drm/amdgpu/vce4: impl vce & mmsch sriov start
authorXiangliang Yu <Xiangliang.Yu@amd.com>
Mon, 13 Mar 2017 03:10:12 +0000 (11:10 +0800)
committerAlex Deucher <alexander.deucher@amd.com>
Thu, 30 Mar 2017 03:55:13 +0000 (23:55 -0400)
For MM sriov, need use MMSCH to init engine and the init procedures
are all saved in mm table.

Signed-off-by: Xiangliang Yu <Xiangliang.Yu@amd.com>
Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c

index b1b887ea2f5eb7674ac6de9fcb624a35b0703dbc..15321495e224b2fbbeea29a35203b5962c1b8197 100644 (file)
@@ -30,6 +30,7 @@
 #include "amdgpu_vce.h"
 #include "soc15d.h"
 #include "soc15_common.h"
+#include "mmsch_v1_0.h"
 
 #include "vega10/soc15ip.h"
 #include "vega10/VCE/vce_4_0_offset.h"
@@ -48,6 +49,63 @@ static void vce_v4_0_mc_resume(struct amdgpu_device *adev);
 static void vce_v4_0_set_ring_funcs(struct amdgpu_device *adev);
 static void vce_v4_0_set_irq_funcs(struct amdgpu_device *adev);
 
+static inline void mmsch_insert_direct_wt(struct mmsch_v1_0_cmd_direct_write *direct_wt,
+                                         uint32_t *init_table,
+                                         uint32_t reg_offset,
+                                         uint32_t value)
+{
+       direct_wt->cmd_header.reg_offset = reg_offset;
+       direct_wt->reg_value = value;
+       memcpy((void *)init_table, direct_wt, sizeof(struct mmsch_v1_0_cmd_direct_write));
+}
+
+static inline void mmsch_insert_direct_rd_mod_wt(struct mmsch_v1_0_cmd_direct_read_modify_write *direct_rd_mod_wt,
+                                                uint32_t *init_table,
+                                                uint32_t reg_offset,
+                                                uint32_t mask, uint32_t data)
+{
+       direct_rd_mod_wt->cmd_header.reg_offset = reg_offset;
+       direct_rd_mod_wt->mask_value = mask;
+       direct_rd_mod_wt->write_data = data;
+       memcpy((void *)init_table, direct_rd_mod_wt,
+              sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write));
+}
+
+static inline void mmsch_insert_direct_poll(struct mmsch_v1_0_cmd_direct_polling *direct_poll,
+                                           uint32_t *init_table,
+                                           uint32_t reg_offset,
+                                           uint32_t mask, uint32_t wait)
+{
+       direct_poll->cmd_header.reg_offset = reg_offset;
+       direct_poll->mask_value = mask;
+       direct_poll->wait_value = wait;
+       memcpy((void *)init_table, direct_poll, sizeof(struct mmsch_v1_0_cmd_direct_polling));
+}
+
+#define INSERT_DIRECT_RD_MOD_WT(reg, mask, data) { \
+       mmsch_insert_direct_rd_mod_wt(&direct_rd_mod_wt, \
+                                     init_table, (reg), \
+                                     (mask), (data)); \
+       init_table += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
+       table_size += sizeof(struct mmsch_v1_0_cmd_direct_read_modify_write)/4; \
+}
+
+#define INSERT_DIRECT_WT(reg, value) { \
+       mmsch_insert_direct_wt(&direct_wt, \
+                              init_table, (reg), \
+                              (value)); \
+       init_table += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
+       table_size += sizeof(struct mmsch_v1_0_cmd_direct_write)/4; \
+}
+
+#define INSERT_DIRECT_POLL(reg, mask, wait) { \
+       mmsch_insert_direct_poll(&direct_poll, \
+                                init_table, (reg), \
+                                (mask), (wait)); \
+       init_table += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
+       table_size += sizeof(struct mmsch_v1_0_cmd_direct_polling)/4; \
+}
+
 /**
  * vce_v4_0_ring_get_rptr - get read pointer
  *
@@ -146,6 +204,148 @@ static int vce_v4_0_firmware_loaded(struct amdgpu_device *adev)
        return -ETIMEDOUT;
 }
 
+static int vce_v4_0_mmsch_start(struct amdgpu_device *adev,
+                               struct amdgpu_mm_table *table)
+{
+       uint32_t data = 0, loop;
+       uint64_t addr = table->gpu_addr;
+       struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)table->cpu_addr;
+       uint32_t size;
+
+       size = header->header_size + header->vce_table_size + header->uvd_table_size;
+
+       /* 1, write to vce_mmsch_vf_ctx_addr_lo/hi register with GPU mc addr of memory descriptor location */
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_ADDR_LO), lower_32_bits(addr));
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_ADDR_HI), upper_32_bits(addr));
+
+       /* 2, update vmid of descriptor */
+       data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_VMID));
+       data &= ~VCE_MMSCH_VF_VMID__VF_CTX_VMID_MASK;
+       data |= (0 << VCE_MMSCH_VF_VMID__VF_CTX_VMID__SHIFT); /* use domain0 for MM scheduler */
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_VMID), data);
+
+       /* 3, notify mmsch about the size of this descriptor */
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_CTX_SIZE), size);
+
+       /* 4, set resp to zero */
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP), 0);
+
+       /* 5, kick off the initialization and wait until VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero */
+       WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_HOST), 0x10000001);
+
+       data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP));
+       loop = 1000;
+       while ((data & 0x10000002) != 0x10000002) {
+               udelay(10);
+               data = RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP));
+               loop--;
+               if (!loop)
+                       break;
+       }
+
+       if (!loop) {
+               dev_err(adev->dev, "failed to init MMSCH, mmVCE_MMSCH_VF_MAILBOX_RESP = %x\n", data);
+               return -EBUSY;
+       }
+
+       return 0;
+}
+
+static int vce_v4_0_sriov_start(struct amdgpu_device *adev)
+{
+       struct amdgpu_ring *ring;
+       uint32_t offset, size;
+       uint32_t table_size = 0;
+       struct mmsch_v1_0_cmd_direct_write direct_wt = {0};
+       struct mmsch_v1_0_cmd_direct_read_modify_write direct_rd_mod_wt = {0};
+       struct mmsch_v1_0_cmd_direct_polling direct_poll = {0};
+       struct mmsch_v1_0_cmd_end end = {0};
+       uint32_t *init_table = adev->virt.mm_table.cpu_addr;
+       struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)init_table;
+
+       direct_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_WRITE;
+       direct_rd_mod_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE;
+       direct_poll.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_POLLING;
+       end.cmd_header.command_type = MMSCH_COMMAND__END;
+
+       if (header->vce_table_offset == 0 && header->vce_table_size == 0) {
+               header->version = MMSCH_VERSION;
+               header->header_size = sizeof(struct mmsch_v1_0_init_header) >> 2;
+
+               if (header->uvd_table_offset == 0 && header->uvd_table_size == 0)
+                       header->vce_table_offset = header->header_size;
+               else
+                       header->vce_table_offset = header->uvd_table_size + header->uvd_table_offset;
+
+               init_table += header->vce_table_offset;
+
+               ring = &adev->vce.ring[0];
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_RPTR), ring->wptr);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR), ring->wptr);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_LO), lower_32_bits(ring->gpu_addr));
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_BASE_HI), upper_32_bits(ring->gpu_addr));
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_SIZE), ring->ring_size / 4);
+
+               /* BEGING OF MC_RESUME */
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_A), ~(1 << 16), 0);
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING), ~0xFF9FF000, 0x1FF000);
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_REG_CLOCK_GATING), ~0x3F, 0x3F);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_B), 0x1FF);
+
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL), 0x398000);
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CACHE_CTRL), ~0x1, 0);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL), 0);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL1), 0);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VM_CTRL), 0);
+
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0), adev->vce.gpu_addr >> 8);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR1), adev->vce.gpu_addr >> 8);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR2), adev->vce.gpu_addr >> 8);
+
+               offset = AMDGPU_VCE_FIRMWARE_OFFSET;
+               size = VCE_V4_0_FW_SIZE;
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0), offset & 0x7FFFFFFF);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE0), size);
+
+               offset += size;
+               size = VCE_V4_0_STACK_SIZE;
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET1), offset & 0x7FFFFFFF);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE1), size);
+
+               offset += size;
+               size = VCE_V4_0_DATA_SIZE;
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET2), offset & 0x7FFFFFFF);
+               INSERT_DIRECT_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE2), size);
+
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_CTRL2), ~0x100, 0);
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_EN),
+                               0xffffffff, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
+
+               /* end of MC_RESUME */
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CNTL),
+                               ~0x200001, VCE_VCPU_CNTL__CLK_EN_MASK);
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_SOFT_RESET),
+                               ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, 0);
+
+               INSERT_DIRECT_POLL(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
+                               VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK,
+                               VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK);
+
+               /* clear BUSY flag */
+               INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(VCE, 0, mmVCE_STATUS),
+                               ~VCE_STATUS__JOB_BUSY_MASK, 0);
+
+               /* add end packet */
+               memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end));
+               table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4;
+               header->vce_table_size = table_size;
+
+               return vce_v4_0_mmsch_start(adev, &adev->virt.mm_table);
+       }
+
+       return -EINVAL; /* already initializaed ? */
+}
+
 /**
  * vce_v4_0_start - start VCE block
  *
@@ -339,7 +539,10 @@ static int vce_v4_0_hw_init(void *handle)
        int r, i;
        struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-       r = vce_v4_0_start(adev);
+       if (amdgpu_sriov_vf(adev))
+               r = vce_v4_0_sriov_start(adev);
+       else
+               r = vce_v4_0_start(adev);
        if (r)
                return r;