RESET_WAVES
};
-struct cik_sdma_rlc_registers;
+struct vi_sdma_mqd;
/*
* Register access functions
uint32_t queue_id, uint32_t __user *wptr,
uint32_t wptr_shift, uint32_t wptr_mask,
struct mm_struct *mm);
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
+static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+ uint32_t __user *wptr, struct mm_struct *mm);
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id);
static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
static const struct kfd2kgd_calls kfd2kgd = {
.init_gtt_mem_allocation = alloc_gtt_mem,
.free_gtt_mem = free_gtt_mem,
- .get_vmem_size = get_vmem_size,
+ .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_vm_alloc_pasid,
.init_interrupts = kgd_init_interrupts,
.hqd_load = kgd_hqd_load,
.hqd_sdma_load = kgd_hqd_sdma_load,
+ .hqd_dump = kgd_hqd_dump,
+ .hqd_sdma_dump = kgd_hqd_sdma_dump,
.hqd_is_occupied = kgd_hqd_is_occupied,
.hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
.hqd_destroy = kgd_hqd_destroy,
.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
};
struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void)
return 0;
}
-static inline uint32_t get_sdma_base_addr(struct cik_sdma_rlc_registers *m)
+static inline uint32_t get_sdma_base_addr(struct vi_sdma_mqd *m)
{
- return 0;
+ uint32_t retval;
+
+ retval = m->sdma_engine_id * SDMA1_REGISTER_OFFSET +
+ m->sdma_queue_id * KFD_VI_SDMA_QUEUE_OFFSET;
+ pr_debug("kfd: sdma base address: 0x%x\n", retval);
+
+ return retval;
}
static inline struct vi_mqd *get_mqd(void *mqd)
return (struct vi_mqd *)mqd;
}
-static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
+static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd)
{
- return (struct cik_sdma_rlc_registers *)mqd;
+ return (struct vi_sdma_mqd *)mqd;
}
static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
return 0;
}
-static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t i = 0, reg;
+#define HQD_N_REGS (54+4)
+#define DUMP_REG(addr) do { \
+ if (WARN_ON_ONCE(i >= HQD_N_REGS)) \
+ break; \
+ (*dump)[i][0] = (addr) << 2; \
+ (*dump)[i++][1] = RREG32(addr); \
+ } while (0)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ acquire_queue(kgd, pipe_id, queue_id);
+
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE2);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE3);
+
+ for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_DONES; reg++)
+ DUMP_REG(reg);
+
+ release_queue(kgd);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
+ return 0;
+}
+
+static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd,
+ uint32_t __user *wptr, struct mm_struct *mm)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ struct vi_sdma_mqd *m;
+ unsigned long end_jiffies;
+ uint32_t sdma_base_addr;
+ uint32_t data;
+
+ m = get_sdma_mqd(mqd);
+ sdma_base_addr = get_sdma_base_addr(m);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL,
+ m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK));
+
+ end_jiffies = msecs_to_jiffies(2000) + jiffies;
+ while (true) {
+ data = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS);
+ if (data & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK)
+ break;
+ if (time_after(jiffies, end_jiffies))
+ return -ETIME;
+ usleep_range(500, 1000);
+ }
+ if (m->sdma_engine_id) {
+ data = RREG32(mmSDMA1_GFX_CONTEXT_CNTL);
+ data = REG_SET_FIELD(data, SDMA1_GFX_CONTEXT_CNTL,
+ RESUME_CTX, 0);
+ WREG32(mmSDMA1_GFX_CONTEXT_CNTL, data);
+ } else {
+ data = RREG32(mmSDMA0_GFX_CONTEXT_CNTL);
+ data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL,
+ RESUME_CTX, 0);
+ WREG32(mmSDMA0_GFX_CONTEXT_CNTL, data);
+ }
+
+ data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA0_RLC0_DOORBELL,
+ ENABLE, 1);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, data);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, m->sdmax_rlcx_rb_rptr);
+
+ if (read_user_wptr(mm, wptr, data))
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, data);
+ else
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR,
+ m->sdmax_rlcx_rb_rptr);
+
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_VIRTUAL_ADDR,
+ m->sdmax_rlcx_virtual_addr);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, m->sdmax_rlcx_rb_base);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE_HI,
+ m->sdmax_rlcx_rb_base_hi);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_LO,
+ m->sdmax_rlcx_rb_rptr_addr_lo);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_HI,
+ m->sdmax_rlcx_rb_rptr_addr_hi);
+
+ data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA0_RLC0_RB_CNTL,
+ RB_ENABLE, 1);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, data);
+
+ return 0;
+}
+
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET +
+ queue_id * KFD_VI_SDMA_QUEUE_OFFSET;
+ uint32_t i = 0, reg;
+#undef HQD_N_REGS
+#define HQD_N_REGS (19+4+2+3+7)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_VIRTUAL_ADDR; reg <= mmSDMA0_RLC0_WATERMARK;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_CSA_ADDR_LO; reg <= mmSDMA0_RLC0_CSA_ADDR_HI;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_IB_SUB_REMAIN; reg <= mmSDMA0_RLC0_DUMMY_REG;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_MIDCMD_DATA0; reg <= mmSDMA0_RLC0_MIDCMD_CNTL;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
return 0;
}
static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
- struct cik_sdma_rlc_registers *m;
+ struct vi_sdma_mqd *m;
uint32_t sdma_base_addr;
uint32_t sdma_rlc_rb_cntl;
unsigned int utimeout)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
- struct cik_sdma_rlc_registers *m;
+ struct vi_sdma_mqd *m;
uint32_t sdma_base_addr;
uint32_t temp;
- int timeout = utimeout;
+ unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;
m = get_sdma_mqd(mqd);
sdma_base_addr = get_sdma_base_addr(m);
while (true) {
temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS);
- if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT)
+ if (temp & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK)
break;
- if (timeout <= 0)
+ if (time_after(jiffies, end_jiffies))
return -ETIME;
- msleep(20);
- timeout -= 20;
+ usleep_range(500, 1000);
}
WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0);
- WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, 0);
- WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, 0);
- WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, 0);
+ WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL,
+ RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL) |
+ SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK);
+
+ m->sdmax_rlcx_rb_rptr = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR);
return 0;
}
projects / linux.git / blobdiff