2) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
fw_data = (const __le32 *)
(adev->gfx.me_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 3) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 4) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec2_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
}
for (i = 0; i < table_size; i ++) {
dst_ptr[bo_offset + i] =
cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
}
bo_offset += table_size;
}
}
static void gfx_v8_0_rlc_fini(struct amdgpu_device *adev)
{
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj, NULL, NULL);
amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj, NULL, NULL);
return 5;
else
return 4;
}
static int gfx_v8_0_rlc_init(struct amdgpu_device *adev)
{
volatile u32 *dst_ptr;
u32 dws;
const struct cs_section_def *cs_data;
int r;
@@ -1381,44 +1301,18 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev)
cs_data = adev->gfx.rlc.cs_data;
if (cs_data) {
/* clear state block */
adev->gfx.rlc.clear_state_size = dws = gfx_v8_0_get_csb_size(adev);
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
gfx_v8_0_rlc_fini(adev);
/* init clear state block */
r = amdgpu_gfx_rlc_init_csb(adev);
if (r)
return r;
}
/* set up the cs buffer */
dst_ptr = adev->gfx.rlc.cs_ptr;
gfx_v8_0_get_csb_buffer(adev, dst_ptr);
amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
}
if ((adev->asic_type == CHIP_CARRIZO) ||
(adev->asic_type == CHIP_STONEY)) {
adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */
r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size,
PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
if (r) {
dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r);
r = amdgpu_gfx_rlc_init_cpt(adev);
if (r)
return r;
}
cz_init_cp_jump_table(adev);
amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
}
return 0;
@@ -1443,7 +1337,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
mec_hpd_size = adev->gfx.num_compute_rings * GFX8_MEC_HPD_SIZE;
r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.mec.hpd_eop_obj,
&adev->gfx.mec.hpd_eop_gpu_addr,
(void **)&hpd);
@@ -1629,7 +1523,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
return 0;
/* bail if the compute ring is not ready */
if (!ring->ready)
if (!ring->sched.ready)
return 0;
tmp = RREG32(mmGB_EDC_MODE);
@@ -2088,7 +1982,7 @@ static int gfx_v8_0_sw_init(void *handle)
return r;
}
r = gfx_v8_0_rlc_init(adev);
r = adev->gfx.rlc.funcs->init(adev);
if (r) {
DRM_ERROR("Failed to init rlc BOs!\n");
return r;
@@ -2181,7 +2075,7 @@ static int gfx_v8_0_sw_fini(void *handle)
amdgpu_gfx_kiq_fini(adev);
gfx_v8_0_mec_fini(adev);
gfx_v8_0_rlc_fini(adev);
amdgpu_gfx_rlc_fini(adev);
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
@@ -4175,10 +4069,10 @@ static void gfx_v8_0_rlc_start(struct amdgpu_device *adev)
static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev)
{
gfx_v8_0_rlc_stop(adev);
gfx_v8_0_rlc_reset(adev);
adev->gfx.rlc.funcs->stop(adev);
adev->gfx.rlc.funcs->reset(adev);
gfx_v8_0_init_pg(adev);
gfx_v8_0_rlc_start(adev);
adev->gfx.rlc.funcs->start(adev);
return 0;
}
@@ -4197,7 +4091,7 @@ static void gfx_v8_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, 1);
tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, 1);
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
adev->gfx.gfx_ring[i].ready = false;
adev->gfx.gfx_ring[i].sched.ready = false;
}
WREG32(mmCP_ME_CNTL, tmp);
udelay(50);
@@ -4379,10 +4273,8 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
/* start the ring */
amdgpu_ring_clear_ring(ring);
gfx_v8_0_cp_gfx_start(adev);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r)
ring->ready = false;
ring->sched.ready = true;
r = amdgpu_ring_test_helper(ring);
return r;
}
@@ -4396,8 +4288,8 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
} else {
WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
for (i = 0; i < adev->gfx.num_compute_rings; i++)
adev->gfx.compute_ring[i].ready = false;
adev->gfx.kiq.ring.ready = false;
adev->gfx.compute_ring[i].sched.ready = false;
adev->gfx.kiq.ring.sched.ready = false;
}
udelay(50);
}
@@ -4473,11 +4365,9 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
}
r = amdgpu_ring_test_ring(kiq_ring);
if (r) {
r = amdgpu_ring_test_helper(kiq_ring);
if (r)
DRM_ERROR("KCQ enable failed\n");
kiq_ring->ready = false;
}
return r;
}
@@ -4781,7 +4671,7 @@ static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev)
amdgpu_bo_kunmap(ring->mqd_obj);
ring->mqd_ptr = NULL;
amdgpu_bo_unreserve(ring->mqd_obj);
ring->ready = true;
ring->sched.ready = true;
return 0;
}
@@ -4820,10 +4710,7 @@ static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev)
*/
for (i = adev->gfx.num_compute_rings - 1; i >= 0; i--) {
ring = &adev->gfx.compute_ring[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r)
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
}
done:
@@ -4867,7 +4754,7 @@ static int gfx_v8_0_hw_init(void *handle)
gfx_v8_0_init_golden_registers(adev);
gfx_v8_0_constants_init(adev);
r = gfx_v8_0_rlc_resume(adev);
r = adev->gfx.rlc.funcs->resume(adev);
if (r)
return r;
@@ -4899,7 +4786,7 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
amdgpu_ring_write(kiq_ring, 0);
amdgpu_ring_write(kiq_ring, 0);
}
r = amdgpu_ring_test_ring(kiq_ring);
r = amdgpu_ring_test_helper(kiq_ring);
if (r)
DRM_ERROR("KCQ disable failed\n");
@@ -4973,16 +4860,16 @@ static int gfx_v8_0_hw_fini(void *handle)
pr_debug("For SRIOV client, shouldn't do anything.\n");
return 0;
}
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if (!gfx_v8_0_wait_for_idle(adev))
gfx_v8_0_cp_enable(adev, false);
else
pr_err("cp is busy, skip halt cp\n");
if (!gfx_v8_0_wait_for_rlc_idle(adev))
gfx_v8_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
else
pr_err("rlc is busy, skip halt rlc\n");
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
return 0;
}
@@ -5071,7 +4958,7 @@ static int gfx_v8_0_pre_soft_reset(void *handle)
srbm_soft_reset = adev->gfx.srbm_soft_reset;
/* stop the rlc */
gfx_v8_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) ||
REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX))
@@ -5197,7 +5084,7 @@ static int gfx_v8_0_post_soft_reset(void *handle)
REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX))
gfx_v8_0_cp_gfx_resume(adev);
gfx_v8_0_rlc_start(adev);
adev->gfx.rlc.funcs->start(adev);
return 0;
}
@@ -5445,7 +5332,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
AMD_PG_SUPPORT_RLC_SMU_HS |
AMD_PG_SUPPORT_CP |
AMD_PG_SUPPORT_GFX_DMG))
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
switch (adev->asic_type) {
case CHIP_CARRIZO:
case CHIP_STONEY:
@@ -5499,7 +5386,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
AMD_PG_SUPPORT_RLC_SMU_HS |
AMD_PG_SUPPORT_CP |
AMD_PG_SUPPORT_GFX_DMG))
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
return 0;
}
@@ -5593,57 +5480,53 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev,
#define RLC_GPR_REG2__MESSAGE__SHIFT 0x00000001
#define RLC_GPR_REG2__MESSAGE_MASK 0x0000001e
static void iceland_enter_rlc_safe_mode(struct amdgpu_device *adev)
static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev)
{
u32 data;
uint32_t rlc_setting;
rlc_setting = RREG32(mmRLC_CNTL);
if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
return false;
return true;
}
static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
{
uint32_t data;
unsigned i;
data = RREG32(mmRLC_CNTL);
if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK))
return;
data |= RLC_SAFE_MODE__CMD_MASK;
data &= ~RLC_SAFE_MODE__MESSAGE_MASK;
data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
WREG32(mmRLC_SAFE_MODE, data);
if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) {
data |= RLC_SAFE_MODE__CMD_MASK;
data &= ~RLC_SAFE_MODE__MESSAGE_MASK;
data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
WREG32(mmRLC_SAFE_MODE, data);
for (i = 0; i < adev->usec_timeout; i++) {
if ((RREG32(mmRLC_GPM_STAT) &
(RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK |
RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) ==
(RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK |
RLC_GPM_STAT__GFX_POWER_STATUS_MASK))
break;
udelay(1);
}
for (i = 0; i < adev->usec_timeout; i++) {
if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
break;
udelay(1);
}
adev->gfx.rlc.in_safe_mode = true;
/* wait for RLC_SAFE_MODE */
for (i = 0; i < adev->usec_timeout; i++) {
if ((RREG32(mmRLC_GPM_STAT) &
(RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK |
RLC_GPM_STAT__GFX_POWER_STATUS_MASK)) ==
(RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK |
RLC_GPM_STAT__GFX_POWER_STATUS_MASK))
break;
udelay(1);
}
for (i = 0; i < adev->usec_timeout; i++) {
if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
break;
udelay(1);
}
}
static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev)
static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
{
u32 data = 0;
uint32_t data;
unsigned i;
data = RREG32(mmRLC_CNTL);
if (!(data & RLC_CNTL__RLC_ENABLE_F32_MASK))
return;
if (adev->cg_flags & (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) {
if (adev->gfx.rlc.in_safe_mode) {
data |= RLC_SAFE_MODE__CMD_MASK;
data &= ~RLC_SAFE_MODE__MESSAGE_MASK;
WREG32(mmRLC_SAFE_MODE, data);
adev->gfx.rlc.in_safe_mode = false;
}
}
data |= RLC_SAFE_MODE__CMD_MASK;
data &= ~RLC_SAFE_MODE__MESSAGE_MASK;
WREG32(mmRLC_SAFE_MODE, data);
for (i = 0; i < adev->usec_timeout; i++) {
if (!REG_GET_FIELD(RREG32(mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
@@ -5653,8 +5536,17 @@ static void iceland_exit_rlc_safe_mode(struct amdgpu_device *adev)
}
static const struct amdgpu_rlc_funcs iceland_rlc_funcs = {
.enter_safe_mode = iceland_enter_rlc_safe_mode,
.exit_safe_mode = iceland_exit_rlc_safe_mode
.is_rlc_enabled = gfx_v8_0_is_rlc_enabled,
.set_safe_mode = gfx_v8_0_set_safe_mode,
.unset_safe_mode = gfx_v8_0_unset_safe_mode,
.init = gfx_v8_0_rlc_init,
.get_csb_size = gfx_v8_0_get_csb_size,
.get_csb_buffer = gfx_v8_0_get_csb_buffer,
.get_cp_table_num = gfx_v8_0_cp_jump_table_num,
.resume = gfx_v8_0_rlc_resume,
.stop = gfx_v8_0_rlc_stop,
.reset = gfx_v8_0_rlc_reset,
.start = gfx_v8_0_rlc_start
};
static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
@@ -5662,7 +5554,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
{
uint32_t temp, data;
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
/* It is disabled by HW by default */
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
@@ -5758,7 +5650,7 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
gfx_v8_0_wait_for_rlc_serdes(adev);
}
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -5768,7 +5660,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL);
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
temp1 = data1 = RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
@@ -5851,7 +5743,7 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
gfx_v8_0_wait_for_rlc_serdes(adev);
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
bool enable)
@@ -6131,9 +6023,11 @@ static void gfx_v8_0_ring_emit_vgt_flush(struct amdgpu_ring *ring)
}
static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 header, control = 0;
if (ib->flags & AMDGPU_IB_FLAG_CE)
@@ -6161,9 +6055,11 @@ static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
}
static void gfx_v8_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
@@ -6738,12 +6634,39 @@ static int gfx_v8_0_eop_irq(struct amdgpu_device *adev,
return 0;
}
static void gfx_v8_0_fault(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
u8 me_id, pipe_id, queue_id;
struct amdgpu_ring *ring;
int i;
me_id = (entry->ring_id & 0x0c) >> 2;
pipe_id = (entry->ring_id & 0x03) >> 0;
queue_id = (entry->ring_id & 0x70) >> 4;
switch (me_id) {
case 0:
drm_sched_fault(&adev->gfx.gfx_ring[0].sched);
break;
case 1:
case 2:
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
ring = &adev->gfx.compute_ring[i];
if (ring->me == me_id && ring->pipe == pipe_id &&
ring->queue == queue_id)
drm_sched_fault(&ring->sched);
}
break;
}
}
static int gfx_v8_0_priv_reg_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal register access in command stream\n");
schedule_work(&adev->reset_work);
gfx_v8_0_fault(adev, entry);
return 0;
}
@@ -6752,7 +6675,7 @@ static int gfx_v8_0_priv_inst_irq(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal instruction in command stream\n");
schedule_work(&adev->reset_work);
gfx_v8_0_fault(adev, entry);
return 0;
}
@@ -6976,10 +6899,8 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = {
17 + /* gfx_v8_0_ring_emit_vm_flush */
7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_kiq x3 for user fence, vm fence */
.emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_compute */
.emit_ib = gfx_v8_0_ring_emit_ib_compute,
.emit_fence = gfx_v8_0_ring_emit_fence_kiq,
.test_ring = gfx_v8_0_ring_test_ring,
.test_ib = gfx_v8_0_ring_test_ib,
.insert_nop = amdgpu_ring_insert_nop,
.pad_ib = amdgpu_ring_generic_pad_ib,
.emit_rreg = gfx_v8_0_ring_emit_rreg,

382
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
View File

@@ -41,7 +41,7 @@
#include "ivsrcid/gfx/irqsrcs_gfx_9_0.h"
#define GFX9_NUM_GFX_RINGS 1
#define GFX9_MEC_HPD_SIZE 2048
#define GFX9_MEC_HPD_SIZE 4096
#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
#define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L
@@ -396,18 +396,14 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_gfx_scratch_get(adev, &scratch);
if (r) {
DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
if (r)
return r;
}
WREG32(scratch, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
ring->idx, r);
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
if (r)
goto error_free_scratch;
amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -419,14 +415,11 @@ static int gfx_v9_0_ring_test_ring(struct amdgpu_ring *ring)
break;
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
ring->idx, scratch, tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_scratch:
amdgpu_gfx_scratch_free(adev, scratch);
return r;
}
@@ -443,19 +436,16 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 16, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err1;
}
ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
ib.ptr[2] = lower_32_bits(gpu_addr);
@@ -469,22 +459,17 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
r = -ETIMEDOUT;
goto err2;
r = -ETIMEDOUT;
goto err2;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err2;
goto err2;
}
tmp = adev->wb.wb[index];
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
r = 0;
} else {
DRM_ERROR("ib test on ring %d failed\n", ring->idx);
r = -EINVAL;
}
if (tmp == 0xDEADBEEF)
r = 0;
else
r = -EINVAL;
err2:
amdgpu_ib_free(adev, &ib, NULL);
@@ -1065,85 +1050,13 @@ static void gfx_v9_0_enable_lbpw(struct amdgpu_device *adev, bool enable)
WREG32_FIELD15(GC, 0, RLC_LB_CNTL, LOAD_BALANCE_ENABLE, enable ? 1 : 0);
}
static void rv_init_cp_jump_table(struct amdgpu_device *adev)
static int gfx_v9_0_cp_jump_table_num(struct amdgpu_device *adev)
{
const __le32 *fw_data;
volatile u32 *dst_ptr;
int me, i, max_me = 5;
u32 bo_offset = 0;
u32 table_offset, table_size;
/* write the cp table buffer */
dst_ptr = adev->gfx.rlc.cp_table_ptr;
for (me = 0; me < max_me; me++) {
if (me == 0) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
fw_data = (const __le32 *)
(adev->gfx.ce_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 1) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
fw_data = (const __le32 *)
(adev->gfx.pfp_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 2) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
fw_data = (const __le32 *)
(adev->gfx.me_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 3) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
} else if (me == 4) {
const struct gfx_firmware_header_v1_0 *hdr =
(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
fw_data = (const __le32 *)
(adev->gfx.mec2_fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
table_offset = le32_to_cpu(hdr->jt_offset);
table_size = le32_to_cpu(hdr->jt_size);
}
for (i = 0; i < table_size; i ++) {
dst_ptr[bo_offset + i] =
cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
}
bo_offset += table_size;
}
}
static void gfx_v9_0_rlc_fini(struct amdgpu_device *adev)
{
/* clear state block */
amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
/* jump table block */
amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
return 5;
}
static int gfx_v9_0_rlc_init(struct amdgpu_device *adev)
{
volatile u32 *dst_ptr;
u32 dws;
const struct cs_section_def *cs_data;
int r;
@@ -1152,45 +1065,18 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev)
cs_data = adev->gfx.rlc.cs_data;
if (cs_data) {
/* clear state block */
adev->gfx.rlc.clear_state_size = dws = gfx_v9_0_get_csb_size(adev);
r = amdgpu_bo_create_reserved(adev, dws * 4, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.clear_state_obj,
&adev->gfx.rlc.clear_state_gpu_addr,
(void **)&adev->gfx.rlc.cs_ptr);
if (r) {
dev_err(adev->dev, "(%d) failed to create rlc csb bo\n",
r);
gfx_v9_0_rlc_fini(adev);
/* init clear state block */
r = amdgpu_gfx_rlc_init_csb(adev);
if (r)
return r;
}
/* set up the cs buffer */
dst_ptr = adev->gfx.rlc.cs_ptr;
gfx_v9_0_get_csb_buffer(adev, dst_ptr);
amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
}
if (adev->asic_type == CHIP_RAVEN) {
/* TODO: double check the cp_table_size for RV */
adev->gfx.rlc.cp_table_size = ALIGN(96 * 5 * 4, 2048) + (64 * 1024); /* JT + GDS */
r = amdgpu_bo_create_reserved(adev, adev->gfx.rlc.cp_table_size,
PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.rlc.cp_table_obj,
&adev->gfx.rlc.cp_table_gpu_addr,
(void **)&adev->gfx.rlc.cp_table_ptr);
if (r) {
dev_err(adev->dev,
"(%d) failed to create cp table bo\n", r);
gfx_v9_0_rlc_fini(adev);
r = amdgpu_gfx_rlc_init_cpt(adev);
if (r)
return r;
}
rv_init_cp_jump_table(adev);
amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj);
amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
}
switch (adev->asic_type) {
@@ -1264,7 +1150,7 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev)
mec_hpd_size = adev->gfx.num_compute_rings * GFX9_MEC_HPD_SIZE;
r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_DOMAIN_VRAM,
&adev->gfx.mec.hpd_eop_obj,
&adev->gfx.mec.hpd_eop_gpu_addr,
(void **)&hpd);
@@ -1635,8 +1521,8 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev)
/* Clear GDS reserved memory */
r = amdgpu_ring_alloc(ring, 17);
if (r) {
DRM_ERROR("amdgpu: NGG failed to lock ring %d (%d).\n",
ring->idx, r);
DRM_ERROR("amdgpu: NGG failed to lock ring %s (%d).\n",
ring->name, r);
return r;
}
@@ -1748,7 +1634,7 @@ static int gfx_v9_0_sw_init(void *handle)
return r;
}
r = gfx_v9_0_rlc_init(adev);
r = adev->gfx.rlc.funcs->init(adev);
if (r) {
DRM_ERROR("Failed to init rlc BOs!\n");
return r;
@@ -2498,12 +2384,12 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
return 0;
}
gfx_v9_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
/* disable CG */
WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, 0);
gfx_v9_0_rlc_reset(adev);
adev->gfx.rlc.funcs->reset(adev);
gfx_v9_0_init_pg(adev);
@@ -2514,15 +2400,24 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
return r;
}
if (adev->asic_type == CHIP_RAVEN ||
adev->asic_type == CHIP_VEGA20) {
if (amdgpu_lbpw != 0)
switch (adev->asic_type) {
case CHIP_RAVEN:
if (amdgpu_lbpw == 0)
gfx_v9_0_enable_lbpw(adev, false);
else
gfx_v9_0_enable_lbpw(adev, true);
break;
case CHIP_VEGA20:
if (amdgpu_lbpw > 0)
gfx_v9_0_enable_lbpw(adev, true);
else
gfx_v9_0_enable_lbpw(adev, false);
break;
default:
break;
}
gfx_v9_0_rlc_start(adev);
adev->gfx.rlc.funcs->start(adev);
return 0;
}
@@ -2537,7 +2432,7 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
if (!enable) {
for (i = 0; i < adev->gfx.num_gfx_rings; i++)
adev->gfx.gfx_ring[i].ready = false;
adev->gfx.gfx_ring[i].sched.ready = false;
}
WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp);
udelay(50);
@@ -2727,7 +2622,7 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev)
/* start the ring */
gfx_v9_0_cp_gfx_start(adev);
ring->ready = true;
ring->sched.ready = true;
return 0;
}
@@ -2742,8 +2637,8 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
WREG32_SOC15(GC, 0, mmCP_MEC_CNTL,
(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
for (i = 0; i < adev->gfx.num_compute_rings; i++)
adev->gfx.compute_ring[i].ready = false;
adev->gfx.kiq.ring.ready = false;
adev->gfx.compute_ring[i].sched.ready = false;
adev->gfx.kiq.ring.sched.ready = false;
}
udelay(50);
}
@@ -2866,11 +2761,9 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev)
amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
}
r = amdgpu_ring_test_ring(kiq_ring);
if (r) {
r = amdgpu_ring_test_helper(kiq_ring);
if (r)
DRM_ERROR("KCQ enable failed\n");
kiq_ring->ready = false;
}
return r;
}
@@ -3249,7 +3142,7 @@ static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev)
amdgpu_bo_kunmap(ring->mqd_obj);
ring->mqd_ptr = NULL;
amdgpu_bo_unreserve(ring->mqd_obj);
ring->ready = true;
ring->sched.ready = true;
return 0;
}
@@ -3314,19 +3207,13 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
return r;
ring = &adev->gfx.gfx_ring[0];
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
ring = &adev->gfx.compute_ring[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r)
ring->ready = false;
amdgpu_ring_test_helper(ring);
}
gfx_v9_0_enable_gui_idle_interrupt(adev, true);
@@ -3353,7 +3240,7 @@ static int gfx_v9_0_hw_init(void *handle)
if (r)
return r;
r = gfx_v9_0_rlc_resume(adev);
r = adev->gfx.rlc.funcs->resume(adev);
if (r)
return r;
@@ -3391,7 +3278,7 @@ static int gfx_v9_0_kcq_disable(struct amdgpu_device *adev)
amdgpu_ring_write(kiq_ring, 0);
amdgpu_ring_write(kiq_ring, 0);
}
r = amdgpu_ring_test_ring(kiq_ring);
r = amdgpu_ring_test_helper(kiq_ring);
if (r)
DRM_ERROR("KCQ disable failed\n");
@@ -3433,7 +3320,7 @@ static int gfx_v9_0_hw_fini(void *handle)
}
gfx_v9_0_cp_enable(adev, false);
gfx_v9_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
gfx_v9_0_csb_vram_unpin(adev);
@@ -3508,7 +3395,7 @@ static int gfx_v9_0_soft_reset(void *handle)
if (grbm_soft_reset) {
/* stop the rlc */
gfx_v9_0_rlc_stop(adev);
adev->gfx.rlc.funcs->stop(adev);
/* Disable GFX parsing/prefetching */
gfx_v9_0_cp_gfx_enable(adev, false);
@@ -3607,64 +3494,47 @@ static int gfx_v9_0_late_init(void *handle)
return 0;
}
static void gfx_v9_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
static bool gfx_v9_0_is_rlc_enabled(struct amdgpu_device *adev)
{
uint32_t rlc_setting, data;
unsigned i;
if (adev->gfx.rlc.in_safe_mode)
return;
uint32_t rlc_setting;
/* if RLC is not enabled, do nothing */
rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL);
if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
return;
return false;
if (adev->cg_flags &
(AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
AMD_CG_SUPPORT_GFX_3D_CGCG)) {
data = RLC_SAFE_MODE__CMD_MASK;
data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
return true;
}
/* wait for RLC_SAFE_MODE */
for (i = 0; i < adev->usec_timeout; i++) {
if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
break;
udelay(1);
}
adev->gfx.rlc.in_safe_mode = true;
static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev)
{
uint32_t data;
unsigned i;
data = RLC_SAFE_MODE__CMD_MASK;
data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
/* wait for RLC_SAFE_MODE */
for (i = 0; i < adev->usec_timeout; i++) {
if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
break;
udelay(1);
}
}
static void gfx_v9_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev)
{
uint32_t rlc_setting, data;
uint32_t data;
if (!adev->gfx.rlc.in_safe_mode)
return;
/* if RLC is not enabled, do nothing */
rlc_setting = RREG32_SOC15(GC, 0, mmRLC_CNTL);
if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
return;
if (adev->cg_flags &
(AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG)) {
/*
* Try to exit safe mode only if it is already in safe
* mode.
*/
data = RLC_SAFE_MODE__CMD_MASK;
WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
adev->gfx.rlc.in_safe_mode = false;
}
data = RLC_SAFE_MODE__CMD_MASK;
WREG32_SOC15(GC, 0, mmRLC_SAFE_MODE, data);
}
static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
bool enable)
{
gfx_v9_0_enter_rlc_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) {
gfx_v9_0_enable_gfx_cg_power_gating(adev, true);
@@ -3675,7 +3545,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
gfx_v9_0_enable_gfx_pipeline_powergating(adev, false);
}
gfx_v9_0_exit_rlc_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev,
@@ -3773,7 +3643,7 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
{
uint32_t data, def;
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
/* Enable 3D CGCG/CGLS */
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)) {
@@ -3813,7 +3683,7 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data);
}
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -3821,7 +3691,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
{
uint32_t def, data;
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
@@ -3861,7 +3731,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data);
}
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
@@ -3890,8 +3760,17 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
}
static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = {
.enter_safe_mode = gfx_v9_0_enter_rlc_safe_mode,
.exit_safe_mode = gfx_v9_0_exit_rlc_safe_mode
.is_rlc_enabled = gfx_v9_0_is_rlc_enabled,
.set_safe_mode = gfx_v9_0_set_safe_mode,
.unset_safe_mode = gfx_v9_0_unset_safe_mode,
.init = gfx_v9_0_rlc_init,
.get_csb_size = gfx_v9_0_get_csb_size,
.get_csb_buffer = gfx_v9_0_get_csb_buffer,
.get_cp_table_num = gfx_v9_0_cp_jump_table_num,
.resume = gfx_v9_0_rlc_resume,
.stop = gfx_v9_0_rlc_stop,
.reset = gfx_v9_0_rlc_reset,
.start = gfx_v9_0_rlc_start
};
static int gfx_v9_0_set_powergating_state(void *handle,
@@ -4072,9 +3951,11 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
}
static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 header, control = 0;
if (ib->flags & AMDGPU_IB_FLAG_CE)
@@ -4103,20 +3984,22 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
}
static void gfx_v9_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
amdgpu_ring_write(ring,
amdgpu_ring_write(ring,
#ifdef __BIG_ENDIAN
(2 << 0) |
(2 << 0) |
#endif
lower_32_bits(ib->gpu_addr));
amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
amdgpu_ring_write(ring, control);
lower_32_bits(ib->gpu_addr));
amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
amdgpu_ring_write(ring, control);
}
static void gfx_v9_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
@@ -4695,12 +4578,39 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev,
return 0;
}
static void gfx_v9_0_fault(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
u8 me_id, pipe_id, queue_id;
struct amdgpu_ring *ring;
int i;
me_id = (entry->ring_id & 0x0c) >> 2;
pipe_id = (entry->ring_id & 0x03) >> 0;
queue_id = (entry->ring_id & 0x70) >> 4;
switch (me_id) {
case 0:
drm_sched_fault(&adev->gfx.gfx_ring[0].sched);
break;
case 1:
case 2:
for (i = 0; i < adev->gfx.num_compute_rings; i++) {
ring = &adev->gfx.compute_ring[i];
if (ring->me == me_id && ring->pipe == pipe_id &&
ring->queue == queue_id)
drm_sched_fault(&ring->sched);
}
break;
}
}
static int gfx_v9_0_priv_reg_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal register access in command stream\n");
schedule_work(&adev->reset_work);
gfx_v9_0_fault(adev, entry);
return 0;
}
@@ -4709,7 +4619,7 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal instruction in command stream\n");
schedule_work(&adev->reset_work);
gfx_v9_0_fault(adev, entry);
return 0;
}
@@ -4836,10 +4746,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
2 + /* gfx_v9_0_ring_emit_vm_flush */
8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */
.emit_ib_size = 4, /* gfx_v9_0_ring_emit_ib_compute */
.emit_ib = gfx_v9_0_ring_emit_ib_compute,
.emit_fence = gfx_v9_0_ring_emit_fence_kiq,
.test_ring = gfx_v9_0_ring_test_ring,
.test_ib = gfx_v9_0_ring_test_ib,
.insert_nop = amdgpu_ring_insert_nop,
.pad_ib = amdgpu_ring_generic_pad_ib,
.emit_rreg = gfx_v9_0_ring_emit_rreg,

19
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
View File

@@ -35,20 +35,25 @@ u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev)
return (u64)RREG32_SOC15(GC, 0, mmMC_VM_FB_OFFSET) << 24;
}
static void gfxhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev)
void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo);
/* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */
int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32
- mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
lower_32_bits(value));
WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
offset * vmid, lower_32_bits(page_table_base));
WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
upper_32_bits(value));
WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
offset * vmid, upper_32_bits(page_table_base));
}
static void gfxhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev)
{
gfxhub_v1_0_init_gart_pt_regs(adev);
uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
gfxhub_v1_0_setup_vm_pt_regs(adev, 0, pt_base);
WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
(u32)(adev->gmc.gart_start >> 12));

2
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.h
View File

@@ -30,5 +30,7 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
bool value);
void gfxhub_v1_0_init(struct amdgpu_device *adev);
u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev);
void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base);
#endif

5
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
View File

@@ -358,7 +358,8 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
return 0;
}
static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid)
static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev,
uint32_t vmid, uint32_t flush_type)
{
WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
}
@@ -580,7 +581,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
else
gmc_v6_0_set_fault_enable_default(adev, true);
gmc_v6_0_flush_gpu_tlb(adev, 0);
gmc_v6_0_flush_gpu_tlb(adev, 0, 0);
dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
(unsigned long long)table_addr);

5
drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
View File

@@ -430,7 +430,8 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
*
* Flush the TLB for the requested page table (CIK).
*/
static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid)
static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev,
uint32_t vmid, uint32_t flush_type)
{
/* bits 0-15 are the VM contexts0-15 */
WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
@@ -698,7 +699,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev)
WREG32(mmCHUB_CONTROL, tmp);
}
gmc_v7_0_flush_gpu_tlb(adev, 0);
gmc_v7_0_flush_gpu_tlb(adev, 0, 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
(unsigned long long)table_addr);

4
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
View File

@@ -611,7 +611,7 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
* Flush the TLB for the requested page table (CIK).
*/
static void gmc_v8_0_flush_gpu_tlb(struct amdgpu_device *adev,
uint32_t vmid)
uint32_t vmid, uint32_t flush_type)
{
/* bits 0-15 are the VM contexts0-15 */
WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
@@ -920,7 +920,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev)
else
gmc_v8_0_set_fault_enable_default(adev, true);
gmc_v8_0_flush_gpu_tlb(adev, 0);
gmc_v8_0_flush_gpu_tlb(adev, 0, 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
(unsigned long long)table_addr);

110
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
View File

@@ -293,14 +293,14 @@ static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev)
adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs;
}
static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid)
static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
uint32_t flush_type)
{
u32 req = 0;
/* invalidate using legacy mode on vmid*/
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ,
PER_VMID_INVALIDATE_REQ, 1 << vmid);
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0);
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
req = REG_SET_FIELD(req, VM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
@@ -312,48 +312,6 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid)
return req;
}
static signed long amdgpu_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
uint32_t reg0, uint32_t reg1,
uint32_t ref, uint32_t mask)
{
signed long r, cnt = 0;
unsigned long flags;
uint32_t seq;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *ring = &kiq->ring;
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 0;
failed_kiq:
pr_err("failed to invalidate tlb with kiq\n");
return r;
}
/*
* GART
* VMID 0 is the physical GPU addresses as used by the kernel.
@@ -362,64 +320,47 @@ failed_kiq:
*/
/**
* gmc_v9_0_flush_gpu_tlb - gart tlb flush callback
* gmc_v9_0_flush_gpu_tlb - tlb flush with certain type
*
* @adev: amdgpu_device pointer
* @vmid: vm instance to flush
* @flush_type: the flush type
*
* Flush the TLB for the requested page table.
* Flush the TLB for the requested page table using certain type.
*/
static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev,
uint32_t vmid)
uint32_t vmid, uint32_t flush_type)
{
/* Use register 17 for GART */
const unsigned eng = 17;
unsigned i, j;
int r;
for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) {
struct amdgpu_vmhub *hub = &adev->vmhub[i];
u32 tmp = gmc_v9_0_get_invalidate_req(vmid);
u32 tmp = gmc_v9_0_get_invalidate_req(vmid, flush_type);
if (adev->gfx.kiq.ring.ready &&
(amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
!adev->in_gpu_reset) {
r = amdgpu_kiq_reg_write_reg_wait(adev, hub->vm_inv_eng0_req + eng,
hub->vm_inv_eng0_ack + eng, tmp, 1 << vmid);
if (!r)
continue;
if (i == AMDGPU_GFXHUB && !adev->in_gpu_reset &&
adev->gfx.kiq.ring.sched.ready &&
(amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
uint32_t req = hub->vm_inv_eng0_req + eng;
uint32_t ack = hub->vm_inv_eng0_ack + eng;
amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, tmp,
1 << vmid);
continue;
}
spin_lock(&adev->gmc.invalidate_lock);
WREG32_NO_KIQ(hub->vm_inv_eng0_req + eng, tmp);
/* Busy wait for ACK.*/
for (j = 0; j < 100; j++) {
tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_ack + eng);
tmp &= 1 << vmid;
if (tmp)
break;
cpu_relax();
}
if (j < 100) {
spin_unlock(&adev->gmc.invalidate_lock);
continue;
}
/* Wait for ACK with a delay.*/
for (j = 0; j < adev->usec_timeout; j++) {
tmp = RREG32_NO_KIQ(hub->vm_inv_eng0_ack + eng);
tmp &= 1 << vmid;
if (tmp)
if (tmp & (1 << vmid))
break;
udelay(1);
}
if (j < adev->usec_timeout) {
spin_unlock(&adev->gmc.invalidate_lock);
continue;
}
spin_unlock(&adev->gmc.invalidate_lock);
if (j < adev->usec_timeout)
continue;
DRM_ERROR("Timeout waiting for VM flush ACK!\n");
}
}
@@ -429,7 +370,7 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
{
struct amdgpu_device *adev = ring->adev;
struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub];
uint32_t req = gmc_v9_0_get_invalidate_req(vmid);
uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0);
unsigned eng = ring->vm_inv_eng;
amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid),
@@ -739,9 +680,8 @@ static int gmc_v9_0_late_init(void *handle)
unsigned vmhub = ring->funcs->vmhub;
ring->vm_inv_eng = vm_inv_eng[vmhub]++;
dev_info(adev->dev, "ring %u(%s) uses VM inv eng %u on hub %u\n",
ring->idx, ring->name, ring->vm_inv_eng,
ring->funcs->vmhub);
dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n",
ring->name, ring->vm_inv_eng, ring->funcs->vmhub);
}
/* Engine 16 is used for KFD and 17 for GART flushes */
@@ -1122,7 +1062,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
gfxhub_v1_0_set_fault_enable_default(adev, value);
mmhub_v1_0_set_fault_enable_default(adev, value);
gmc_v9_0_flush_gpu_tlb(adev, 0);
gmc_v9_0_flush_gpu_tlb(adev, 0, 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),

6
drivers/gpu/drm/amd/amdgpu/kv_dpm.c
View File

@@ -508,19 +508,19 @@ static int kv_enable_didt(struct amdgpu_device *adev, bool enable)
pi->caps_db_ramping ||
pi->caps_td_ramping ||
pi->caps_tcp_ramping) {
adev->gfx.rlc.funcs->enter_safe_mode(adev);
amdgpu_gfx_rlc_enter_safe_mode(adev);
if (enable) {
ret = kv_program_pt_config_registers(adev, didt_config_kv);
if (ret) {
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
return ret;
}
}
kv_do_enable_didt(adev, enable);
adev->gfx.rlc.funcs->exit_safe_mode(adev);
amdgpu_gfx_rlc_exit_safe_mode(adev);
}
return 0;

19
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
View File

@@ -52,20 +52,25 @@ u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev)
return base;
}
static void mmhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev)
void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo);
/* two registers distance between mmVM_CONTEXT0_* to mmVM_CONTEXT1_* */
int offset = mmVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32
- mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
lower_32_bits(value));
WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
offset * vmid, lower_32_bits(page_table_base));
WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
upper_32_bits(value));
WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
offset * vmid, upper_32_bits(page_table_base));
}
static void mmhub_v1_0_init_gart_aperture_regs(struct amdgpu_device *adev)
{
mmhub_v1_0_init_gart_pt_regs(adev);
uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
mmhub_v1_0_setup_vm_pt_regs(adev, 0, pt_base);
WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
(u32)(adev->gmc.gart_start >> 12));

2
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h
View File

@@ -34,5 +34,7 @@ int mmhub_v1_0_set_clockgating(struct amdgpu_device *adev,
void mmhub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags);
void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev,
bool enable);
void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base);
#endif

141
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c
View File

@@ -34,6 +34,7 @@
#include "nbio/nbio_7_4_offset.h"
MODULE_FIRMWARE("amdgpu/vega20_sos.bin");
MODULE_FIRMWARE("amdgpu/vega20_ta.bin");
/* address block */
#define smnMP1_FIRMWARE_FLAGS 0x3010024
@@ -98,7 +99,8 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
const char *chip_name;
char fw_name[30];
int err = 0;
const struct psp_firmware_header_v1_0 *hdr;
const struct psp_firmware_header_v1_0 *sos_hdr;
const struct ta_firmware_header_v1_0 *ta_hdr;
DRM_DEBUG("\n");
@@ -119,16 +121,32 @@ static int psp_v11_0_init_microcode(struct psp_context *psp)
if (err)
goto out;
hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.sos_fw->data;
adev->psp.sos_fw_version = le32_to_cpu(hdr->header.ucode_version);
adev->psp.sos_feature_version = le32_to_cpu(hdr->ucode_feature_version);
adev->psp.sos_bin_size = le32_to_cpu(hdr->sos_size_bytes);
adev->psp.sys_bin_size = le32_to_cpu(hdr->header.ucode_size_bytes) -
le32_to_cpu(hdr->sos_size_bytes);
adev->psp.sys_start_addr = (uint8_t *)hdr +
le32_to_cpu(hdr->header.ucode_array_offset_bytes);
sos_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.sos_fw->data;
adev->psp.sos_fw_version = le32_to_cpu(sos_hdr->header.ucode_version);
adev->psp.sos_feature_version = le32_to_cpu(sos_hdr->ucode_feature_version);
adev->psp.sos_bin_size = le32_to_cpu(sos_hdr->sos_size_bytes);
adev->psp.sys_bin_size = le32_to_cpu(sos_hdr->header.ucode_size_bytes) -
le32_to_cpu(sos_hdr->sos_size_bytes);
adev->psp.sys_start_addr = (uint8_t *)sos_hdr +
le32_to_cpu(sos_hdr->header.ucode_array_offset_bytes);
adev->psp.sos_start_addr = (uint8_t *)adev->psp.sys_start_addr +
le32_to_cpu(hdr->sos_offset_bytes);
le32_to_cpu(sos_hdr->sos_offset_bytes);
snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ta.bin", chip_name);
err = request_firmware(&adev->psp.ta_fw, fw_name, adev->dev);
if (err)
goto out;
err = amdgpu_ucode_validate(adev->psp.ta_fw);
if (err)
goto out;
ta_hdr = (const struct ta_firmware_header_v1_0 *)adev->psp.ta_fw->data;
adev->psp.ta_xgmi_ucode_version = le32_to_cpu(ta_hdr->ta_xgmi_ucode_version);
adev->psp.ta_xgmi_ucode_size = le32_to_cpu(ta_hdr->ta_xgmi_size_bytes);
adev->psp.ta_xgmi_start_addr = (uint8_t *)ta_hdr +
le32_to_cpu(ta_hdr->header.ucode_array_offset_bytes);
return 0;
out:
if (err) {
@@ -167,7 +185,7 @@ static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp)
/* Copy PSP System Driver binary to memory */
memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size);
/* Provide the sys driver to bootrom */
/* Provide the sys driver to bootloader */
WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36,
(uint32_t)(psp->fw_pri_mc_addr >> 20));
psp_gfxdrv_command_reg = 1 << 16;
@@ -208,7 +226,7 @@ static int psp_v11_0_bootloader_load_sos(struct psp_context *psp)
/* Copy Secure OS binary to PSP memory */
memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size);
/* Provide the PSP secure OS to bootrom */
/* Provide the PSP secure OS to bootloader */
WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36,
(uint32_t)(psp->fw_pri_mc_addr >> 20));
psp_gfxdrv_command_reg = 2 << 16;
@@ -552,24 +570,110 @@ static int psp_v11_0_mode1_reset(struct psp_context *psp)
static int psp_v11_0_xgmi_get_topology_info(struct psp_context *psp,
int number_devices, struct psp_xgmi_topology_info *topology)
{
struct ta_xgmi_shared_memory *xgmi_cmd;
struct ta_xgmi_cmd_get_topology_info_input *topology_info_input;
struct ta_xgmi_cmd_get_topology_info_output *topology_info_output;
int i;
int ret;
if (!topology || topology->num_nodes > TA_XGMI__MAX_CONNECTED_NODES)
return -EINVAL;
xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf;
memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
/* Fill in the shared memory with topology information as input */
topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info;
xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO;
topology_info_input->num_nodes = number_devices;
for (i = 0; i < topology_info_input->num_nodes; i++) {
topology_info_input->nodes[i].node_id = topology->nodes[i].node_id;
topology_info_input->nodes[i].num_hops = topology->nodes[i].num_hops;
topology_info_input->nodes[i].is_sharing_enabled = topology->nodes[i].is_sharing_enabled;
topology_info_input->nodes[i].sdma_engine = topology->nodes[i].sdma_engine;
}
/* Invoke xgmi ta to get the topology information */
ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO);
if (ret)
return ret;
/* Read the output topology information from the shared memory */
topology_info_output = &xgmi_cmd->xgmi_out_message.get_topology_info;
topology->num_nodes = xgmi_cmd->xgmi_out_message.get_topology_info.num_nodes;
for (i = 0; i < topology->num_nodes; i++) {
topology->nodes[i].node_id = topology_info_output->nodes[i].node_id;
topology->nodes[i].num_hops = topology_info_output->nodes[i].num_hops;
topology->nodes[i].is_sharing_enabled = topology_info_output->nodes[i].is_sharing_enabled;
topology->nodes[i].sdma_engine = topology_info_output->nodes[i].sdma_engine;
}
return 0;
}
static int psp_v11_0_xgmi_set_topology_info(struct psp_context *psp,
int number_devices, struct psp_xgmi_topology_info *topology)
{
return 0;
struct ta_xgmi_shared_memory *xgmi_cmd;
struct ta_xgmi_cmd_get_topology_info_input *topology_info_input;
int i;
if (!topology || topology->num_nodes > TA_XGMI__MAX_CONNECTED_NODES)
return -EINVAL;
xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf;
memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info;
xgmi_cmd->cmd_id = TA_COMMAND_XGMI__SET_TOPOLOGY_INFO;
topology_info_input->num_nodes = number_devices;
for (i = 0; i < topology_info_input->num_nodes; i++) {
topology_info_input->nodes[i].node_id = topology->nodes[i].node_id;
topology_info_input->nodes[i].num_hops = topology->nodes[i].num_hops;
topology_info_input->nodes[i].is_sharing_enabled = topology->nodes[i].is_sharing_enabled;
topology_info_input->nodes[i].sdma_engine = topology->nodes[i].sdma_engine;
}
/* Invoke xgmi ta to set topology information */
return psp_xgmi_invoke(psp, TA_COMMAND_XGMI__SET_TOPOLOGY_INFO);
}
static u64 psp_v11_0_xgmi_get_hive_id(struct psp_context *psp)
{
u64 hive_id = 0;
struct ta_xgmi_shared_memory *xgmi_cmd;
int ret;
/* Remove me when we can get correct hive_id through PSP */
if (psp->adev->gmc.xgmi.num_physical_nodes)
hive_id = 0x123456789abcdef;
xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf;
memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
return hive_id;
xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_HIVE_ID;
/* Invoke xgmi ta to get hive id */
ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
if (ret)
return 0;
else
return xgmi_cmd->xgmi_out_message.get_hive_id.hive_id;
}
static u64 psp_v11_0_xgmi_get_node_id(struct psp_context *psp)
{
struct ta_xgmi_shared_memory *xgmi_cmd;
int ret;
xgmi_cmd = (struct ta_xgmi_shared_memory*)psp->xgmi_context.xgmi_shared_buf;
memset(xgmi_cmd, 0, sizeof(struct ta_xgmi_shared_memory));
xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_NODE_ID;
/* Invoke xgmi ta to get the node id */
ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
if (ret)
return 0;
else
return xgmi_cmd->xgmi_out_message.get_node_id.node_id;
}
static const struct psp_funcs psp_v11_0_funcs = {
@@ -587,6 +691,7 @@ static const struct psp_funcs psp_v11_0_funcs = {
.xgmi_get_topology_info = psp_v11_0_xgmi_get_topology_info,
.xgmi_set_topology_info = psp_v11_0_xgmi_set_topology_info,
.xgmi_get_hive_id = psp_v11_0_xgmi_get_hive_id,
.xgmi_get_node_id = psp_v11_0_xgmi_get_node_id,
};
void psp_v11_0_set_psp_funcs(struct psp_context *psp)

4
drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
View File

@@ -194,7 +194,7 @@ static int psp_v3_1_bootloader_load_sysdrv(struct psp_context *psp)
/* Copy PSP System Driver binary to memory */
memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size);
/* Provide the sys driver to bootrom */
/* Provide the sys driver to bootloader */
WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36,
(uint32_t)(psp->fw_pri_mc_addr >> 20));
psp_gfxdrv_command_reg = 1 << 16;
@@ -254,7 +254,7 @@ static int psp_v3_1_bootloader_load_sos(struct psp_context *psp)
/* Copy Secure OS binary to PSP memory */
memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size);
/* Provide the PSP secure OS to bootrom */
/* Provide the PSP secure OS to bootloader */
WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36,
(uint32_t)(psp->fw_pri_mc_addr >> 20));
psp_gfxdrv_command_reg = 2 << 16;

69
drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
View File

@@ -225,7 +225,7 @@ static void sdma_v2_4_ring_set_wptr(struct amdgpu_ring *ring)
static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
int i;
for (i = 0; i < count; i++)
@@ -245,9 +245,12 @@ static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
* Schedule an IB in the DMA ring (VI).
*/
static void sdma_v2_4_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
/* IB packet must end on a 8 DW boundary */
sdma_v2_4_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8);
@@ -349,8 +352,8 @@ static void sdma_v2_4_gfx_stop(struct amdgpu_device *adev)
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0);
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
}
sdma0->ready = false;
sdma1->ready = false;
sdma0->sched.ready = false;
sdma1->sched.ready = false;
}
/**
@@ -471,17 +474,15 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
/* enable DMA IBs */
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
ring->ready = true;
ring->sched.ready = true;
}
sdma_v2_4_enable(adev, true);
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
@@ -550,21 +551,16 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
if (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_device_wb_free(adev, index);
return r;
}
if (r)
goto error_free_wb;
amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR));
@@ -581,15 +577,11 @@ static int sdma_v2_4_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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_device_wb_free(adev, index);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_wb:
amdgpu_device_wb_free(adev, index);
return r;
}
@@ -612,20 +604,16 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err0;
}
ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR);
@@ -644,21 +632,16 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
else
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
@@ -760,7 +743,7 @@ static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe,
*/
static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
u32 pad_count;
int i;
@@ -1105,8 +1088,14 @@ static int sdma_v2_4_process_illegal_inst_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
u8 instance_id, queue_id;
DRM_ERROR("Illegal instruction in SDMA command stream\n");
schedule_work(&adev->reset_work);
instance_id = (entry->ring_id & 0x3) >> 0;
queue_id = (entry->ring_id & 0xc) >> 2;
if (instance_id <= 1 && queue_id == 0)
drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched);
return 0;
}

69
drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
View File

@@ -399,7 +399,7 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring)
static void sdma_v3_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
int i;
for (i = 0; i < count; i++)
@@ -419,9 +419,12 @@ static void sdma_v3_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
* Schedule an IB in the DMA ring (VI).
*/
static void sdma_v3_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
/* IB packet must end on a 8 DW boundary */
sdma_v3_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8);
@@ -523,8 +526,8 @@ static void sdma_v3_0_gfx_stop(struct amdgpu_device *adev)
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0);
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
}
sdma0->ready = false;
sdma1->ready = false;
sdma0->sched.ready = false;
sdma1->sched.ready = false;
}
/**
@@ -739,7 +742,7 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
/* enable DMA IBs */
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
ring->ready = true;
ring->sched.ready = true;
}
/* unhalt the MEs */
@@ -749,11 +752,9 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
@@ -822,21 +823,16 @@ static int sdma_v3_0_ring_test_ring(struct amdgpu_ring *ring)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
if (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_device_wb_free(adev, index);
return r;
}
if (r)
goto error_free_wb;
amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR));
@@ -853,15 +849,11 @@ static int sdma_v3_0_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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_device_wb_free(adev, index);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_wb:
amdgpu_device_wb_free(adev, index);
return r;
}
@@ -884,20 +876,16 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err0;
}
ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR);
@@ -916,21 +904,16 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
else
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
dma_fence_put(f);
@@ -1031,7 +1014,7 @@ static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib, uint64_t pe,
*/
static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
u32 pad_count;
int i;
@@ -1440,8 +1423,14 @@ static int sdma_v3_0_process_illegal_inst_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
u8 instance_id, queue_id;
DRM_ERROR("Illegal instruction in SDMA command stream\n");
schedule_work(&adev->reset_work);
instance_id = (entry->ring_id & 0x3) >> 0;
queue_id = (entry->ring_id & 0xc) >> 2;
if (instance_id <= 1 && queue_id == 0)
drm_sched_fault(&adev->sdma.instance[instance_id].ring.sched);
return 0;
}

702
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
View File

@@ -54,6 +54,11 @@ MODULE_FIRMWARE("amdgpu/raven2_sdma.bin");
#define SDMA0_POWER_CNTL__ON_OFF_CONDITION_HOLD_TIME_MASK 0x000000F8L
#define SDMA0_POWER_CNTL__ON_OFF_STATUS_DURATION_TIME_MASK 0xFC000000L
#define WREG32_SDMA(instance, offset, value) \
WREG32(sdma_v4_0_get_reg_offset(adev, (instance), (offset)), value)
#define RREG32_SDMA(instance, offset) \
RREG32(sdma_v4_0_get_reg_offset(adev, (instance), (offset)))
static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev);
static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev);
static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev);
@@ -367,16 +372,11 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring)
wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs]));
DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr);
} else {
u32 lowbit, highbit;
lowbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR)) >> 2;
highbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI)) >> 2;
DRM_DEBUG("wptr [%i]high== 0x%08x low==0x%08x\n",
ring->me, highbit, lowbit);
wptr = highbit;
wptr = RREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR_HI);
wptr = wptr << 32;
wptr |= lowbit;
wptr |= RREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR);
DRM_DEBUG("wptr before shift [%i] wptr == 0x%016llx\n",
ring->me, wptr);
}
return wptr >> 2;
@@ -417,14 +417,67 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring)
lower_32_bits(ring->wptr << 2),
ring->me,
upper_32_bits(ring->wptr << 2));
WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2));
WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2));
WREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR,
lower_32_bits(ring->wptr << 2));
WREG32_SDMA(ring->me, mmSDMA0_GFX_RB_WPTR_HI,
upper_32_bits(ring->wptr << 2));
}
}
/**
* sdma_v4_0_page_ring_get_wptr - get the current write pointer
*
* @ring: amdgpu ring pointer
*
* Get the current wptr from the hardware (VEGA10+).
*/
static uint64_t sdma_v4_0_page_ring_get_wptr(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
u64 wptr;
if (ring->use_doorbell) {
/* XXX check if swapping is necessary on BE */
wptr = READ_ONCE(*((u64 *)&adev->wb.wb[ring->wptr_offs]));
} else {
wptr = RREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR_HI);
wptr = wptr << 32;
wptr |= RREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR);
}
return wptr >> 2;
}
/**
* sdma_v4_0_ring_set_wptr - commit the write pointer
*
* @ring: amdgpu ring pointer
*
* Write the wptr back to the hardware (VEGA10+).
*/
static void sdma_v4_0_page_ring_set_wptr(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
if (ring->use_doorbell) {
u64 *wb = (u64 *)&adev->wb.wb[ring->wptr_offs];
/* XXX check if swapping is necessary on BE */
WRITE_ONCE(*wb, (ring->wptr << 2));
WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
} else {
uint64_t wptr = ring->wptr << 2;
WREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR,
lower_32_bits(wptr));
WREG32_SDMA(ring->me, mmSDMA0_PAGE_RB_WPTR_HI,
upper_32_bits(wptr));
}
}
static void sdma_v4_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
int i;
for (i = 0; i < count; i++)
@@ -444,9 +497,12 @@ static void sdma_v4_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
* Schedule an IB in the DMA ring (VEGA10).
*/
static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
/* IB packet must end on a 8 DW boundary */
sdma_v4_0_ring_insert_nop(ring, (10 - (lower_32_bits(ring->wptr) & 7)) % 8);
@@ -568,16 +624,16 @@ static void sdma_v4_0_gfx_stop(struct amdgpu_device *adev)
amdgpu_ttm_set_buffer_funcs_status(adev, false);
for (i = 0; i < adev->sdma.num_instances; i++) {
rb_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
rb_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
ib_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl);
ib_cntl = RREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL);
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl);
}
sdma0->ready = false;
sdma1->ready = false;
sdma0->sched.ready = false;
sdma1->sched.ready = false;
}
/**
@@ -592,6 +648,39 @@ static void sdma_v4_0_rlc_stop(struct amdgpu_device *adev)
/* XXX todo */
}
/**
* sdma_v4_0_page_stop - stop the page async dma engines
*
* @adev: amdgpu_device pointer
*
* Stop the page async dma ring buffers (VEGA10).
*/
static void sdma_v4_0_page_stop(struct amdgpu_device *adev)
{
struct amdgpu_ring *sdma0 = &adev->sdma.instance[0].page;
struct amdgpu_ring *sdma1 = &adev->sdma.instance[1].page;
u32 rb_cntl, ib_cntl;
int i;
if ((adev->mman.buffer_funcs_ring == sdma0) ||
(adev->mman.buffer_funcs_ring == sdma1))
amdgpu_ttm_set_buffer_funcs_status(adev, false);
for (i = 0; i < adev->sdma.num_instances; i++) {
rb_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL,
RB_ENABLE, 0);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl);
ib_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL);
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL,
IB_ENABLE, 0);
WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl);
}
sdma0->sched.ready = false;
sdma1->sched.ready = false;
}
/**
* sdma_v_0_ctx_switch_enable - stop the async dma engines context switch
*
@@ -630,18 +719,15 @@ static void sdma_v4_0_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
}
for (i = 0; i < adev->sdma.num_instances; i++) {
f32_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL));
f32_cntl = RREG32_SDMA(i, mmSDMA0_CNTL);
f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_CNTL,
AUTO_CTXSW_ENABLE, enable ? 1 : 0);
if (enable && amdgpu_sdma_phase_quantum) {
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE0_QUANTUM),
phase_quantum);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE1_QUANTUM),
phase_quantum);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_PHASE2_QUANTUM),
phase_quantum);
WREG32_SDMA(i, mmSDMA0_PHASE0_QUANTUM, phase_quantum);
WREG32_SDMA(i, mmSDMA0_PHASE1_QUANTUM, phase_quantum);
WREG32_SDMA(i, mmSDMA0_PHASE2_QUANTUM, phase_quantum);
}
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL), f32_cntl);
WREG32_SDMA(i, mmSDMA0_CNTL, f32_cntl);
}
}
@@ -662,156 +748,217 @@ static void sdma_v4_0_enable(struct amdgpu_device *adev, bool enable)
if (enable == false) {
sdma_v4_0_gfx_stop(adev);
sdma_v4_0_rlc_stop(adev);
if (adev->sdma.has_page_queue)
sdma_v4_0_page_stop(adev);
}
for (i = 0; i < adev->sdma.num_instances; i++) {
f32_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
f32_cntl = RREG32_SDMA(i, mmSDMA0_F32_CNTL);
f32_cntl = REG_SET_FIELD(f32_cntl, SDMA0_F32_CNTL, HALT, enable ? 0 : 1);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), f32_cntl);
WREG32_SDMA(i, mmSDMA0_F32_CNTL, f32_cntl);
}
}
/**
* sdma_v4_0_rb_cntl - get parameters for rb_cntl
*/
static uint32_t sdma_v4_0_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl)
{
/* Set ring buffer size in dwords */
uint32_t rb_bufsz = order_base_2(ring->ring_size / 4);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
#ifdef __BIG_ENDIAN
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
RPTR_WRITEBACK_SWAP_ENABLE, 1);
#endif
return rb_cntl;
}
/**
* sdma_v4_0_gfx_resume - setup and start the async dma engines
*
* @adev: amdgpu_device pointer
* @i: instance to resume
*
* Set up the gfx DMA ring buffers and enable them (VEGA10).
* Returns 0 for success, error for failure.
*/
static int sdma_v4_0_gfx_resume(struct amdgpu_device *adev)
static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i)
{
struct amdgpu_ring *ring;
struct amdgpu_ring *ring = &adev->sdma.instance[i].ring;
u32 rb_cntl, ib_cntl, wptr_poll_cntl;
u32 rb_bufsz;
u32 wb_offset;
u32 doorbell;
u32 doorbell_offset;
u32 temp;
u64 wptr_gpu_addr;
int i, r;
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
wb_offset = (ring->rptr_offs * 4);
wb_offset = (ring->rptr_offs * 4);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL), 0);
rb_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL);
rb_cntl = sdma_v4_0_rb_cntl(ring, rb_cntl);
WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl);
/* Set ring buffer size in dwords */
rb_bufsz = order_base_2(ring->ring_size / 4);
rb_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL));
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SIZE, rb_bufsz);
#ifdef __BIG_ENDIAN
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_SWAP_ENABLE, 1);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
RPTR_WRITEBACK_SWAP_ENABLE, 1);
#endif
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
/* Initialize the ring buffer's read and write pointers */
WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR, 0);
WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_HI, 0);
WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR, 0);
WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_HI, 0);
/* Initialize the ring buffer's read and write pointers */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR), 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_HI), 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), 0);
/* set the wb address whether it's enabled or not */
WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_ADDR_HI,
upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
WREG32_SDMA(i, mmSDMA0_GFX_RB_RPTR_ADDR_LO,
lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC);
/* set the wb address whether it's enabled or not */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_HI),
upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_RPTR_ADDR_LO),
lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL,
RPTR_WRITEBACK_ENABLE, 1);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RPTR_WRITEBACK_ENABLE, 1);
WREG32_SDMA(i, mmSDMA0_GFX_RB_BASE, ring->gpu_addr >> 8);
WREG32_SDMA(i, mmSDMA0_GFX_RB_BASE_HI, ring->gpu_addr >> 40);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE), ring->gpu_addr >> 8);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_BASE_HI), ring->gpu_addr >> 40);
ring->wptr = 0;
ring->wptr = 0;
/* before programing wptr to a less value, need set minor_ptr_update first */
WREG32_SDMA(i, mmSDMA0_GFX_MINOR_PTR_UPDATE, 1);
/* before programing wptr to a less value, need set minor_ptr_update first */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 1);
doorbell = RREG32_SDMA(i, mmSDMA0_GFX_DOORBELL);
doorbell_offset = RREG32_SDMA(i, mmSDMA0_GFX_DOORBELL_OFFSET);
if (!amdgpu_sriov_vf(adev)) { /* only bare-metal use register write for wptr */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr) << 2);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr) << 2);
}
doorbell = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL));
doorbell_offset = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET));
if (ring->use_doorbell) {
doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 1);
doorbell_offset = REG_SET_FIELD(doorbell_offset, SDMA0_GFX_DOORBELL_OFFSET,
doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE,
ring->use_doorbell);
doorbell_offset = REG_SET_FIELD(doorbell_offset,
SDMA0_GFX_DOORBELL_OFFSET,
OFFSET, ring->doorbell_index);
} else {
doorbell = REG_SET_FIELD(doorbell, SDMA0_GFX_DOORBELL, ENABLE, 0);
}
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL), doorbell);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_DOORBELL_OFFSET), doorbell_offset);
adev->nbio_funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
ring->doorbell_index);
WREG32_SDMA(i, mmSDMA0_GFX_DOORBELL, doorbell);
WREG32_SDMA(i, mmSDMA0_GFX_DOORBELL_OFFSET, doorbell_offset);
adev->nbio_funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
ring->doorbell_index);
if (amdgpu_sriov_vf(adev))
sdma_v4_0_ring_set_wptr(ring);
sdma_v4_0_ring_set_wptr(ring);
/* set minor_ptr_update to 0 after wptr programed */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_MINOR_PTR_UPDATE), 0);
/* set minor_ptr_update to 0 after wptr programed */
WREG32_SDMA(i, mmSDMA0_GFX_MINOR_PTR_UPDATE, 0);
/* set utc l1 enable flag always to 1 */
temp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL));
temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_CNTL), temp);
/* setup the wptr shadow polling */
wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO,
lower_32_bits(wptr_gpu_addr));
WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI,
upper_32_bits(wptr_gpu_addr));
wptr_poll_cntl = RREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL);
wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
SDMA0_GFX_RB_WPTR_POLL_CNTL,
F32_POLL_ENABLE, amdgpu_sriov_vf(adev));
WREG32_SDMA(i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL, wptr_poll_cntl);
if (!amdgpu_sriov_vf(adev)) {
/* unhalt engine */
temp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL));
temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_F32_CNTL), temp);
}
/* enable DMA RB */
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
WREG32_SDMA(i, mmSDMA0_GFX_RB_CNTL, rb_cntl);
/* setup the wptr shadow polling */
wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO),
lower_32_bits(wptr_gpu_addr));
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI),
upper_32_bits(wptr_gpu_addr));
wptr_poll_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL));
if (amdgpu_sriov_vf(adev))
wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1);
else
wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), wptr_poll_cntl);
/* enable DMA RB */
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1);
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_RB_CNTL), rb_cntl);
ib_cntl = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL));
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
ib_cntl = RREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL);
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_ENABLE, 1);
#ifdef __BIG_ENDIAN
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_GFX_IB_CNTL, IB_SWAP_ENABLE, 1);
#endif
/* enable DMA IBs */
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
/* enable DMA IBs */
WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl);
ring->ready = true;
ring->sched.ready = true;
}
if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
sdma_v4_0_ctx_switch_enable(adev, true);
sdma_v4_0_enable(adev, true);
}
/**
* sdma_v4_0_page_resume - setup and start the async dma engines
*
* @adev: amdgpu_device pointer
* @i: instance to resume
*
* Set up the page DMA ring buffers and enable them (VEGA10).
* Returns 0 for success, error for failure.
*/
static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i)
{
struct amdgpu_ring *ring = &adev->sdma.instance[i].page;
u32 rb_cntl, ib_cntl, wptr_poll_cntl;
u32 wb_offset;
u32 doorbell;
u32 doorbell_offset;
u64 wptr_gpu_addr;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
return r;
}
wb_offset = (ring->rptr_offs * 4);
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
rb_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL);
rb_cntl = sdma_v4_0_rb_cntl(ring, rb_cntl);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl);
}
/* Initialize the ring buffer's read and write pointers */
WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR, 0);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_HI, 0);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR, 0);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_HI, 0);
return 0;
/* set the wb address whether it's enabled or not */
WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_ADDR_HI,
upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_RPTR_ADDR_LO,
lower_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC);
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL,
RPTR_WRITEBACK_ENABLE, 1);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_BASE, ring->gpu_addr >> 8);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_BASE_HI, ring->gpu_addr >> 40);
ring->wptr = 0;
/* before programing wptr to a less value, need set minor_ptr_update first */
WREG32_SDMA(i, mmSDMA0_PAGE_MINOR_PTR_UPDATE, 1);
doorbell = RREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL);
doorbell_offset = RREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL_OFFSET);
doorbell = REG_SET_FIELD(doorbell, SDMA0_PAGE_DOORBELL, ENABLE,
ring->use_doorbell);
doorbell_offset = REG_SET_FIELD(doorbell_offset,
SDMA0_PAGE_DOORBELL_OFFSET,
OFFSET, ring->doorbell_index);
WREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL, doorbell);
WREG32_SDMA(i, mmSDMA0_PAGE_DOORBELL_OFFSET, doorbell_offset);
/* TODO: enable doorbell support */
/*adev->nbio_funcs->sdma_doorbell_range(adev, i, ring->use_doorbell,
ring->doorbell_index);*/
sdma_v4_0_ring_set_wptr(ring);
/* set minor_ptr_update to 0 after wptr programed */
WREG32_SDMA(i, mmSDMA0_PAGE_MINOR_PTR_UPDATE, 0);
/* setup the wptr shadow polling */
wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_ADDR_LO,
lower_32_bits(wptr_gpu_addr));
WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_ADDR_HI,
upper_32_bits(wptr_gpu_addr));
wptr_poll_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL);
wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl,
SDMA0_PAGE_RB_WPTR_POLL_CNTL,
F32_POLL_ENABLE, amdgpu_sriov_vf(adev));
WREG32_SDMA(i, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL, wptr_poll_cntl);
/* enable DMA RB */
rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_PAGE_RB_CNTL, RB_ENABLE, 1);
WREG32_SDMA(i, mmSDMA0_PAGE_RB_CNTL, rb_cntl);
ib_cntl = RREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL);
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL, IB_ENABLE, 1);
#ifdef __BIG_ENDIAN
ib_cntl = REG_SET_FIELD(ib_cntl, SDMA0_PAGE_IB_CNTL, IB_SWAP_ENABLE, 1);
#endif
/* enable DMA IBs */
WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl);
ring->sched.ready = true;
}
static void
@@ -922,12 +1069,14 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev)
(adev->sdma.instance[i].fw->data +
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_ADDR), 0);
WREG32_SDMA(i, mmSDMA0_UCODE_ADDR, 0);
for (j = 0; j < fw_size; j++)
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_DATA), le32_to_cpup(fw_data++));
WREG32_SDMA(i, mmSDMA0_UCODE_DATA,
le32_to_cpup(fw_data++));
WREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_UCODE_ADDR), adev->sdma.instance[i].fw_version);
WREG32_SDMA(i, mmSDMA0_UCODE_ADDR,
adev->sdma.instance[i].fw_version);
}
return 0;
@@ -943,33 +1092,78 @@ static int sdma_v4_0_load_microcode(struct amdgpu_device *adev)
*/
static int sdma_v4_0_start(struct amdgpu_device *adev)
{
int r = 0;
struct amdgpu_ring *ring;
int i, r;
if (amdgpu_sriov_vf(adev)) {
sdma_v4_0_ctx_switch_enable(adev, false);
sdma_v4_0_enable(adev, false);
} else {
/* set RB registers */
r = sdma_v4_0_gfx_resume(adev);
return r;
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
r = sdma_v4_0_load_microcode(adev);
if (r)
return r;
}
/* unhalt the MEs */
sdma_v4_0_enable(adev, true);
/* enable sdma ring preemption */
sdma_v4_0_ctx_switch_enable(adev, true);
}
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
r = sdma_v4_0_load_microcode(adev);
/* start the gfx rings and rlc compute queues */
for (i = 0; i < adev->sdma.num_instances; i++) {
uint32_t temp;
WREG32_SDMA(i, mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL, 0);
sdma_v4_0_gfx_resume(adev, i);
if (adev->sdma.has_page_queue)
sdma_v4_0_page_resume(adev, i);
/* set utc l1 enable flag always to 1 */
temp = RREG32_SDMA(i, mmSDMA0_CNTL);
temp = REG_SET_FIELD(temp, SDMA0_CNTL, UTC_L1_ENABLE, 1);
WREG32_SDMA(i, mmSDMA0_CNTL, temp);
if (!amdgpu_sriov_vf(adev)) {
/* unhalt engine */
temp = RREG32_SDMA(i, mmSDMA0_F32_CNTL);
temp = REG_SET_FIELD(temp, SDMA0_F32_CNTL, HALT, 0);
WREG32_SDMA(i, mmSDMA0_F32_CNTL, temp);
}
}
if (amdgpu_sriov_vf(adev)) {
sdma_v4_0_ctx_switch_enable(adev, true);
sdma_v4_0_enable(adev, true);
} else {
r = sdma_v4_0_rlc_resume(adev);
if (r)
return r;
}
/* unhalt the MEs */
sdma_v4_0_enable(adev, true);
/* enable sdma ring preemption */
sdma_v4_0_ctx_switch_enable(adev, true);
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
/* start the gfx rings and rlc compute queues */
r = sdma_v4_0_gfx_resume(adev);
if (r)
return r;
r = sdma_v4_0_rlc_resume(adev);
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
if (adev->sdma.has_page_queue) {
struct amdgpu_ring *page = &adev->sdma.instance[i].page;
r = amdgpu_ring_test_helper(page);
if (r)
return r;
if (adev->mman.buffer_funcs_ring == page)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
}
return r;
}
@@ -993,21 +1187,16 @@ static int sdma_v4_0_ring_test_ring(struct amdgpu_ring *ring)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
if (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_device_wb_free(adev, index);
return r;
}
if (r)
goto error_free_wb;
amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR));
@@ -1024,15 +1213,11 @@ static int sdma_v4_0_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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_device_wb_free(adev, index);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_wb:
amdgpu_device_wb_free(adev, index);
return r;
}
@@ -1055,20 +1240,16 @@ static int sdma_v4_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err0;
}
ib.ptr[0] = SDMA_PKT_HEADER_OP(SDMA_OP_WRITE) |
SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_WRITE_LINEAR);
@@ -1087,21 +1268,17 @@ static int sdma_v4_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
else
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
dma_fence_put(f);
@@ -1206,7 +1383,7 @@ static void sdma_v4_0_vm_set_pte_pde(struct amdgpu_ib *ib,
*/
static void sdma_v4_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
{
struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
struct amdgpu_sdma_instance *sdma = amdgpu_sdma_get_instance_from_ring(ring);
u32 pad_count;
int i;
@@ -1276,10 +1453,18 @@ static int sdma_v4_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
if (adev->asic_type == CHIP_RAVEN)
if (adev->asic_type == CHIP_RAVEN) {
adev->sdma.num_instances = 1;
else
adev->sdma.has_page_queue = false;
} else {
adev->sdma.num_instances = 2;
/* TODO: Page queue breaks driver reload under SRIOV */
if ((adev->asic_type == CHIP_VEGA10) && amdgpu_sriov_vf((adev)))
adev->sdma.has_page_queue = false;
else if (adev->asic_type != CHIP_VEGA20 &&
adev->asic_type != CHIP_VEGA12)
adev->sdma.has_page_queue = true;
}
sdma_v4_0_set_ring_funcs(adev);
sdma_v4_0_set_buffer_funcs(adev);
@@ -1340,6 +1525,21 @@ static int sdma_v4_0_sw_init(void *handle)
AMDGPU_SDMA_IRQ_TRAP1);
if (r)
return r;
if (adev->sdma.has_page_queue) {
ring = &adev->sdma.instance[i].page;
ring->ring_obj = NULL;
ring->use_doorbell = false;
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
&adev->sdma.trap_irq,
(i == 0) ?
AMDGPU_SDMA_IRQ_TRAP0 :
AMDGPU_SDMA_IRQ_TRAP1);
if (r)
return r;
}
}
return r;
@@ -1350,8 +1550,11 @@ static int sdma_v4_0_sw_fini(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
int i;
for (i = 0; i < adev->sdma.num_instances; i++)
for (i = 0; i < adev->sdma.num_instances; i++) {
amdgpu_ring_fini(&adev->sdma.instance[i].ring);
if (adev->sdma.has_page_queue)
amdgpu_ring_fini(&adev->sdma.instance[i].page);
}
for (i = 0; i < adev->sdma.num_instances; i++) {
release_firmware(adev->sdma.instance[i].fw);
@@ -1414,7 +1617,7 @@ static bool sdma_v4_0_is_idle(void *handle)
u32 i;
for (i = 0; i < adev->sdma.num_instances; i++) {
u32 tmp = RREG32(sdma_v4_0_get_reg_offset(adev, i, mmSDMA0_STATUS_REG));
u32 tmp = RREG32_SDMA(i, mmSDMA0_STATUS_REG);
if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK))
return false;
@@ -1430,8 +1633,8 @@ static int sdma_v4_0_wait_for_idle(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
for (i = 0; i < adev->usec_timeout; i++) {
sdma0 = RREG32(sdma_v4_0_get_reg_offset(adev, 0, mmSDMA0_STATUS_REG));
sdma1 = RREG32(sdma_v4_0_get_reg_offset(adev, 1, mmSDMA0_STATUS_REG));
sdma0 = RREG32_SDMA(0, mmSDMA0_STATUS_REG);
sdma1 = RREG32_SDMA(1, mmSDMA0_STATUS_REG);
if (sdma0 & sdma1 & SDMA0_STATUS_REG__IDLE_MASK)
return 0;
@@ -1452,16 +1655,13 @@ static int sdma_v4_0_set_trap_irq_state(struct amdgpu_device *adev,
unsigned type,
enum amdgpu_interrupt_state state)
{
unsigned int instance = (type == AMDGPU_SDMA_IRQ_TRAP0) ? 0 : 1;
u32 sdma_cntl;
u32 reg_offset = (type == AMDGPU_SDMA_IRQ_TRAP0) ?
sdma_v4_0_get_reg_offset(adev, 0, mmSDMA0_CNTL) :
sdma_v4_0_get_reg_offset(adev, 1, mmSDMA0_CNTL);
sdma_cntl = RREG32(reg_offset);
sdma_cntl = RREG32_SDMA(instance, mmSDMA0_CNTL);
sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA0_CNTL, TRAP_ENABLE,
state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
WREG32(reg_offset, sdma_cntl);
WREG32_SDMA(instance, mmSDMA0_CNTL, sdma_cntl);
return 0;
}
@@ -1470,39 +1670,32 @@ static int sdma_v4_0_process_trap_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
uint32_t instance;
DRM_DEBUG("IH: SDMA trap\n");
switch (entry->client_id) {
case SOC15_IH_CLIENTID_SDMA0:
switch (entry->ring_id) {
case 0:
amdgpu_fence_process(&adev->sdma.instance[0].ring);
break;
case 1:
/* XXX compute */
break;
case 2:
/* XXX compute */
break;
case 3:
/* XXX page queue*/
break;
}
instance = 0;
break;
case SOC15_IH_CLIENTID_SDMA1:
switch (entry->ring_id) {
case 0:
amdgpu_fence_process(&adev->sdma.instance[1].ring);
break;
case 1:
/* XXX compute */
break;
case 2:
/* XXX compute */
break;
case 3:
/* XXX page queue*/
break;
}
instance = 1;
break;
default:
return 0;
}
switch (entry->ring_id) {
case 0:
amdgpu_fence_process(&adev->sdma.instance[instance].ring);
break;
case 1:
/* XXX compute */
break;
case 2:
/* XXX compute */
break;
case 3:
amdgpu_fence_process(&adev->sdma.instance[instance].page);
break;
}
return 0;
@@ -1512,12 +1705,29 @@ static int sdma_v4_0_process_illegal_inst_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
int instance;
DRM_ERROR("Illegal instruction in SDMA command stream\n");
schedule_work(&adev->reset_work);
switch (entry->client_id) {
case SOC15_IH_CLIENTID_SDMA0:
instance = 0;
break;
case SOC15_IH_CLIENTID_SDMA1:
instance = 1;
break;
default:
return 0;
}
switch (entry->ring_id) {
case 0:
drm_sched_fault(&adev->sdma.instance[instance].ring.sched);
break;
}
return 0;
}
static void sdma_v4_0_update_medium_grain_clock_gating(
struct amdgpu_device *adev,
bool enable)
@@ -1730,6 +1940,38 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = {
.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 = {
.type = AMDGPU_RING_TYPE_SDMA,
.align_mask = 0xf,
.nop = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP),
.support_64bit_ptrs = true,
.vmhub = AMDGPU_MMHUB,
.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 void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev)
{
int i;
@@ -1737,6 +1979,10 @@ static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev)
for (i = 0; i < adev->sdma.num_instances; i++) {
adev->sdma.instance[i].ring.funcs = &sdma_v4_0_ring_funcs;
adev->sdma.instance[i].ring.me = i;
if (adev->sdma.has_page_queue) {
adev->sdma.instance[i].page.funcs = &sdma_v4_0_page_ring_funcs;
adev->sdma.instance[i].page.me = i;
}
}
}
@@ -1818,7 +2064,10 @@ static const struct amdgpu_buffer_funcs sdma_v4_0_buffer_funcs = {
static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev)
{
adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs;
adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
if (adev->sdma.has_page_queue)
adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].page;
else
adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring;
}
static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = {
@@ -1836,7 +2085,10 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev)
adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs;
for (i = 0; i < adev->sdma.num_instances; i++) {
sched = &adev->sdma.instance[i].ring.sched;
if (adev->sdma.has_page_queue)
sched = &adev->sdma.instance[i].page.sched;
else
sched = &adev->sdma.instance[i].ring.sched;
adev->vm_manager.vm_pte_rqs[i] =
&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
}

68
drivers/gpu/drm/amd/amdgpu/si_dma.c
View File

@@ -61,9 +61,11 @@ static void si_dma_ring_set_wptr(struct amdgpu_ring *ring)
}
static void si_dma_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
/* The indirect buffer packet must end on an 8 DW boundary in the DMA ring.
* Pad as necessary with NOPs.
*/
@@ -122,7 +124,7 @@ static void si_dma_stop(struct amdgpu_device *adev)
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, false);
ring->ready = false;
ring->sched.ready = false;
}
}
@@ -175,13 +177,11 @@ static int si_dma_start(struct amdgpu_device *adev)
WREG32(DMA_RB_WPTR + sdma_offsets[i], lower_32_bits(ring->wptr) << 2);
WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_RB_ENABLE);
ring->ready = true;
ring->sched.ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
return r;
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_buffer_funcs_status(adev, true);
@@ -209,21 +209,16 @@ static int si_dma_ring_test_ring(struct amdgpu_ring *ring)
u64 gpu_addr;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
if (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, 4);
if (r) {
DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
amdgpu_device_wb_free(adev, index);
return r;
}
if (r)
goto error_free_wb;
amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1));
amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
@@ -238,15 +233,11 @@ static int si_dma_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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_device_wb_free(adev, index);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
error_free_wb:
amdgpu_device_wb_free(adev, index);
return r;
}
@@ -269,20 +260,16 @@ static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = amdgpu_device_wb_get(adev, &index);
if (r) {
dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r);
if (r)
return r;
}
gpu_addr = adev->wb.gpu_addr + (index * 4);
tmp = 0xCAFEDEAD;
adev->wb.wb[index] = cpu_to_le32(tmp);
memset(&ib, 0, sizeof(ib));
r = amdgpu_ib_get(adev, NULL, 256, &ib);
if (r) {
DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
if (r)
goto err0;
}
ib.ptr[0] = DMA_PACKET(DMA_PACKET_WRITE, 0, 0, 0, 1);
ib.ptr[1] = lower_32_bits(gpu_addr);
@@ -295,21 +282,16 @@ static int si_dma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
r = dma_fence_wait_timeout(f, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out\n");
r = -ETIMEDOUT;
goto err1;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
goto err1;
}
tmp = le32_to_cpu(adev->wb.wb[index]);
if (tmp == 0xDEADBEEF) {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
if (tmp == 0xDEADBEEF)
r = 0;
} else {
DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
else
r = -EINVAL;
}
err1:
amdgpu_ib_free(adev, &ib, NULL);
@@ -658,15 +640,6 @@ static int si_dma_process_trap_irq(struct amdgpu_device *adev,
return 0;
}
static int si_dma_process_illegal_inst_irq(struct amdgpu_device *adev,
struct amdgpu_irq_src *source,
struct amdgpu_iv_entry *entry)
{
DRM_ERROR("Illegal instruction in SDMA command stream\n");
schedule_work(&adev->reset_work);
return 0;
}
static int si_dma_set_clockgating_state(void *handle,
enum amd_clockgating_state state)
{
@@ -781,15 +754,10 @@ static const struct amdgpu_irq_src_funcs si_dma_trap_irq_funcs = {
.process = si_dma_process_trap_irq,
};
static const struct amdgpu_irq_src_funcs si_dma_illegal_inst_irq_funcs = {
.process = si_dma_process_illegal_inst_irq,
};
static void si_dma_set_irq_funcs(struct amdgpu_device *adev)
{
adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST;
adev->sdma.trap_irq.funcs = &si_dma_trap_irq_funcs;
adev->sdma.illegal_inst_irq.funcs = &si_dma_illegal_inst_irq_funcs;
}
/**

130
drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h Normal file
View File

@@ -0,0 +1,130 @@
/*
* Copyright 2018 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
#ifndef _TA_XGMI_IF_H
#define _TA_XGMI_IF_H
/* Responses have bit 31 set */
#define RSP_ID_MASK (1U << 31)
#define RSP_ID(cmdId) (((uint32_t)(cmdId)) | RSP_ID_MASK)
enum ta_command_xgmi {
TA_COMMAND_XGMI__INITIALIZE = 0x00,
TA_COMMAND_XGMI__GET_NODE_ID = 0x01,
TA_COMMAND_XGMI__GET_HIVE_ID = 0x02,
TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO = 0x03,
TA_COMMAND_XGMI__SET_TOPOLOGY_INFO = 0x04
};
/* XGMI related enumerations */
/**********************************************************/;
enum ta_xgmi_connected_nodes {
TA_XGMI__MAX_CONNECTED_NODES = 64
};
enum ta_xgmi_status {
TA_XGMI_STATUS__SUCCESS = 0x00,
TA_XGMI_STATUS__GENERIC_FAILURE = 0x01,
TA_XGMI_STATUS__NULL_POINTER = 0x02,
TA_XGMI_STATUS__INVALID_PARAMETER = 0x03,
TA_XGMI_STATUS__NOT_INITIALIZED = 0x04,
TA_XGMI_STATUS__INVALID_NODE_NUM = 0x05,
TA_XGMI_STATUS__INVALID_NODE_ID = 0x06,
TA_XGMI_STATUS__INVALID_TOPOLOGY = 0x07,
TA_XGMI_STATUS__FAILED_ID_GEN = 0x08,
TA_XGMI_STATUS__FAILED_TOPOLOGY_INIT = 0x09,
TA_XGMI_STATUS__SET_SHARING_ERROR = 0x0A
};
enum ta_xgmi_assigned_sdma_engine {
TA_XGMI_ASSIGNED_SDMA_ENGINE__NOT_ASSIGNED = -1,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA0 = 0,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA1 = 1,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA2 = 2,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA3 = 3,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA4 = 4,
TA_XGMI_ASSIGNED_SDMA_ENGINE__SDMA5 = 5
};
/* input/output structures for XGMI commands */
/**********************************************************/
struct ta_xgmi_node_info {
uint64_t node_id;
uint8_t num_hops;
uint8_t is_sharing_enabled;
enum ta_xgmi_assigned_sdma_engine sdma_engine;
};
struct ta_xgmi_cmd_initialize_output {
uint32_t status;
};
struct ta_xgmi_cmd_get_node_id_output {
uint64_t node_id;
};
struct ta_xgmi_cmd_get_hive_id_output {
uint64_t hive_id;
};
struct ta_xgmi_cmd_get_topology_info_input {
uint32_t num_nodes;
struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES];
};
struct ta_xgmi_cmd_get_topology_info_output {
uint32_t num_nodes;
struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES];
};
struct ta_xgmi_cmd_set_topology_info_input {
uint32_t num_nodes;
struct ta_xgmi_node_info nodes[TA_XGMI__MAX_CONNECTED_NODES];
};
/**********************************************************/
/* Common input structure for XGMI callbacks */
union ta_xgmi_cmd_input {
struct ta_xgmi_cmd_get_topology_info_input get_topology_info;
struct ta_xgmi_cmd_set_topology_info_input set_topology_info;
};
/* Common output structure for XGMI callbacks */
union ta_xgmi_cmd_output {
struct ta_xgmi_cmd_initialize_output initialize;
struct ta_xgmi_cmd_get_node_id_output get_node_id;
struct ta_xgmi_cmd_get_hive_id_output get_hive_id;
struct ta_xgmi_cmd_get_topology_info_output get_topology_info;
};
/**********************************************************/
struct ta_xgmi_shared_memory {
uint32_t cmd_id;
uint32_t resp_id;
enum ta_xgmi_status xgmi_status;
uint32_t reserved;
union ta_xgmi_cmd_input xgmi_in_message;
union ta_xgmi_cmd_output xgmi_out_message;
};
#endif //_TA_XGMI_IF_H

29
drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
View File

@@ -162,12 +162,9 @@ static int uvd_v4_2_hw_init(void *handle)
uvd_v4_2_enable_mgcg(adev, true);
amdgpu_asic_set_uvd_clocks(adev, 10000, 10000);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
@@ -218,7 +215,7 @@ static int uvd_v4_2_hw_fini(void *handle)
if (RREG32(mmUVD_STATUS) != 0)
uvd_v4_2_stop(adev);
ring->ready = false;
ring->sched.ready = false;
return 0;
}
@@ -484,11 +481,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring)
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);
if (r)
return r;
}
amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
@@ -499,14 +494,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -519,8 +509,9 @@ static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring)
* Write ring commands to execute the indirect buffer
*/
static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
amdgpu_ring_write(ring, PACKET0(mmUVD_RBC_IB_BASE, 0));
amdgpu_ring_write(ring, ib->gpu_addr);

28
drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
View File

@@ -158,12 +158,9 @@ static int uvd_v5_0_hw_init(void *handle)
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;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
@@ -215,7 +212,7 @@ static int uvd_v5_0_hw_fini(void *handle)
if (RREG32(mmUVD_STATUS) != 0)
uvd_v5_0_stop(adev);
ring->ready = false;
ring->sched.ready = false;
return 0;
}
@@ -500,11 +497,8 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring)
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);
if (r)
return r;
}
amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
@@ -515,14 +509,9 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -535,8 +524,9 @@ static int uvd_v5_0_ring_test_ring(struct amdgpu_ring *ring)
* Write ring commands to execute the indirect buffer
*/
static void uvd_v5_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_64BIT_BAR_LOW, 0));
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

76
drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
View File

@@ -175,11 +175,8 @@ static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring)
int r;
r = amdgpu_ring_alloc(ring, 16);
if (r) {
DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n",
ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring, HEVC_ENC_CMD_END);
amdgpu_ring_commit(ring);
@@ -189,14 +186,8 @@ static int uvd_v6_0_enc_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("ring test on %d succeeded in %d usecs\n",
ring->idx, i);
} else {
DRM_ERROR("amdgpu: ring %d test failed\n",
ring->idx);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
}
return r;
}
@@ -336,31 +327,24 @@ static int uvd_v6_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = uvd_v6_0_enc_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r);
if (r)
goto error;
}
r = uvd_v6_0_enc_get_destroy_msg(ring, 1, &fence);
if (r) {
DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: IB test timed out.\n");
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r);
} else {
DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx);
else if (r > 0)
r = 0;
}
error:
dma_fence_put(fence);
return r;
}
static int uvd_v6_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -476,12 +460,9 @@ static int uvd_v6_0_hw_init(void *handle)
uvd_v6_0_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
uvd_v6_0_enable_mgcg(adev, true);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
@@ -513,12 +494,9 @@ static int uvd_v6_0_hw_init(void *handle)
if (uvd_v6_0_enc_support(adev)) {
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst->ring_enc[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
}
}
@@ -548,7 +526,7 @@ static int uvd_v6_0_hw_fini(void *handle)
if (RREG32(mmUVD_STATUS) != 0)
uvd_v6_0_stop(adev);
ring->ready = false;
ring->sched.ready = false;
return 0;
}
@@ -969,11 +947,9 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring)
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);
if (r)
return r;
}
amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
amdgpu_ring_commit(ring);
@@ -984,14 +960,9 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("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;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -1004,9 +975,12 @@ static int uvd_v6_0_ring_test_ring(struct amdgpu_ring *ring)
* Write ring commands to execute the indirect buffer
*/
static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, PACKET0(mmUVD_LMI_RBC_IB_VMID, 0));
amdgpu_ring_write(ring, vmid);
@@ -1027,8 +1001,12 @@ static void uvd_v6_0_ring_emit_ib(struct amdgpu_ring *ring,
* Write enc ring commands to execute the indirect buffer
*/
static void uvd_v6_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM);
amdgpu_ring_write(ring, vmid);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

74
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
View File

@@ -183,11 +183,8 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring)
return 0;
r = amdgpu_ring_alloc(ring, 16);
if (r) {
DRM_ERROR("amdgpu: uvd enc failed to lock (%d)ring %d (%d).\n",
ring->me, ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring, HEVC_ENC_CMD_END);
amdgpu_ring_commit(ring);
@@ -197,14 +194,8 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n",
ring->me, ring->idx, i);
} else {
DRM_ERROR("amdgpu: (%d)ring %d test failed\n",
ring->me, ring->idx);
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
}
return r;
}
@@ -343,27 +334,19 @@ static int uvd_v7_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout)
long r;
r = uvd_v7_0_enc_get_create_msg(ring, 1, NULL);
if (r) {
DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ring->me, r);
if (r)
goto error;
}
r = uvd_v7_0_enc_get_destroy_msg(ring, 1, &fence);
if (r) {
DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ring->me, r);
if (r)
goto error;
}
r = dma_fence_wait_timeout(fence, false, timeout);
if (r == 0) {
DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ring->me);
if (r == 0)
r = -ETIMEDOUT;
} else if (r < 0) {
DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ring->me, r);
} else {
DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ring->me, ring->idx);
else if (r > 0)
r = 0;
}
error:
dma_fence_put(fence);
return r;
@@ -540,12 +523,9 @@ static int uvd_v7_0_hw_init(void *handle)
ring = &adev->uvd.inst[j].ring;
if (!amdgpu_sriov_vf(adev)) {
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
r = amdgpu_ring_alloc(ring, 10);
if (r) {
@@ -582,12 +562,9 @@ static int uvd_v7_0_hw_init(void *handle)
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst[j].ring_enc[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
}
}
done:
@@ -619,7 +596,7 @@ static int uvd_v7_0_hw_fini(void *handle)
for (i = 0; i < adev->uvd.num_uvd_inst; ++i) {
if (adev->uvd.harvest_config & (1 << i))
continue;
adev->uvd.inst[i].ring.ready = false;
adev->uvd.inst[i].ring.sched.ready = false;
}
return 0;
@@ -1235,11 +1212,9 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring)
WREG32_SOC15(UVD, ring->me, mmUVD_CONTEXT_ID, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
if (r) {
DRM_ERROR("amdgpu: (%d)cp failed to lock ring %d (%d).\n",
ring->me, ring->idx, r);
if (r)
return r;
}
amdgpu_ring_write(ring,
PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_CONTEXT_ID), 0));
amdgpu_ring_write(ring, 0xDEADBEEF);
@@ -1251,14 +1226,9 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring)
DRM_UDELAY(1);
}
if (i < adev->usec_timeout) {
DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n",
ring->me, ring->idx, i);
} else {
DRM_ERROR("(%d)amdgpu: ring %d test failed (0x%08X)\n",
ring->me, ring->idx, tmp);
r = -EINVAL;
}
if (i >= adev->usec_timeout)
r = -ETIMEDOUT;
return r;
}
@@ -1300,10 +1270,12 @@ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p,
* Write ring commands to execute the indirect buffer
*/
static void uvd_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_job *job,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
bool ctx_switch)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring,
PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_LMI_RBC_IB_VMID), 0));
@@ -1329,8 +1301,12 @@ static void uvd_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
* Write enc ring commands to execute the indirect buffer
*/
static void uvd_v7_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, HEVC_ENC_CMD_IB_VM);
amdgpu_ring_write(ring, vmid);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

6
drivers/gpu/drm/amd/amdgpu/vce_v2_0.c
View File

@@ -463,15 +463,11 @@ static int vce_v2_0_hw_init(void *handle)
amdgpu_asic_set_vce_clocks(adev, 10000, 10000);
vce_v2_0_enable_mgcg(adev, true, false);
for (i = 0; i < adev->vce.num_rings; i++)
adev->vce.ring[i].ready = false;
for (i = 0; i < adev->vce.num_rings; i++) {
r = amdgpu_ring_test_ring(&adev->vce.ring[i]);
r = amdgpu_ring_test_helper(&adev->vce.ring[i]);
if (r)
return r;
else
adev->vce.ring[i].ready = true;
}
DRM_INFO("VCE initialized successfully.\n");

13
drivers/gpu/drm/amd/amdgpu/vce_v3_0.c
View File

@@ -474,15 +474,10 @@ static int vce_v3_0_hw_init(void *handle)
amdgpu_asic_set_vce_clocks(adev, 10000, 10000);
for (i = 0; i < adev->vce.num_rings; i++)
adev->vce.ring[i].ready = false;
for (i = 0; i < adev->vce.num_rings; i++) {
r = amdgpu_ring_test_ring(&adev->vce.ring[i]);
r = amdgpu_ring_test_helper(&adev->vce.ring[i]);
if (r)
return r;
else
adev->vce.ring[i].ready = true;
}
DRM_INFO("VCE initialized successfully.\n");
@@ -838,8 +833,12 @@ out:
}
static void vce_v3_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, VCE_CMD_IB_VM);
amdgpu_ring_write(ring, vmid);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

15
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
View File

@@ -519,15 +519,10 @@ static int vce_v4_0_hw_init(void *handle)
if (r)
return r;
for (i = 0; i < adev->vce.num_rings; i++)
adev->vce.ring[i].ready = false;
for (i = 0; i < adev->vce.num_rings; i++) {
r = amdgpu_ring_test_ring(&adev->vce.ring[i]);
r = amdgpu_ring_test_helper(&adev->vce.ring[i]);
if (r)
return r;
else
adev->vce.ring[i].ready = true;
}
DRM_INFO("VCE initialized successfully.\n");
@@ -549,7 +544,7 @@ static int vce_v4_0_hw_fini(void *handle)
}
for (i = 0; i < adev->vce.num_rings; i++)
adev->vce.ring[i].ready = false;
adev->vce.ring[i].sched.ready = false;
return 0;
}
@@ -951,9 +946,11 @@ static int vce_v4_0_set_powergating_state(void *handle,
}
#endif
static void vce_v4_0_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
static void vce_v4_0_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_job *job,
struct amdgpu_ib *ib, bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, VCE_CMD_IB_VM);
amdgpu_ring_write(ring, vmid);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));

42
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
View File

@@ -176,30 +176,22 @@ static int vcn_v1_0_hw_init(void *handle)
struct amdgpu_ring *ring = &adev->vcn.ring_dec;
int i, r;
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
ring = &adev->vcn.ring_enc[i];
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
ring->sched.ready = true;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
}
ring = &adev->vcn.ring_jpeg;
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
r = amdgpu_ring_test_helper(ring);
if (r)
goto done;
}
done:
if (!r)
@@ -224,7 +216,7 @@ static int vcn_v1_0_hw_fini(void *handle)
if (RREG32_SOC15(VCN, 0, mmUVD_STATUS))
vcn_v1_0_stop(adev);
ring->ready = false;
ring->sched.ready = false;
return 0;
}
@@ -1366,10 +1358,12 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
* Write ring commands to execute the indirect buffer
*/
static void vcn_v1_0_dec_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring,
PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0));
@@ -1524,8 +1518,12 @@ static void vcn_v1_0_enc_ring_insert_end(struct amdgpu_ring *ring)
* Write enc ring commands to execute the indirect buffer
*/
static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib, unsigned int vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring, VCN_ENC_CMD_IB);
amdgpu_ring_write(ring, vmid);
amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr));
@@ -1725,10 +1723,12 @@ static void vcn_v1_0_jpeg_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u6
* Write ring commands to execute the indirect buffer.
*/
static void vcn_v1_0_jpeg_ring_emit_ib(struct amdgpu_ring *ring,
struct amdgpu_ib *ib,
unsigned vmid, bool ctx_switch)
struct amdgpu_job *job,
struct amdgpu_ib *ib,
bool ctx_switch)
{
struct amdgpu_device *adev = ring->adev;
unsigned vmid = AMDGPU_JOB_GET_VMID(job);
amdgpu_ring_write(ring,
PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0));