Merge branch 'drm-next-4.13' of git://people.freedesktop.org/~agd5f/linux into drm...
authorDave Airlie <airlied@redhat.com>
Thu, 15 Jun 2017 23:54:02 +0000 (09:54 +1000)
committerDave Airlie <airlied@redhat.com>
Thu, 15 Jun 2017 23:56:53 +0000 (09:56 +1000)
New radeon and amdgpu features for 4.13:
- Lots of Vega10 bug fixes
- Preliminary Raven support
- KIQ support for compute rings
- MEC queue management rework from Andres
- Audio support for DCE6
- SR-IOV improvements
- Improved module parameters for controlling radeon vs amdgpu support
  for SI and CIK
- Bug fixes
- General code cleanups

[airlied: dropped drmP.h header from one file was needed and build broke]

* 'drm-next-4.13' of git://people.freedesktop.org/~agd5f/linux: (362 commits)
  drm/amdgpu: Fix compiler warnings
  drm/amdgpu: vm_update_ptes remove code duplication
  drm/amd/amdgpu: Port VCN over to new SOC15 macros
  drm/amd/amdgpu: Port PSP v10.0 over to new SOC15 macros
  drm/amd/amdgpu: Port PSP v3.1 over to new SOC15 macros
  drm/amd/amdgpu: Port NBIO v7.0 driver over to new SOC15 macros
  drm/amd/amdgpu: Port NBIO v6.1 driver over to new SOC15 macros
  drm/amd/amdgpu: Port UVD 7.0 over to new SOC15 macros
  drm/amd/amdgpu: Port MMHUB over to new SOC15 macros
  drm/amd/amdgpu: Cleanup gfxhub read-modify-write patterns
  drm/amd/amdgpu: Port GFXHUB over to new SOC15 macros
  drm/amd/amdgpu: Add offset variant to SOC15 macros
  drm/amd/powerplay: add avfs control for Vega10
  drm/amdgpu: add virtual display support for raven
  drm/amdgpu/gfx9: fix compute ring doorbell index
  drm/amd/amdgpu: Rename KIQ ring to avoid spaces
  drm/amd/amdgpu: gfx9 tidy ups (v2)
  drm/amdgpu: add contiguous flag in ucode bo create
  drm/amdgpu: fix missed gpu info firmware when cache firmware during S3
  drm/amdgpu: export test ib debugfs interface
  ...

28 files changed:
1  2 
drivers/gpu/drm/amd/amdgpu/Makefile
drivers/gpu/drm/amd/amdgpu/amdgpu.h
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
drivers/gpu/drm/amd/amdgpu/cik.c
drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
drivers/gpu/drm/amd/amdgpu/psp_v10_0.c
drivers/gpu/drm/amd/amdgpu/psp_v3_1.c
drivers/gpu/drm/amd/amdgpu/si.c
drivers/gpu/drm/amd/amdgpu/soc15.c
drivers/gpu/drm/amd/amdgpu/vega10_ih.c
drivers/gpu/drm/amd/amdgpu/vi.c
drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c
drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c
drivers/gpu/drm/radeon/cik.c
drivers/gpu/drm/radeon/radeon.h
drivers/gpu/drm/radeon/radeon_drv.c
drivers/gpu/drm/radeon/radeon_kms.c

Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
index 0000000000000000000000000000000000000000,00c0f281f972db148bcd1ce2caa9ee3be2b04152..20c1e539ff35e3e78d542fb099bea7af292a6ad1
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,309 +1,308 @@@
 -#include "drmP.h"
+ /*
+  * 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: Huang Rui
+  *
+  */
+ #include <linux/firmware.h>
+ #include "amdgpu.h"
+ #include "amdgpu_psp.h"
+ #include "amdgpu_ucode.h"
+ #include "soc15_common.h"
+ #include "psp_v10_0.h"
+ #include "vega10/soc15ip.h"
+ #include "raven1/MP/mp_10_0_offset.h"
+ #include "raven1/GC/gc_9_1_offset.h"
+ #include "raven1/SDMA0/sdma0_4_1_offset.h"
+ static int
+ psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type)
+ {
+       switch(ucode->ucode_id) {
+       case AMDGPU_UCODE_ID_SDMA0:
+               *type = GFX_FW_TYPE_SDMA0;
+               break;
+       case AMDGPU_UCODE_ID_SDMA1:
+               *type = GFX_FW_TYPE_SDMA1;
+               break;
+       case AMDGPU_UCODE_ID_CP_CE:
+               *type = GFX_FW_TYPE_CP_CE;
+               break;
+       case AMDGPU_UCODE_ID_CP_PFP:
+               *type = GFX_FW_TYPE_CP_PFP;
+               break;
+       case AMDGPU_UCODE_ID_CP_ME:
+               *type = GFX_FW_TYPE_CP_ME;
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC1:
+               *type = GFX_FW_TYPE_CP_MEC;
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC1_JT:
+               *type = GFX_FW_TYPE_CP_MEC_ME1;
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC2:
+               *type = GFX_FW_TYPE_CP_MEC;
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC2_JT:
+               *type = GFX_FW_TYPE_CP_MEC_ME2;
+               break;
+       case AMDGPU_UCODE_ID_RLC_G:
+               *type = GFX_FW_TYPE_RLC_G;
+               break;
+       case AMDGPU_UCODE_ID_SMC:
+               *type = GFX_FW_TYPE_SMU;
+               break;
+       case AMDGPU_UCODE_ID_UVD:
+               *type = GFX_FW_TYPE_UVD;
+               break;
+       case AMDGPU_UCODE_ID_VCE:
+               *type = GFX_FW_TYPE_VCE;
+               break;
+       case AMDGPU_UCODE_ID_MAXIMUM:
+       default:
+               return -EINVAL;
+       }
+       return 0;
+ }
+ int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cmd_resp *cmd)
+ {
+       int ret;
+       uint64_t fw_mem_mc_addr = ucode->mc_addr;
+       struct  common_firmware_header *header;
+       memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp));
+       header = (struct common_firmware_header *)ucode->fw;
+       cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW;
+       cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = (uint32_t)fw_mem_mc_addr;
+       cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = (uint32_t)((uint64_t)fw_mem_mc_addr >> 32);
+       cmd->cmd.cmd_load_ip_fw.fw_size = le32_to_cpu(header->ucode_size_bytes);
+       ret = psp_v10_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type);
+       if (ret)
+               DRM_ERROR("Unknown firmware type\n");
+       return ret;
+ }
+ int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type)
+ {
+       int ret = 0;
+       unsigned int psp_ring_reg = 0;
+       struct psp_ring *ring;
+       struct amdgpu_device *adev = psp->adev;
+       ring = &psp->km_ring;
+       ring->ring_type = ring_type;
+       /* allocate 4k Page of Local Frame Buffer memory for ring */
+       ring->ring_size = 0x1000;
+       ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE,
+                                     AMDGPU_GEM_DOMAIN_VRAM,
+                                     &adev->firmware.rbuf,
+                                     &ring->ring_mem_mc_addr,
+                                     (void **)&ring->ring_mem);
+       if (ret) {
+               ring->ring_size = 0;
+               return ret;
+       }
+       /* Write low address of the ring to C2PMSG_69 */
+       psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr);
+       WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg);
+       /* Write high address of the ring to C2PMSG_70 */
+       psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr);
+       WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg);
+       /* Write size of ring to C2PMSG_71 */
+       psp_ring_reg = ring->ring_size;
+       WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg);
+       /* Write the ring initialization command to C2PMSG_64 */
+       psp_ring_reg = ring_type;
+       psp_ring_reg = psp_ring_reg << 16;
+       WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg);
+       /* Wait for response flag (bit 31) in C2PMSG_64 */
+       psp_ring_reg = 0;
+       while ((psp_ring_reg & 0x80000000) == 0) {
+               psp_ring_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64);
+       }
+       return 0;
+ }
+ int psp_v10_0_cmd_submit(struct psp_context *psp,
+                       struct amdgpu_firmware_info *ucode,
+                       uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr,
+                       int index)
+ {
+       unsigned int psp_write_ptr_reg = 0;
+       struct psp_gfx_rb_frame * write_frame = psp->km_ring.ring_mem;
+       struct psp_ring *ring = &psp->km_ring;
+       struct amdgpu_device *adev = psp->adev;
+       /* KM (GPCOM) prepare write pointer */
+       psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67);
+       /* Update KM RB frame pointer to new frame */
+       if ((psp_write_ptr_reg % ring->ring_size) == 0)
+               write_frame = ring->ring_mem;
+       else
+               write_frame = ring->ring_mem + (psp_write_ptr_reg / (sizeof(struct psp_gfx_rb_frame) / 4));
+       /* Update KM RB frame */
+       write_frame->cmd_buf_addr_hi = (unsigned int)(cmd_buf_mc_addr >> 32);
+       write_frame->cmd_buf_addr_lo = (unsigned int)(cmd_buf_mc_addr);
+       write_frame->fence_addr_hi = (unsigned int)(fence_mc_addr >> 32);
+       write_frame->fence_addr_lo = (unsigned int)(fence_mc_addr);
+       write_frame->fence_value = index;
+       /* Update the write Pointer in DWORDs */
+       psp_write_ptr_reg += sizeof(struct psp_gfx_rb_frame) / 4;
+       psp_write_ptr_reg = (psp_write_ptr_reg >= ring->ring_size) ? 0 : psp_write_ptr_reg;
+       WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg);
+       return 0;
+ }
+ static int
+ psp_v10_0_sram_map(unsigned int *sram_offset, unsigned int *sram_addr_reg_offset,
+                 unsigned int *sram_data_reg_offset,
+                 enum AMDGPU_UCODE_ID ucode_id)
+ {
+       int ret = 0;
+       switch(ucode_id) {
+ /* TODO: needs to confirm */
+ #if 0
+       case AMDGPU_UCODE_ID_SMC:
+               *sram_offset = 0;
+               *sram_addr_reg_offset = 0;
+               *sram_data_reg_offset = 0;
+               break;
+ #endif
+       case AMDGPU_UCODE_ID_CP_CE:
+               *sram_offset = 0x0;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_CP_PFP:
+               *sram_offset = 0x0;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_CP_ME:
+               *sram_offset = 0x0;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_ME_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_ME_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC1:
+               *sram_offset = 0x10000;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_CP_MEC2:
+               *sram_offset = 0x10000;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_MEC2_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_MEC2_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_RLC_G:
+               *sram_offset = 0x2000;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_DATA);
+               break;
+       case AMDGPU_UCODE_ID_SDMA0:
+               *sram_offset = 0x0;
+               *sram_addr_reg_offset = SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_UCODE_ADDR);
+               *sram_data_reg_offset = SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_UCODE_DATA);
+               break;
+ /* TODO: needs to confirm */
+ #if 0
+       case AMDGPU_UCODE_ID_SDMA1:
+               *sram_offset = ;
+               *sram_addr_reg_offset = ;
+               break;
+       case AMDGPU_UCODE_ID_UVD:
+               *sram_offset = ;
+               *sram_addr_reg_offset = ;
+               break;
+       case AMDGPU_UCODE_ID_VCE:
+               *sram_offset = ;
+               *sram_addr_reg_offset = ;
+               break;
+ #endif
+       case AMDGPU_UCODE_ID_MAXIMUM:
+       default:
+               ret = -EINVAL;
+               break;
+       }
+       return ret;
+ }
+ bool psp_v10_0_compare_sram_data(struct psp_context *psp,
+                               struct amdgpu_firmware_info *ucode,
+                               enum AMDGPU_UCODE_ID ucode_type)
+ {
+       int err = 0;
+       unsigned int fw_sram_reg_val = 0;
+       unsigned int fw_sram_addr_reg_offset = 0;
+       unsigned int fw_sram_data_reg_offset = 0;
+       unsigned int ucode_size;
+       uint32_t *ucode_mem = NULL;
+       struct amdgpu_device *adev = psp->adev;
+       err = psp_v10_0_sram_map(&fw_sram_reg_val, &fw_sram_addr_reg_offset,
+                               &fw_sram_data_reg_offset, ucode_type);
+       if (err)
+               return false;
+       WREG32(fw_sram_addr_reg_offset, fw_sram_reg_val);
+       ucode_size = ucode->ucode_size;
+       ucode_mem = (uint32_t *)ucode->kaddr;
+       while (!ucode_size) {
+               fw_sram_reg_val = RREG32(fw_sram_data_reg_offset);
+               if (*ucode_mem != fw_sram_reg_val)
+                       return false;
+               ucode_mem++;
+               /* 4 bytes */
+               ucode_size -= 4;
+       }
+       return true;
+ }
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge