From 4667fbe2f7c54628c48aedc8c8472cd12a5f7fcd Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Wed, 24 Nov 2021 17:24:58 +0800
Subject: drm/amdgpu: do gfxhub init for all XCDs
Each XCD needs to do gfxhub init
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 581 +++++++++++++++++--------------
1 file changed, 317 insertions(+), 264 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index c59c6c85fbff..79af32bb078c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -43,19 +43,25 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
uint64_t page_table_base)
{
struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ int i;
- WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
- hub->ctx_addr_distance * vmid,
- lower_32_bits(page_table_base));
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ WREG32_SOC15_OFFSET(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+ hub->ctx_addr_distance * vmid,
+ lower_32_bits(page_table_base));
- WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
- hub->ctx_addr_distance * vmid,
- upper_32_bits(page_table_base));
+ WREG32_SOC15_OFFSET(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+ hub->ctx_addr_distance * vmid,
+ upper_32_bits(page_table_base));
+ }
}
static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
{
uint64_t pt_base;
+ int i;
if (adev->gmc.pdb0_bo)
pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
@@ -67,26 +73,36 @@ static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
/* If use GART for FB translation, vmid0 page table covers both
* vram and system memory (gart)
*/
- if (adev->gmc.pdb0_bo) {
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
- (u32)(adev->gmc.fb_start >> 12));
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
- (u32)(adev->gmc.fb_start >> 44));
-
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
- (u32)(adev->gmc.gart_end >> 12));
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
- (u32)(adev->gmc.gart_end >> 44));
- } else {
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
- (u32)(adev->gmc.gart_start >> 12));
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
- (u32)(adev->gmc.gart_start >> 44));
-
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
- (u32)(adev->gmc.gart_end >> 12));
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
- (u32)(adev->gmc.gart_end >> 44));
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ if (adev->gmc.pdb0_bo) {
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+ (u32)(adev->gmc.fb_start >> 12));
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+ (u32)(adev->gmc.fb_start >> 44));
+
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+ (u32)(adev->gmc.gart_end >> 12));
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+ (u32)(adev->gmc.gart_end >> 44));
+ } else {
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+ (u32)(adev->gmc.gart_start >> 12));
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+ (u32)(adev->gmc.gart_start >> 44));
+
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+ (u32)(adev->gmc.gart_end >> 12));
+ WREG32_SOC15(GC, i,
+ regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+ (u32)(adev->gmc.gart_end >> 44));
+ }
}
}
@@ -94,160 +110,183 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
{
uint64_t value;
uint32_t tmp;
+ int i;
- /* Program the AGP BAR */
- WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BASE, 0);
- WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
- WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
-
- if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
- /* Program the system aperture low logical page number. */
- WREG32_SOC15_RLC(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
- min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
-
- if (adev->apu_flags & AMD_APU_IS_RAVEN2)
- /*
- * Raven2 has a HW issue that it is unable to use the
- * vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
- * So here is the workaround that increase system
- * aperture high address (add 1) to get rid of the VM
- * fault and hardware hang.
- */
- WREG32_SOC15_RLC(GC, 0,
- regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
- max((adev->gmc.fb_end >> 18) + 0x1,
- adev->gmc.agp_end >> 18));
- else
- WREG32_SOC15_RLC(GC, 0,
- regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
- max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
-
- /* Set default page address. */
- value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
- WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
- (u32)(value >> 12));
- WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
- (u32)(value >> 44));
-
- /* Program "protection fault". */
- WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
- (u32)(adev->dummy_page_addr >> 12));
- WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
- (u32)((u64)adev->dummy_page_addr >> 44));
-
- tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
- ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
- WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
- }
-
- /* In the case squeezing vram into GART aperture, we don't use
- * FB aperture and AGP aperture. Disable them.
- */
- if (adev->gmc.pdb0_bo) {
- WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_TOP, 0);
- WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
- WREG32_SOC15(GC, 0, regMC_VM_AGP_TOP, 0);
- WREG32_SOC15(GC, 0, regMC_VM_AGP_BOT, 0xFFFFFF);
- WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
- WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ /* Program the AGP BAR */
+ WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BASE, 0);
+ WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+ WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+ if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
+ /* Program the system aperture low logical page number. */
+ WREG32_SOC15_RLC(GC, i, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+ min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+
+ if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+ /*
+ * Raven2 has a HW issue that it is unable to use the
+ * vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
+ * So here is the workaround that increase system
+ * aperture high address (add 1) to get rid of the VM
+ * fault and hardware hang.
+ */
+ WREG32_SOC15_RLC(GC, i,
+ regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+ max((adev->gmc.fb_end >> 18) + 0x1,
+ adev->gmc.agp_end >> 18));
+ else
+ WREG32_SOC15_RLC(GC, i,
+ regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+ max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+ /* Set default page address. */
+ value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+ WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+ (u32)(value >> 12));
+ WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+ (u32)(value >> 44));
+
+ /* Program "protection fault". */
+ WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+ (u32)(adev->dummy_page_addr >> 12));
+ WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+ (u32)((u64)adev->dummy_page_addr >> 44));
+
+ tmp = RREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL2);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
+ ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+ WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+ }
+
+ /* In the case squeezing vram into GART aperture, we don't use
+ * FB aperture and AGP aperture. Disable them.
+ */
+ if (adev->gmc.pdb0_bo) {
+ WREG32_SOC15(GC, i, regMC_VM_FB_LOCATION_TOP, 0);
+ WREG32_SOC15(GC, i, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
+ WREG32_SOC15(GC, i, regMC_VM_AGP_TOP, 0);
+ WREG32_SOC15(GC, i, regMC_VM_AGP_BOT, 0xFFFFFF);
+ WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
+ WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+ }
}
}
static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
{
uint32_t tmp;
+ int i;
- /* Setup TLB control */
- tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
-
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
- ENABLE_ADVANCED_DRIVER_MODEL, 1);
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
- SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
- MTYPE, MTYPE_UC);/* XXX for emulation. */
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
-
- WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ /* Setup TLB control */
+ tmp = RREG32_SOC15(GC, i, regMC_VM_MX_L1_TLB_CNTL);
+
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+ ENABLE_L1_TLB, 1);
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+ SYSTEM_ACCESS_MODE, 3);
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+ ENABLE_ADVANCED_DRIVER_MODEL, 1);
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+ SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+ MTYPE, MTYPE_UC);/* XXX for emulation. */
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
+
+ WREG32_SOC15_RLC(GC, i, regMC_VM_MX_L1_TLB_CNTL, tmp);
+ }
}
static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
{
uint32_t tmp;
+ int i;
- /* Setup L2 cache */
- tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
- /* XXX for emulation, Refer to closed source code.*/
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
- 0);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
- WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL, tmp);
-
- tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL2);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
- WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL2, tmp);
-
- tmp = regVM_L2_CNTL3_DEFAULT;
- if (adev->gmc.translate_further) {
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
- L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
- } else {
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
- L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
- }
- WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL3, tmp);
-
- tmp = regVM_L2_CNTL4_DEFAULT;
- if (adev->gmc.xgmi.connected_to_cpu) {
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
- } else {
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ /* Setup L2 cache */
+ tmp = RREG32_SOC15(GC, i, regVM_L2_CNTL);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
+ /* XXX for emulation, Refer to closed source code.*/
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
+ 0);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+ WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL, tmp);
+
+ tmp = RREG32_SOC15(GC, i, regVM_L2_CNTL2);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+ WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL2, tmp);
+
+ tmp = regVM_L2_CNTL3_DEFAULT;
+ if (adev->gmc.translate_further) {
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+ L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+ } else {
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+ L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+ }
+ WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL3, tmp);
+
+ tmp = regVM_L2_CNTL4_DEFAULT;
+ if (adev->gmc.xgmi.connected_to_cpu) {
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
+ } else {
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+ }
+ WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL4, tmp);
}
- WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL4, tmp);
}
static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
{
uint32_t tmp;
+ int i;
- tmp = RREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
- adev->gmc.vmid0_page_table_depth);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
- adev->gmc.vmid0_page_table_block_size);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
- RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
- WREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL, tmp);
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ tmp = RREG32_SOC15(GC, i, regVM_CONTEXT0_CNTL);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
+ adev->gmc.vmid0_page_table_depth);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
+ adev->gmc.vmid0_page_table_block_size);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
+ RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+ WREG32_SOC15(GC, i, regVM_CONTEXT0_CNTL, tmp);
+ }
}
static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
{
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
- 0XFFFFFFFF);
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
- 0x0000000F);
-
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
- 0);
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
- 0);
-
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
- WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+ int i;
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+ 0XFFFFFFFF);
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+ 0x0000000F);
+
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+ 0);
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+ 0);
+
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+ WREG32_SOC15(GC, i,
+ regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+ }
}
static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
@@ -255,7 +294,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
unsigned num_level, block_size;
uint32_t tmp;
- int i;
+ int i, j;
num_level = adev->vm_manager.num_level;
block_size = adev->vm_manager.block_size;
@@ -264,81 +303,89 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
else
block_size -= 9;
- for (i = 0; i <= 14; i++) {
- tmp = RREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL, i);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
- num_level);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
- 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- PAGE_TABLE_BLOCK_SIZE,
- block_size);
- /* Send no-retry XNACK on fault to suppress VM fault storm.
- * On Aldebaran, XNACK can be enabled in the SQ per-process.
- * Retry faults need to be enabled for that to work.
- */
- tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
- RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
- !adev->gmc.noretry ||
- adev->asic_type == CHIP_ALDEBARAN);
- WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL,
- i * hub->ctx_distance, tmp);
- WREG32_SOC15_OFFSET(GC, 0,
- regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
- i * hub->ctx_addr_distance, 0);
- WREG32_SOC15_OFFSET(GC, 0,
- regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
- i * hub->ctx_addr_distance, 0);
- WREG32_SOC15_OFFSET(GC, 0,
- regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
- i * hub->ctx_addr_distance,
- lower_32_bits(adev->vm_manager.max_pfn - 1));
- WREG32_SOC15_OFFSET(GC, 0,
- regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
- i * hub->ctx_addr_distance,
- upper_32_bits(adev->vm_manager.max_pfn - 1));
+ for (j = 0; j < adev->gfx.num_xcd; j++) {
+ for (i = 0; i <= 14; i++) {
+ tmp = RREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL, i);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+ num_level);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
+ 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ PAGE_TABLE_BLOCK_SIZE,
+ block_size);
+ /* Send no-retry XNACK on fault to suppress VM fault storm.
+ * On Aldebaran, XNACK can be enabled in the SQ per-process.
+ * Retry faults need to be enabled for that to work.
+ */
+ tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+ RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+ !adev->gmc.noretry ||
+ adev->asic_type == CHIP_ALDEBARAN);
+ WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL,
+ i * hub->ctx_distance, tmp);
+ WREG32_SOC15_OFFSET(GC, j,
+ regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+ i * hub->ctx_addr_distance, 0);
+ WREG32_SOC15_OFFSET(GC, j,
+ regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+ i * hub->ctx_addr_distance, 0);
+ WREG32_SOC15_OFFSET(GC, j,
+ regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+ i * hub->ctx_addr_distance,
+ lower_32_bits(adev->vm_manager.max_pfn - 1));
+ WREG32_SOC15_OFFSET(GC, j,
+ regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+ i * hub->ctx_addr_distance,
+ upper_32_bits(adev->vm_manager.max_pfn - 1));
+ }
}
}
static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
{
struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
- unsigned i;
-
- for (i = 0 ; i < 18; ++i) {
- WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
- i * hub->eng_addr_distance, 0xffffffff);
- WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
- i * hub->eng_addr_distance, 0x1f);
+ unsigned i, j;
+
+ for (j = 0; j < adev->gfx.num_xcd; j++) {
+ for (i = 0 ; i < 18; ++i) {
+ WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+ i * hub->eng_addr_distance, 0xffffffff);
+ WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+ i * hub->eng_addr_distance, 0x1f);
+ }
}
}
static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
{
- if (amdgpu_sriov_vf(adev) && adev->asic_type != CHIP_ARCTURUS) {
+ int i;
+
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ if (amdgpu_sriov_vf(adev)) {
/*
* MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
* VF copy registers so vbios post doesn't program them, for
* SRIOV driver need to program them
*/
- WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_BASE,
- adev->gmc.vram_start >> 24);
- WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_TOP,
- adev->gmc.vram_end >> 24);
+ WREG32_SOC15_RLC(GC, i, regMC_VM_FB_LOCATION_BASE,
+ adev->gmc.vram_start >> 24);
+ WREG32_SOC15_RLC(GC, i, regMC_VM_FB_LOCATION_TOP,
+ adev->gmc.vram_end >> 24);
+ }
}
/* GART Enable. */
@@ -361,27 +408,29 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
{
struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
u32 tmp;
- u32 i;
-
- /* Disable all tables */
- for (i = 0; i < 16; i++)
- WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_CNTL,
- i * hub->ctx_distance, 0);
-
- /* Setup TLB control */
- tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
- tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
- tmp = REG_SET_FIELD(tmp,
- MC_VM_MX_L1_TLB_CNTL,
- ENABLE_ADVANCED_DRIVER_MODEL,
- 0);
- WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
-
- /* Setup L2 cache */
- tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
- tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
- WREG32_SOC15(GC, 0, regVM_L2_CNTL, tmp);
- WREG32_SOC15(GC, 0, regVM_L2_CNTL3, 0);
+ u32 i, j;
+
+ for (j = 0; j < adev->gfx.num_xcd; j++) {
+ /* Disable all tables */
+ for (i = 0; i < 16; i++)
+ WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT0_CNTL,
+ i * hub->ctx_distance, 0);
+
+ /* Setup TLB control */
+ tmp = RREG32_SOC15(GC, j, regMC_VM_MX_L1_TLB_CNTL);
+ tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+ tmp = REG_SET_FIELD(tmp,
+ MC_VM_MX_L1_TLB_CNTL,
+ ENABLE_ADVANCED_DRIVER_MODEL,
+ 0);
+ WREG32_SOC15_RLC(GC, j, regMC_VM_MX_L1_TLB_CNTL, tmp);
+
+ /* Setup L2 cache */
+ tmp = RREG32_SOC15(GC, j, regVM_L2_CNTL);
+ tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
+ WREG32_SOC15(GC, j, regVM_L2_CNTL, tmp);
+ WREG32_SOC15(GC, j, regVM_L2_CNTL3, 0);
+ }
}
/**
@@ -394,38 +443,42 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
bool value)
{
u32 tmp;
- tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp,
- VM_L2_PROTECTION_FAULT_CNTL,
- TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
- value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
- if (!value) {
+ int i;
+
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ tmp = RREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp,
+ VM_L2_PROTECTION_FAULT_CNTL,
+ TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+ value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- CRASH_ON_NO_RETRY_FAULT, 1);
+ WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
- CRASH_ON_RETRY_FAULT, 1);
+ EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+ if (!value) {
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ CRASH_ON_NO_RETRY_FAULT, 1);
+ tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+ CRASH_ON_RETRY_FAULT, 1);
+ }
+ WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
}
- WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
}
static void gfxhub_v1_2_init(struct amdgpu_device *adev)
--
cgit v1.2.3
From f4caf5842652f08e024741ef6d423cb0c101d863 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Wed, 14 Sep 2022 16:35:50 +0800
Subject: drm/amdgpu: introduce vmhub definition for multi-partition cases (v3)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
v1: Each partition has its own gfxhub or mmhub. adjust
the num of MAX_VMHUBS and the GFXHUB/MMHUB layout (Le)
v2: re-design the AMDGPU_GFXHUB/AMDGPU_MMHUB layout (Le)
v3: apply the gfxhub/mmhub layout to new IPs (Hawking)
v4: fix up gmc11 (Alex)
v5: rebase (Alex)
Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 2 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 4 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 4 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 13 +++--
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 6 +--
drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 8 +--
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 8 +--
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 4 +-
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c | 12 ++---
drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 33 ++++++-------
drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c | 26 +++++-----
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 78 +++++++++++++++---------------
drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c | 4 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mes_v10_1.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c | 12 ++---
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 10 ++--
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 +-
drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c | 2 +-
drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c | 8 +--
drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c | 4 +-
drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c | 2 +-
47 files changed, 204 insertions(+), 204 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index a46285841d17..f0a136d35279 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -736,7 +736,7 @@ int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
for (i = 0; i < adev->num_vmhubs; i++)
amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
} else {
- amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, 0);
+ amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0), 0);
}
return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 87e1a1a9f298..488b3bb6dcb1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -315,7 +315,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.kiq;
ring->xcc_id = xcc_id;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
if (xcc_id >= 1)
ring->doorbell_index = adev->doorbell_index.xcc1_kiq_start +
xcc_id - 1;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 4e2531758866..0a4e5fcfec6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -670,7 +670,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
for (i = 0; i < 16; i++) {
reg = hub->vm_context0_cntl + hub->ctx_distance * i;
- tmp = (hub_type == AMDGPU_GFXHUB_0) ?
+ tmp = (hub_type == AMDGPU_GFXHUB(0)) ?
RREG32_SOC15_IP(GC, reg) :
RREG32_SOC15_IP(MMHUB, reg);
@@ -679,7 +679,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
else
tmp &= ~hub->vm_cntx_cntl_vm_fault;
- (hub_type == AMDGPU_GFXHUB_0) ?
+ (hub_type == AMDGPU_GFXHUB(0)) ?
WREG32_SOC15_IP(GC, reg, tmp) :
WREG32_SOC15_IP(MMHUB, reg, tmp);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index b6bd667df676..c3964c14f215 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -2374,12 +2374,12 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
case AMDGPU_VM_OP_RESERVE_VMID:
/* We only have requirement to reserve vmid from gfxhub */
r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm,
- AMDGPU_GFXHUB_0);
+ AMDGPU_GFXHUB(0));
if (r)
return r;
break;
case AMDGPU_VM_OP_UNRESERVE_VMID:
- amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB_0);
+ amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB(0));
break;
default:
return -EINVAL;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 6f085f0b4ef3..9f5d32b0fda1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -111,11 +111,14 @@ struct amdgpu_mem_stats;
/* Reserve 4MB VRAM for page tables */
#define AMDGPU_VM_RESERVED_VRAM (8ULL << 20)
-/* max number of VMHUB */
-#define AMDGPU_MAX_VMHUBS 3
-#define AMDGPU_GFXHUB_0 0
-#define AMDGPU_MMHUB_0 1
-#define AMDGPU_MMHUB_1 2
+/*
+ * max number of VMHUB
+ * layout: max 8 GFXHUB + 4 MMHUB0 + 1 MMHUB1
+ */
+#define AMDGPU_MAX_VMHUBS 13
+#define AMDGPU_GFXHUB(x) (x)
+#define AMDGPU_MMHUB0(x) (8 + x)
+#define AMDGPU_MMHUB1(x) (8 + 4 + x)
/* Reserve 2MB at top/bottom of address space for kernel use */
#define AMDGPU_VA_RESERVED_SIZE (2ULL << 20)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 8e86b2c23c0a..7b585141e10e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -4461,7 +4461,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -4490,7 +4490,7 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX10_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -4978,7 +4978,7 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
nv_grbm_select(adev, 0, 0, 0, i);
/* CP and shaders */
WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index f77779c31043..790df2cc3480 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -906,7 +906,7 @@ static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
else
ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -937,7 +937,7 @@ static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX11_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -1707,7 +1707,7 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
soc21_grbm_select(adev, 0, 0, 0, i);
/* CP and shaders */
WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -4190,7 +4190,7 @@ static int gfx_v11_0_gfxhub_enable(struct amdgpu_device *adev)
false : true;
adev->gfxhub.funcs->set_fault_enable_default(adev, value);
- amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+ amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
return 0;
}
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 46577b59cb04..91814dc083c9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -2005,7 +2005,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX9_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -2105,7 +2105,7 @@ static int gfx_v9_0_sw_init(void *handle)
/* disable scheduler on the real ring */
ring->no_scheduler = true;
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -2123,7 +2123,7 @@ static int gfx_v9_0_sw_init(void *handle)
ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
ring->is_sw_ring = true;
hw_prio = amdgpu_sw_ring_priority(i);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio,
NULL);
@@ -2393,7 +2393,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
soc15_grbm_select(adev, 0, 0, 0, i, 0);
/* CP and shaders */
if (i == 0) {
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
index d648a29c33e0..ec7c049c5952 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
@@ -1935,7 +1935,7 @@ static bool gfx_v9_4_2_query_uctl2_poison_status(struct amdgpu_device *adev)
u32 status = 0;
struct amdgpu_vmhub *hub;
- hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
status = RREG32(hub->vm_l2_pro_fault_status);
/* reset page fault status */
WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 9d17dcfae130..f5104b982633 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -757,7 +757,7 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
(adev->doorbell_index.mec_ring0 + ring_id) << 1;
ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+ (ring_id * GFX9_MEC_HPD_SIZE);
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "comp_%d.%d.%d.%d",
ring->xcc_id, ring->me, ring->pipe, ring->queue);
@@ -996,7 +996,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
- for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+ for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
for (j = 0; j < adev->gfx.num_xcd; j++) {
soc15_grbm_select(adev, 0, 0, 0, i, j);
/* CP and shaders */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
index ab2325f6c7ac..d94cc1ec7242 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
@@ -40,7 +40,7 @@ static void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -247,7 +247,7 @@ static void gfxhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -307,7 +307,7 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -338,7 +338,7 @@ static int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -411,7 +411,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 79af32bb078c..9c385ce3a8c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -42,7 +42,7 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
for (i = 0; i < adev->gfx.num_xcd; i++) {
@@ -291,7 +291,7 @@ static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i, j;
@@ -357,7 +357,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i, j;
for (j = 0; j < adev->gfx.num_xcd; j++) {
@@ -406,7 +406,7 @@ static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i, j;
@@ -483,7 +483,7 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_2_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
index 9b3a02527318..f173a61c6c15 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
@@ -120,7 +120,7 @@ static u64 gfxhub_v2_0_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -282,7 +282,7 @@ static void gfxhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -331,7 +331,7 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v2_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -360,7 +360,7 @@ static int gfxhub_v2_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v2_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -433,7 +433,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_0_vmhub_funcs = {
static void gfxhub_v2_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
index 4aacbbec31e2..d8fc3e8088cd 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
@@ -123,7 +123,7 @@ static u64 gfxhub_v2_1_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v2_1_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -291,7 +291,7 @@ static void gfxhub_v2_1_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -340,7 +340,7 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v2_1_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -381,7 +381,7 @@ static int gfxhub_v2_1_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v2_1_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -462,7 +462,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_1_vmhub_funcs = {
static void gfxhub_v2_1_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
@@ -651,7 +651,7 @@ static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
int time = 1000;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
index 13712640fa46..c53147f9c9fc 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
@@ -119,7 +119,7 @@ static u64 gfxhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -290,7 +290,7 @@ static void gfxhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -339,7 +339,7 @@ static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v3_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -380,7 +380,7 @@ static int gfxhub_v3_0_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v3_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -463,7 +463,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_vmhub_funcs = {
static void gfxhub_v3_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
index 6e0bd628c889..ae777487d72e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
@@ -122,7 +122,7 @@ static u64 gfxhub_v3_0_3_get_mc_fb_offset(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -295,7 +295,7 @@ static void gfxhub_v3_0_3_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
int i;
uint32_t tmp;
@@ -344,7 +344,7 @@ static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
unsigned i;
for (i = 0 ; i < 18; ++i) {
@@ -373,7 +373,7 @@ static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v3_0_3_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
u32 tmp;
u32 i;
@@ -451,7 +451,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_3_vmhub_funcs = {
static void gfxhub_v3_0_3_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index 5697b66bf0de..ea2a448147e3 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -76,7 +76,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -84,11 +84,11 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
break;
case AMDGPU_IRQ_STATE_ENABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -96,7 +96,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
break;
default:
break;
@@ -149,7 +149,7 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
+ if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
RREG32(hub->vm_l2_pro_fault_status);
@@ -212,8 +212,7 @@ static void gmc_v10_0_set_irq_funcs(struct amdgpu_device *adev)
static bool gmc_v10_0_use_invalidate_semaphore(struct amdgpu_device *adev,
uint32_t vmhub)
{
- return ((vmhub == AMDGPU_MMHUB_0 ||
- vmhub == AMDGPU_MMHUB_1) &&
+ return ((vmhub == AMDGPU_MMHUB0(0)) &&
(!amdgpu_sriov_vf(adev)));
}
@@ -249,7 +248,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
unsigned int i;
unsigned char hub_ip = 0;
- hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
+ hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
GC_HWIP : MMHUB_HWIP;
spin_lock(&adev->gmc.invalidate_lock);
@@ -284,7 +283,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
* Issue a dummy read to wait for the ACK register to be cleared
* to avoid a false ACK due to the new fast GRBM interface.
*/
- if ((vmhub == AMDGPU_GFXHUB_0) &&
+ if ((vmhub == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
hub->eng_distance * eng, hub_ip);
@@ -361,19 +360,19 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
mutex_lock(&adev->mman.gtt_window_lock);
- if (vmhub == AMDGPU_MMHUB_0) {
- gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB_0, 0);
+ if (vmhub == AMDGPU_MMHUB0(0)) {
+ gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB0(0), 0);
mutex_unlock(&adev->mman.gtt_window_lock);
return;
}
- BUG_ON(vmhub != AMDGPU_GFXHUB_0);
+ BUG_ON(vmhub != AMDGPU_GFXHUB(0));
if (!adev->mman.buffer_funcs_enabled ||
!adev->ib_pool_ready ||
amdgpu_in_reset(adev) ||
ring->sched.ready == false) {
- gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB_0, 0);
+ gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB(0), 0);
mutex_unlock(&adev->mman.gtt_window_lock);
return;
}
@@ -466,7 +465,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v10_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
if (!adev->enable_mes)
break;
@@ -534,7 +533,7 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1075,9 +1074,9 @@ static int gmc_v10_0_gart_enable(struct amdgpu_device *adev)
if (!adev->in_s0ix)
adev->gfxhub.funcs->set_fault_enable_default(adev, value);
adev->mmhub.funcs->set_fault_enable_default(adev, value);
- gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+ gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
if (!adev->in_s0ix)
- gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+ gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index 2f570fb5febe..fb2ac31cbba7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -64,7 +64,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -72,11 +72,11 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
break;
case AMDGPU_IRQ_STATE_ENABLE:
/* MM HUB */
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
/* GFX HUB */
/* This works because this interrupt is only
* enabled at init/resume and disabled in
@@ -84,7 +84,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* change over the course of suspend/resume.
*/
if (!adev->in_s0ix)
- amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+ amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
break;
default:
break;
@@ -110,7 +110,7 @@ static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if (entry->vmid_src == AMDGPU_GFXHUB_0)
+ if (entry->vmid_src == AMDGPU_GFXHUB(0))
RREG32(hub->vm_l2_pro_fault_status);
status = RREG32(hub->vm_l2_pro_fault_status);
@@ -170,7 +170,7 @@ static void gmc_v11_0_set_irq_funcs(struct amdgpu_device *adev)
static bool gmc_v11_0_use_invalidate_semaphore(struct amdgpu_device *adev,
uint32_t vmhub)
{
- return ((vmhub == AMDGPU_MMHUB_0) &&
+ return ((vmhub == AMDGPU_MMHUB0(0)) &&
(!amdgpu_sriov_vf(adev)));
}
@@ -202,7 +202,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
unsigned int i;
unsigned char hub_ip = 0;
- hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
+ hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
GC_HWIP : MMHUB_HWIP;
spin_lock(&adev->gmc.invalidate_lock);
@@ -251,7 +251,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
hub->eng_distance * eng, 0, hub_ip);
/* Issue additional private vm invalidation to MMHUB */
- if ((vmhub != AMDGPU_GFXHUB_0) &&
+ if ((vmhub != AMDGPU_GFXHUB(0)) &&
(hub->vm_l2_bank_select_reserved_cid2) &&
!amdgpu_sriov_vf(adev)) {
inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
@@ -284,7 +284,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
uint32_t vmhub, uint32_t flush_type)
{
- if ((vmhub == AMDGPU_GFXHUB_0) && !adev->gfx.is_poweron)
+ if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
return;
/* flush hdp cache */
@@ -369,7 +369,7 @@ static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v11_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
}
}
@@ -435,7 +435,7 @@ static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
if (ring->is_mes_queue)
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
@@ -886,7 +886,7 @@ static int gmc_v11_0_sw_fini(void *handle)
static void gmc_v11_0_init_golden_registers(struct amdgpu_device *adev)
{
if (amdgpu_sriov_vf(adev)) {
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32(hub->vm_contexts_disable, 0);
return;
@@ -921,7 +921,7 @@ static int gmc_v11_0_gart_enable(struct amdgpu_device *adev)
false : true;
adev->mmhub.funcs->set_fault_enable_default(adev, value);
- gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+ gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(adev->gmc.gart_size >> 20),
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 6ae5cee9b64b..193ba4d912a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -491,20 +491,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* fini/suspend, so the overall state doesn't
* change over the course of suspend/resume.
*/
- if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
continue;
- if (j == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP(GC, reg);
- else
+ if (j >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP(MMHUB, reg);
+ else
+ tmp = RREG32_SOC15_IP(GC, reg);
tmp &= ~bits;
- if (j == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP(GC, reg, tmp);
- else
+ if (j >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP(MMHUB, reg, tmp);
+ else
+ WREG32_SOC15_IP(GC, reg, tmp);
}
}
break;
@@ -519,20 +519,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
* fini/suspend, so the overall state doesn't
* change over the course of suspend/resume.
*/
- if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
continue;
- if (j == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP(GC, reg);
- else
+ if (j >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP(MMHUB, reg);
+ else
+ tmp = RREG32_SOC15_IP(GC, reg);
tmp |= bits;
- if (j == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP(GC, reg, tmp);
- else
+ if (j >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP(MMHUB, reg, tmp);
+ else
+ WREG32_SOC15_IP(GC, reg, tmp);
}
}
break;
@@ -605,13 +605,13 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
hub_name = "mmhub0";
- hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
hub_name = "mmhub1";
- hub = &adev->vmhub[AMDGPU_MMHUB_1];
+ hub = &adev->vmhub[AMDGPU_MMHUB1(0)];
} else {
hub_name = "gfxhub0";
- hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
}
memset(&task_info, 0, sizeof(struct amdgpu_task_info));
@@ -636,7 +636,7 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
* be updated to avoid reading an incorrect value due to
* the new fast GRBM interface.
*/
- if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
+ if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
RREG32(hub->vm_l2_pro_fault_status);
@@ -649,7 +649,7 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
dev_err(adev->dev,
"VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
status);
- if (hub == &adev->vmhub[AMDGPU_GFXHUB_0]) {
+ if (hub == &adev->vmhub[AMDGPU_GFXHUB(0)]) {
dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" :
gfxhub_client_ids[cid],
@@ -759,8 +759,8 @@ static bool gmc_v9_0_use_invalidate_semaphore(struct amdgpu_device *adev,
adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
return false;
- return ((vmhub == AMDGPU_MMHUB_0 ||
- vmhub == AMDGPU_MMHUB_1) &&
+ return ((vmhub == AMDGPU_MMHUB0(0) ||
+ vmhub == AMDGPU_MMHUB1(0)) &&
(!amdgpu_sriov_vf(adev)) &&
(!(!(adev->apu_flags & AMD_APU_IS_RAVEN2) &&
(adev->apu_flags & AMD_APU_IS_PICASSO))));
@@ -849,11 +849,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
if (use_semaphore) {
for (j = 0; j < adev->usec_timeout; j++) {
/* a read return value of 1 means semaphore acquire */
- if (vmhub == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
-
+ else
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
if (tmp & 0x1)
break;
udelay(1);
@@ -864,27 +863,26 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
}
do {
- if (vmhub == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
+ else
+ WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
/*
* Issue a dummy read to wait for the ACK register to
* be cleared to avoid a false ACK due to the new fast
* GRBM interface.
*/
- if ((vmhub == AMDGPU_GFXHUB_0) &&
+ if ((vmhub == AMDGPU_GFXHUB(0)) &&
(adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
RREG32_NO_KIQ(hub->vm_inv_eng0_req +
hub->eng_distance * eng);
for (j = 0; j < adev->usec_timeout; j++) {
- if (vmhub == AMDGPU_GFXHUB_0)
- tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
- else
+ if (vmhub >= AMDGPU_MMHUB0(0))
tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
-
+ else
+ tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
if (tmp & (1 << vmid))
break;
udelay(1);
@@ -900,10 +898,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
* add semaphore release after invalidation,
* write with 0 means semaphore release
*/
- if (vmhub == AMDGPU_GFXHUB_0)
- WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+ if (vmhub >= AMDGPU_MMHUB0(0))
+ WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
else
- WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+ WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
}
spin_unlock(&adev->gmc.invalidate_lock);
@@ -994,7 +992,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
i, flush_type);
} else {
gmc_v9_0_flush_gpu_tlb(adev, vmid,
- AMDGPU_GFXHUB_0, flush_type);
+ AMDGPU_GFXHUB(0), flush_type);
}
break;
}
@@ -1060,10 +1058,10 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
uint32_t reg;
/* Do nothing because there's no lut register for mmhub1. */
- if (ring->vm_hub == AMDGPU_MMHUB_1)
+ if (ring->vm_hub == AMDGPU_MMHUB1(0))
return;
- if (ring->vm_hub == AMDGPU_GFXHUB_0)
+ if (ring->vm_hub == AMDGPU_GFXHUB(0))
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
else
reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1947,7 +1945,7 @@ static int gmc_v9_0_hw_init(void *handle)
adev->mmhub.funcs->set_fault_enable_default(adev, value);
}
for (i = 0; i < adev->num_vmhubs; ++i) {
- if (adev->in_s0ix && (i == AMDGPU_GFXHUB_0))
+ if (adev->in_s0ix && (i == AMDGPU_GFXHUB(0)))
continue;
gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
index a3076eb8af6a..71fe7f6f9889 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
@@ -485,7 +485,7 @@ int jpeg_v1_0_sw_init(void *handle)
return r;
ring = &adev->jpeg.inst->ring_dec;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
index 0eddf7c824a7..3a43e42f4834 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c
@@ -86,7 +86,7 @@ static int jpeg_v2_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq,
0, AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
index b040f51d9aa9..259b7ba6a842 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c
@@ -128,9 +128,9 @@ static int jpeg_v2_5_sw_init(void *handle)
ring = &adev->jpeg.inst[i].ring_dec;
ring->use_doorbell = true;
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1 + 8 * i;
sprintf(ring->name, "jpeg_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst[i].irq,
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
index 1c2292cc5f2c..c55386c22311 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c
@@ -101,7 +101,7 @@ static int jpeg_v3_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
index 77e1e64aa1d1..d7d5ffc29393 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c
@@ -108,7 +108,7 @@ static int jpeg_v4_0_sw_init(void *handle)
ring = &adev->jpeg.inst->ring_dec;
ring->use_doorbell = true;
ring->doorbell_index = amdgpu_sriov_vf(adev) ? (((adev->doorbell_index.vcn.vcn_ring0_1) << 1) + 4) : ((adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 1);
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "jpeg_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->jpeg.inst->irq, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
index 4560476c7c31..f1a6abdad21b 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v10_1.c
@@ -149,7 +149,7 @@ static int mes_v10_1_add_hw_queue(struct amdgpu_mes *mes,
{
struct amdgpu_device *adev = mes->adev;
union MESAPI__ADD_QUEUE mes_add_queue_pkt;
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
index 3adb450eec07..9791f3581786 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
@@ -164,7 +164,7 @@ static int mes_v11_0_add_hw_queue(struct amdgpu_mes *mes,
{
struct amdgpu_device *adev = mes->adev;
union MESAPI__ADD_QUEUE mes_add_queue_pkt;
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
uint32_t vm_cntx_cntl = hub->vm_cntx_cntl;
memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt));
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
index 15e7cbeae75b..fb91b31056ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -229,7 +229,7 @@ static void mmhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -285,7 +285,7 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -338,7 +338,7 @@ static int mmhub_v1_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -415,7 +415,7 @@ static void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
index 73afbf2facc9..9086f2fdfaf4 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_7.c
@@ -54,7 +54,7 @@ static u64 mmhub_v1_7_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_7_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -261,7 +261,7 @@ static void mmhub_v1_7_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -319,7 +319,7 @@ static void mmhub_v1_7_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_7_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -348,7 +348,7 @@ static int mmhub_v1_7_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_7_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -425,7 +425,7 @@ static void mmhub_v1_7_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_7_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
index 342d1702104c..9ec06f9db761 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
@@ -53,7 +53,7 @@ static u64 mmhub_v1_8_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v1_8_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -253,7 +253,7 @@ static void mmhub_v1_8_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned num_level, block_size;
uint32_t tmp;
int i;
@@ -311,7 +311,7 @@ static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v1_8_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -352,7 +352,7 @@ static int mmhub_v1_8_gart_enable(struct amdgpu_device *adev)
static void mmhub_v1_8_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -426,7 +426,7 @@ static void mmhub_v1_8_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v1_8_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
index 278e32db878d..8f76c6ecf50a 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_0.c
@@ -187,7 +187,7 @@ mmhub_v2_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET_RLC(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -362,7 +362,7 @@ static void mmhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -412,7 +412,7 @@ static void mmhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v2_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -441,7 +441,7 @@ static int mmhub_v2_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v2_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -520,7 +520,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_0_vmhub_funcs = {
static void mmhub_v2_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
index fcf2813e70db..8bd0fc8d9d25 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v2_3.c
@@ -121,7 +121,7 @@ static void mmhub_v2_3_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, mmMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid, lower_32_bits(page_table_base));
@@ -280,7 +280,7 @@ static void mmhub_v2_3_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -330,7 +330,7 @@ static void mmhub_v2_3_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v2_3_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -373,7 +373,7 @@ static int mmhub_v2_3_gart_enable(struct amdgpu_device *adev)
static void mmhub_v2_3_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -446,7 +446,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v2_3_vmhub_funcs = {
static void mmhub_v2_3_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
index 17a792616979..441379e91cfa 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0.c
@@ -136,7 +136,7 @@ mmhub_v3_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -319,7 +319,7 @@ static void mmhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -369,7 +369,7 @@ static void mmhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -398,7 +398,7 @@ static int mmhub_v3_0_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -477,7 +477,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_vmhub_funcs = {
static void mmhub_v3_0_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
index 26509b6b8c24..12c7f4b46ea9 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_1.c
@@ -138,7 +138,7 @@ static void mmhub_v3_0_1_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -306,7 +306,7 @@ static void mmhub_v3_0_1_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -356,7 +356,7 @@ static void mmhub_v3_0_1_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_1_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -385,7 +385,7 @@ static int mmhub_v3_0_1_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_1_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -459,7 +459,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_1_vmhub_funcs = {
static void mmhub_v3_0_1_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
index 26abbc6a47ab..5dadc85abf7e 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v3_0_2.c
@@ -129,7 +129,7 @@ mmhub_v3_0_2_print_l2_protection_fault_status(struct amdgpu_device *adev,
static void mmhub_v3_0_2_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0, regMMVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -311,7 +311,7 @@ static void mmhub_v3_0_2_disable_identity_aperture(struct amdgpu_device *adev)
static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
int i;
uint32_t tmp;
@@ -361,7 +361,7 @@ static void mmhub_v3_0_2_setup_vmid_config(struct amdgpu_device *adev)
static void mmhub_v3_0_2_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -390,7 +390,7 @@ static int mmhub_v3_0_2_gart_enable(struct amdgpu_device *adev)
static void mmhub_v3_0_2_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i;
@@ -469,7 +469,7 @@ static const struct amdgpu_vmhub_funcs mmhub_v3_0_2_vmhub_funcs = {
static void mmhub_v3_0_2_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
hub->ctx0_ptb_addr_lo32 =
SOC15_REG_OFFSET(MMHUB, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
index 72083e96222f..e790f890aec6 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v9_4.c
@@ -57,7 +57,7 @@ static u64 mmhub_v9_4_get_fb_location(struct amdgpu_device *adev)
static void mmhub_v9_4_setup_hubid_vm_pt_regs(struct amdgpu_device *adev, int hubid,
uint32_t vmid, uint64_t value)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
WREG32_SOC15_OFFSET(MMHUB, 0,
mmVML2VC0_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
@@ -294,7 +294,7 @@ static void mmhub_v9_4_disable_identity_aperture(struct amdgpu_device *adev,
static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned int num_level, block_size;
uint32_t tmp;
int i;
@@ -363,7 +363,7 @@ static void mmhub_v9_4_setup_vmid_config(struct amdgpu_device *adev, int hubid)
static void mmhub_v9_4_program_invalidation(struct amdgpu_device *adev,
int hubid)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
unsigned i;
for (i = 0; i < 18; ++i) {
@@ -404,7 +404,7 @@ static int mmhub_v9_4_gart_enable(struct amdgpu_device *adev)
static void mmhub_v9_4_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+ struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
u32 tmp;
u32 i, j;
@@ -507,8 +507,8 @@ static void mmhub_v9_4_set_fault_enable_default(struct amdgpu_device *adev, bool
static void mmhub_v9_4_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] =
- {&adev->vmhub[AMDGPU_MMHUB_0], &adev->vmhub[AMDGPU_MMHUB_1]};
+ struct amdgpu_vmhub *hub[MMHUB_NUM_INSTANCES] = {
+ &adev->vmhub[AMDGPU_MMHUB0(0)], &adev->vmhub[AMDGPU_MMHUB1(0)]};
int i;
for (i = 0; i < MMHUB_NUM_INSTANCES; i++) {
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 9295ac7edd56..50b6eb9bcfda 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -1825,12 +1825,12 @@ static int sdma_v4_0_sw_init(void *handle)
/*
* On Arcturus, SDMA instance 5~7 has a different vmhub
- * type(AMDGPU_MMHUB_1).
+ * type(AMDGPU_MMHUB1).
*/
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1851,9 +1851,9 @@ static int sdma_v4_0_sw_init(void *handle)
ring->doorbell_index += 0x400;
if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 64dcaa2670dd..7efe7c43fffb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1309,7 +1309,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
/* doorbell size is 2 dwords, get DWORD offset */
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1328,7 +1328,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
*/
ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
ring->doorbell_index += 0x400;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "page%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
index 92e1299be021..a0077cf41295 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
@@ -1389,7 +1389,7 @@ static int sdma_v5_0_sw_init(void *handle)
(adev->doorbell_index.sdma_engine[0] << 1) //get DWORD offset
: (adev->doorbell_index.sdma_engine[1] << 1); // get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
(i == 0) ? AMDGPU_SDMA_IRQ_INSTANCE0 :
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index ca7e8757d78e..efa2c84ee78e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -1253,7 +1253,7 @@ static int sdma_v5_2_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); //get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
AMDGPU_SDMA_IRQ_INSTANCE0 + i,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
index 3d9a80511a45..79d09792d2ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
@@ -1298,7 +1298,7 @@ static int sdma_v6_0_sw_init(void *handle)
ring->doorbell_index =
(adev->doorbell_index.sdma_engine[i] << 1); // get DWORD offset
- ring->vm_hub = AMDGPU_GFXHUB_0;
+ ring->vm_hub = AMDGPU_GFXHUB(0);
sprintf(ring->name, "sdma%d", i);
r = amdgpu_ring_init(adev, ring, 1024,
&adev->sdma.trap_irq,
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
index e32b656b3dab..abaa4463e906 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c
@@ -444,7 +444,7 @@ static int uvd_v7_0_sw_init(void *handle)
continue;
if (!amdgpu_sriov_vf(adev)) {
ring = &adev->uvd.inst[j].ring;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "uvd_%d", ring->me);
r = amdgpu_ring_init(adev, ring, 512,
&adev->uvd.inst[j].irq, 0,
@@ -455,7 +455,7 @@ static int uvd_v7_0_sw_init(void *handle)
for (i = 0; i < adev->uvd.num_enc_rings; ++i) {
ring = &adev->uvd.inst[j].ring_enc[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "uvd_enc_%d.%d", ring->me, i);
if (amdgpu_sriov_vf(adev)) {
ring->use_doorbell = true;
diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
index 57b85bb6a1e4..e0b70cd3b697 100644
--- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c
@@ -466,7 +466,7 @@ static int vce_v4_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(i);
ring = &adev->vce.ring[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vce%d", i);
if (amdgpu_sriov_vf(adev)) {
/* DOORBELL only works under SRIOV */
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
index 761c28fa6ec1..f877c39c7cdd 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c
@@ -120,7 +120,7 @@ static int vcn_v1_0_sw_init(void *handle)
return r;
ring = &adev->vcn.inst->ring_dec;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -142,7 +142,7 @@ static int vcn_v1_0_sw_init(void *handle)
enum amdgpu_ring_priority_level hw_prio = amdgpu_vcn_get_enc_ring_prio(i);
ring = &adev->vcn.inst->ring_enc[i];
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
hw_prio, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
index 7c2b3aa48083..c975aed2f6c7 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_0.c
@@ -129,7 +129,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring->use_doorbell = true;
ring->doorbell_index = adev->doorbell_index.vcn.vcn_ring0_1 << 1;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec");
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst->irq, 0,
@@ -160,7 +160,7 @@ static int vcn_v2_0_sw_init(void *handle)
ring = &adev->vcn.inst->ring_enc[i];
ring->use_doorbell = true;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
if (!amdgpu_sriov_vf(adev))
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + i;
else
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
index ab0b45d0ead1..7044bd7c9f62 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v2_5.c
@@ -188,9 +188,9 @@ static int vcn_v2_5_sw_init(void *handle)
(amdgpu_sriov_vf(adev) ? 2*j : 8*j);
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec_%d", j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[j].irq,
@@ -208,9 +208,9 @@ static int vcn_v2_5_sw_init(void *handle)
(amdgpu_sriov_vf(adev) ? (1 + i + 2*j) : (2 + i + 8*j));
if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(2, 5, 0))
- ring->vm_hub = AMDGPU_MMHUB_1;
+ ring->vm_hub = AMDGPU_MMHUB1(0);
else
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc_%d.%d", j, i);
r = amdgpu_ring_init(adev, ring, 512,
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
index 3eab186261aa..70fefbf26c48 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v3_0.c
@@ -189,7 +189,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 8 * i;
}
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_dec_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
AMDGPU_RING_PRIO_DEFAULT,
@@ -213,7 +213,7 @@ static int vcn_v3_0_sw_init(void *handle)
} else {
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + j + 8 * i;
}
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_enc_%d.%d", i, j);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
hw_prio, &adev->vcn.inst[i].sched_score);
diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
index bf0674039598..81446e6996df 100644
--- a/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/vcn_v4_0.c
@@ -149,7 +149,7 @@ static int vcn_v4_0_sw_init(void *handle)
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + i * (adev->vcn.num_enc_rings + 1) + 1;
else
ring->doorbell_index = (adev->doorbell_index.vcn.vcn_ring0_1 << 1) + 2 + 8 * i;
- ring->vm_hub = AMDGPU_MMHUB_0;
+ ring->vm_hub = AMDGPU_MMHUB0(0);
sprintf(ring->name, "vcn_unified_%d", i);
r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.inst[i].irq, 0,
--
cgit v1.2.3
From b35ce49ab9ca2bc6a59d4441efc5039de80643ce Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Sun, 19 Dec 2021 11:03:59 +0800
Subject: drm/amdgpu: assign register address for vmhub object on each XCD
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Each XCD has its own gfxhub.
v2: switch to the new VMHUB layout
v3: fix mistake
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 58 +++++++++++++++++++-------------
1 file changed, 34 insertions(+), 24 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 9c385ce3a8c4..e5016fea1f28 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -483,33 +483,43 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_2_init(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+ struct amdgpu_vmhub *hub;
+ int i;
- hub->ctx0_ptb_addr_lo32 =
- SOC15_REG_OFFSET(GC, 0,
+ for (i = 0; i < adev->gfx.num_xcd; i++) {
+ hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
+
+ hub->ctx0_ptb_addr_lo32 =
+ SOC15_REG_OFFSET(GC, i,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
- hub->ctx0_ptb_addr_hi32 =
- SOC15_REG_OFFSET(GC, 0,
+ hub->ctx0_ptb_addr_hi32 =
+ SOC15_REG_OFFSET(GC, i,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
- hub->vm_inv_eng0_sem =
- SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_SEM);
- hub->vm_inv_eng0_req =
- SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_REQ);
- hub->vm_inv_eng0_ack =
- SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ACK);
- hub->vm_context0_cntl =
- SOC15_REG_OFFSET(GC, 0, regVM_CONTEXT0_CNTL);
- hub->vm_l2_pro_fault_status =
- SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS);
- hub->vm_l2_pro_fault_cntl =
- SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-
- hub->ctx_distance = regVM_CONTEXT1_CNTL - regVM_CONTEXT0_CNTL;
- hub->ctx_addr_distance = regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
- regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
- hub->eng_distance = regVM_INVALIDATE_ENG1_REQ - regVM_INVALIDATE_ENG0_REQ;
- hub->eng_addr_distance = regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
- regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+ hub->vm_inv_eng0_sem =
+ SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_SEM);
+ hub->vm_inv_eng0_req =
+ SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_REQ);
+ hub->vm_inv_eng0_ack =
+ SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_ACK);
+ hub->vm_context0_cntl =
+ SOC15_REG_OFFSET(GC, i, regVM_CONTEXT0_CNTL);
+ hub->vm_l2_pro_fault_status =
+ SOC15_REG_OFFSET(GC, i,
+ regVM_L2_PROTECTION_FAULT_STATUS);
+ hub->vm_l2_pro_fault_cntl =
+ SOC15_REG_OFFSET(GC, i, regVM_L2_PROTECTION_FAULT_CNTL);
+
+ hub->ctx_distance = regVM_CONTEXT1_CNTL -
+ regVM_CONTEXT0_CNTL;
+ hub->ctx_addr_distance =
+ regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+ regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+ hub->eng_distance = regVM_INVALIDATE_ENG1_REQ -
+ regVM_INVALIDATE_ENG0_REQ;
+ hub->eng_addr_distance =
+ regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+ regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+ }
}
--
cgit v1.2.3
From d25555866172a0454b0dc1374b0ff29e1ae5f676 Mon Sep 17 00:00:00 2001
From: Amber Lin <Amber.Lin@amd.com>
Date: Wed, 23 Mar 2022 11:01:52 -0400
Subject: drm/amdgpu: Set XNACK per process on GC 9.4.3
Set RETRY_PERMISSION_OR_INVALID_PAGE_FAULT bit in VM_CONTEXT1_CNTL
as well so XNACK can be enabled in the SQ per process.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 6 ++++--
drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 2 +-
2 files changed, 5 insertions(+), 3 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index e5016fea1f28..d74621662311 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -328,13 +328,15 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
PAGE_TABLE_BLOCK_SIZE,
block_size);
/* Send no-retry XNACK on fault to suppress VM fault storm.
- * On Aldebaran, XNACK can be enabled in the SQ per-process.
+ * On 9.4.2 and 9.4.3, XNACK can be enabled in
+ * the SQ per-process.
* Retry faults need to be enabled for that to work.
*/
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
!adev->gmc.noretry ||
- adev->asic_type == CHIP_ALDEBARAN);
+ adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
+ adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3));
WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL,
i * hub->ctx_distance, tmp);
WREG32_SOC15_OFFSET(GC, j,
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
index 9ec06f9db761..3883758b7993 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
@@ -288,7 +288,7 @@ static void mmhub_v1_8_setup_vmid_config(struct amdgpu_device *adev)
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
PAGE_TABLE_BLOCK_SIZE,
block_size);
- /* On Aldebaran, XNACK can be enabled in the SQ per-process.
+ /* On 9.4.3, XNACK can be enabled in the SQ per-process.
* Retry faults need to be enabled for that to work.
*/
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
--
cgit v1.2.3
From ed42f2cc3b56955310a16da726886e684ed88432 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Thu, 24 Feb 2022 16:26:07 +0800
Subject: drm/amdgpu: correct the vmhub reference for each XCD in gfxhub init
Correct this though the value is same across different vmhub.
Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index d74621662311..d3424ce97aa8 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -42,10 +42,11 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
uint32_t vmid,
uint64_t page_table_base)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+ struct amdgpu_vmhub *hub;
int i;
for (i = 0; i < adev->gfx.num_xcd; i++) {
+ hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
WREG32_SOC15_OFFSET(GC, i,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
@@ -291,7 +292,7 @@ static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+ struct amdgpu_vmhub *hub;
unsigned num_level, block_size;
uint32_t tmp;
int i, j;
@@ -304,6 +305,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
block_size -= 9;
for (j = 0; j < adev->gfx.num_xcd; j++) {
+ hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0; i <= 14; i++) {
tmp = RREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL, i);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
@@ -359,10 +361,11 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+ struct amdgpu_vmhub *hub;
unsigned i, j;
for (j = 0; j < adev->gfx.num_xcd; j++) {
+ hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0 ; i < 18; ++i) {
WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
i * hub->eng_addr_distance, 0xffffffff);
@@ -408,11 +411,12 @@ static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
{
- struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+ struct amdgpu_vmhub *hub;
u32 tmp;
u32 i, j;
for (j = 0; j < adev->gfx.num_xcd; j++) {
+ hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
/* Disable all tables */
for (i = 0; i < 16; i++)
WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT0_CNTL,
--
cgit v1.2.3
From 9eb7681f760c77adece36bc62953245c9f44a3be Mon Sep 17 00:00:00 2001
From: Shiwu Zhang <shiwu.zhang@amd.com>
Date: Mon, 21 Feb 2022 15:38:39 +0800
Subject: drm/amdgpu: add the support of XGMI link for GC 9.4.3
Add the xgmi LFB_CNTL/LBF_SIZE reg addresses to fetch the xgmi info from.
v2: move get_xgmi_info() to GC_V9_4_3 sepecific source files to utilize
the register definitions specific for GC_V9_4_3
v3: remove the duplicated register definitions
v4: enable xgmi based on asic_type as XGMI_IP ver is not available
yet for IP discovery
Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Ack-by: Lijo Lazar <Lijo.Lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 41 +++++++++++++++++++++++++++++++-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 10 +++++---
2 files changed, 47 insertions(+), 4 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index d3424ce97aa8..1bb17d95f720 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -528,6 +528,45 @@ static void gfxhub_v1_2_init(struct amdgpu_device *adev)
}
}
+static int gfxhub_v1_2_get_xgmi_info(struct amdgpu_device *adev)
+{
+ u32 max_num_physical_nodes;
+ u32 max_physical_node_id;
+ u32 xgmi_lfb_cntl;
+ u32 max_region;
+ u64 seg_size;
+
+ xgmi_lfb_cntl = RREG32_SOC15(GC, 0, regMC_VM_XGMI_LFB_CNTL);
+ seg_size = REG_GET_FIELD(
+ RREG32_SOC15(GC, 0, regMC_VM_XGMI_LFB_SIZE),
+ MC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
+ max_region =
+ REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
+
+
+
+ max_num_physical_nodes = 8;
+ max_physical_node_id = 7;
+
+ /* PF_MAX_REGION=0 means xgmi is disabled */
+ if (max_region || adev->gmc.xgmi.connected_to_cpu) {
+ adev->gmc.xgmi.num_physical_nodes = max_region + 1;
+
+ if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes)
+ return -EINVAL;
+
+ adev->gmc.xgmi.physical_node_id =
+ REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL,
+ PF_LFB_REGION);
+
+ if (adev->gmc.xgmi.physical_node_id > max_physical_node_id)
+ return -EINVAL;
+
+ adev->gmc.xgmi.node_segment_size = seg_size;
+ }
+
+ return 0;
+}
const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
.get_mc_fb_offset = gfxhub_v1_2_get_mc_fb_offset,
@@ -536,5 +575,5 @@ const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
.gart_disable = gfxhub_v1_2_gart_disable,
.set_fault_enable_default = gfxhub_v1_2_set_fault_enable_default,
.init = gfxhub_v1_2_init,
- .get_xgmi_info = gfxhub_v1_1_get_xgmi_info,
+ .get_xgmi_info = gfxhub_v1_2_get_xgmi_info,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 59be0c0293c4..4b2c4ecd7253 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -1419,9 +1419,13 @@ static int gmc_v9_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- /* ARCT and VEGA20 don't have XGMI defined in their IP discovery tables */
- if (adev->asic_type == CHIP_VEGA20 ||
- adev->asic_type == CHIP_ARCTURUS)
+ /*
+ * 9.4.0, 9.4.1 and 9.4.3 don't have XGMI defined
+ * in their IP discovery tables
+ */
+ if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0) ||
+ adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
+ adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
adev->gmc.xgmi.supported = true;
if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(6, 1, 0)) {
--
cgit v1.2.3
From 8078f1c610fdcdd8003e2c538fb04af41fa5c269 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Wed, 29 Jun 2022 11:41:53 +0530
Subject: drm/amdgpu: Change num_xcd to xcc_mask
Instead of number of XCCs, keep a mask of XCCs for the exact XCCs
available on the ASIC. XCC configuration could differ based on
different ASIC configs.
v2:
Rename num_xcd to num_xcc (Hawking)
Use smaller xcc_mask size, changed to u16 (Le)
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 21 ++---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 4 +-
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 133 ++++++++++++++++++-------------
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 67 ++++++++++------
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 3 +-
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 10 +--
7 files changed, 141 insertions(+), 99 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 76438f197de1..069b259f384c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -209,12 +209,12 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
int max_queues_per_mec = min(adev->gfx.mec.num_pipe_per_mec *
adev->gfx.mec.num_queue_per_pipe,
adev->gfx.num_compute_rings);
- int num_xcd = (adev->gfx.num_xcd > 1) ? adev->gfx.num_xcd : 1;
+ int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
if (multipipe_policy) {
/* policy: make queues evenly cross all pipes on MEC1 only
* for multiple xcc, just use the original policy for simplicity */
- for (j = 0; j < num_xcd; j++) {
+ for (j = 0; j < num_xcc; j++) {
for (i = 0; i < max_queues_per_mec; i++) {
pipe = i % adev->gfx.mec.num_pipe_per_mec;
queue = (i / adev->gfx.mec.num_pipe_per_mec) %
@@ -226,13 +226,13 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
}
} else {
/* policy: amdgpu owns all queues in the given pipe */
- for (j = 0; j < num_xcd; j++) {
+ for (j = 0; j < num_xcc; j++) {
for (i = 0; i < max_queues_per_mec; ++i)
set_bit(i, adev->gfx.mec_bitmap[j].queue_bitmap);
}
}
- for (j = 0; j < num_xcd; j++) {
+ for (j = 0; j < num_xcc; j++) {
dev_dbg(adev->dev, "mec queue bitmap weight=%d\n",
bitmap_weight(adev->gfx.mec_bitmap[j].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
}
@@ -1207,23 +1207,24 @@ static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
struct drm_device *ddev = dev_get_drvdata(dev);
struct amdgpu_device *adev = drm_to_adev(ddev);
enum amdgpu_gfx_partition mode;
- int ret;
+ int ret = 0, num_xcc;
- if (adev->gfx.num_xcd % 2 != 0)
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ if (num_xcc % 2 != 0)
return -EINVAL;
if (!strncasecmp("SPX", buf, strlen("SPX"))) {
mode = AMDGPU_SPX_PARTITION_MODE;
} else if (!strncasecmp("DPX", buf, strlen("DPX"))) {
- if (adev->gfx.num_xcd != 4 || adev->gfx.num_xcd != 8)
+ if (num_xcc != 4 || num_xcc != 8)
return -EINVAL;
mode = AMDGPU_DPX_PARTITION_MODE;
} else if (!strncasecmp("TPX", buf, strlen("TPX"))) {
- if (adev->gfx.num_xcd != 6)
+ if (num_xcc != 6)
return -EINVAL;
mode = AMDGPU_TPX_PARTITION_MODE;
} else if (!strncasecmp("QPX", buf, strlen("QPX"))) {
- if (adev->gfx.num_xcd != 8)
+ if (num_xcc != 8)
return -EINVAL;
mode = AMDGPU_QPX_PARTITION_MODE;
} else if (!strncasecmp("CPX", buf, strlen("CPX"))) {
@@ -1253,7 +1254,7 @@ static ssize_t amdgpu_gfx_get_available_compute_partition(struct device *dev,
char *supported_partition;
/* TBD */
- switch (adev->gfx.num_xcd) {
+ switch (NUM_XCC(adev->gfx.xcc_mask)) {
case 8:
supported_partition = "SPX, DPX, QPX, CPX";
break;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index 8df36527aee9..93f9875154db 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -64,6 +64,8 @@ enum amdgpu_gfx_partition {
AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE,
};
+#define NUM_XCC(x) hweight16(x)
+
struct amdgpu_mec {
struct amdgpu_bo *hpd_eop_obj;
u64 hpd_eop_gpu_addr;
@@ -396,7 +398,7 @@ struct amdgpu_gfx {
bool cp_gfx_shadow; /* for gfx11 */
enum amdgpu_gfx_partition partition_mode;
- uint32_t num_xcd;
+ uint16_t xcc_mask;
uint32_t num_xcc_per_xcp;
struct mutex partition_mutex;
};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index 91814dc083c9..da69177dc76f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -4536,7 +4536,7 @@ static int gfx_v9_0_early_init(void *handle)
adev->gfx.num_gfx_rings = 0;
else
adev->gfx.num_gfx_rings = GFX9_NUM_GFX_RINGS;
- adev->gfx.num_xcd = 1;
+ adev->gfx.xcc_mask = 1;
adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
AMDGPU_MAX_COMPUTE_RINGS);
gfx_v9_0_set_kiq_pm4_funcs(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 52185b1d5d31..c776fc5884de 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -177,16 +177,19 @@ static const struct kiq_pm4_funcs gfx_v9_4_3_kiq_pm4_funcs = {
static void gfx_v9_4_3_set_kiq_pm4_funcs(struct amdgpu_device *adev)
{
- int i;
- for (i = 0; i < adev->gfx.num_xcd; i++)
+ int i, num_xcc;
+
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++)
adev->gfx.kiq[i].pmf = &gfx_v9_4_3_kiq_pm4_funcs;
}
static void gfx_v9_4_3_init_golden_registers(struct amdgpu_device *adev)
{
- int i;
+ int i, num_xcc;
- for (i = 2; i < adev->gfx.num_xcd; i++)
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 2; i < num_xcc; i++)
WREG32_SOC15(GC, i, regGRBM_MCM_ADDR, 0x4);
}
@@ -499,7 +502,7 @@ static void gfx_v9_4_3_mec_fini(struct amdgpu_device *adev)
static int gfx_v9_4_3_mec_init(struct amdgpu_device *adev)
{
- int r, i;
+ int r, i, num_xcc;
u32 *hpd;
const __le32 *fw_data;
unsigned fw_size;
@@ -508,7 +511,8 @@ static int gfx_v9_4_3_mec_init(struct amdgpu_device *adev)
const struct gfx_firmware_header_v1_0 *mec_hdr;
- for (i = 0; i < adev->gfx.num_xcd; i++)
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++)
bitmap_zero(adev->gfx.mec_bitmap[i].queue_bitmap,
AMDGPU_MAX_COMPUTE_QUEUES);
@@ -683,23 +687,24 @@ static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
enum amdgpu_gfx_partition mode)
{
u32 tmp = 0;
- int num_xcc_per_partition, i;
+ int num_xcc_per_partition, i, num_xcc;
if (mode == adev->gfx.partition_mode)
return mode;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
switch (mode) {
case AMDGPU_SPX_PARTITION_MODE:
- num_xcc_per_partition = adev->gfx.num_xcd;
+ num_xcc_per_partition = num_xcc;
break;
case AMDGPU_DPX_PARTITION_MODE:
- num_xcc_per_partition = adev->gfx.num_xcd / 2;
+ num_xcc_per_partition = num_xcc / 2;
break;
case AMDGPU_TPX_PARTITION_MODE:
- num_xcc_per_partition = adev->gfx.num_xcd / 3;
+ num_xcc_per_partition = num_xcc / 3;
break;
case AMDGPU_QPX_PARTITION_MODE:
- num_xcc_per_partition = adev->gfx.num_xcd / 4;
+ num_xcc_per_partition = num_xcc / 4;
break;
case AMDGPU_CPX_PARTITION_MODE:
num_xcc_per_partition = 1;
@@ -712,7 +717,7 @@ static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
* Stop user queues and threads, and make sure GPU is empty of work.
*/
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ for (i = 0; i < num_xcc; i++) {
tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, NUM_XCC_IN_XCP,
num_xcc_per_partition);
tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, VIRTUAL_XCC_ID,
@@ -836,7 +841,7 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
static int gfx_v9_4_3_sw_init(void *handle)
{
- int i, j, k, r, ring_id, xcc_id;
+ int i, j, k, r, ring_id, xcc_id, num_xcc;
struct amdgpu_kiq *kiq;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -844,6 +849,8 @@ static int gfx_v9_4_3_sw_init(void *handle)
adev->gfx.mec.num_pipe_per_mec = 4;
adev->gfx.mec.num_queue_per_pipe = 8;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
/* EOP Event */
r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_EOP_INTERRUPT, &adev->gfx.eop_irq);
if (r)
@@ -877,8 +884,7 @@ static int gfx_v9_4_3_sw_init(void *handle)
/* set up the compute queues - allocate horizontally across pipes */
ring_id = 0;
- for (xcc_id = 0; xcc_id < adev->gfx.num_xcd; xcc_id++) {
-
+ for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
for (k = 0; k < adev->gfx.mec.num_pipe_per_mec;
@@ -930,14 +936,14 @@ static int gfx_v9_4_3_sw_init(void *handle)
static int gfx_v9_4_3_sw_fini(void *handle)
{
- int i;
+ int i, num_xcc;
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- for (i = 0; i < adev->gfx.num_compute_rings *
- adev->gfx.num_xcd; i++)
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < adev->gfx.num_compute_rings * num_xcc; i++)
amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ for (i = 0; i < num_xcc; i++) {
amdgpu_gfx_mqd_sw_fini(adev, i);
amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[i].ring);
amdgpu_gfx_kiq_fini(adev, i);
@@ -1050,9 +1056,10 @@ static void gfx_v9_4_3_init_gds_vmid(struct amdgpu_device *adev, int xcc_id)
static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
{
u32 tmp;
- int i, j;
+ int i, j, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
WREG32_FIELD15_PREREG(GC, i, GRBM_CNTL, READ_TIMEOUT, 0xff);
gfx_v9_4_3_setup_rb(adev, i);
}
@@ -1064,7 +1071,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
- for (j = 0; j < adev->gfx.num_xcd; j++) {
+ for (j = 0; j < num_xcc; j++) {
soc15_grbm_select(adev, 0, 0, 0, i, j);
/* CP and shaders */
if (i == 0) {
@@ -1092,7 +1099,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
mutex_unlock(&adev->srbm_mutex);
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ for (i = 0; i < num_xcc; i++) {
gfx_v9_4_3_init_compute_vmid(adev, i);
gfx_v9_4_3_init_gds_vmid(adev, i);
}
@@ -1150,8 +1157,10 @@ static void gfx_v9_4_3_disable_gpa_mode(struct amdgpu_device *adev, int xcc_id)
static void gfx_v9_4_3_program_xcc_id(struct amdgpu_device *adev, int xcc_id)
{
uint32_t tmp = 0;
+ int num_xcc;
- switch (adev->gfx.num_xcd) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ switch (num_xcc) {
/* directly config VIRTUAL_XCC_ID to 0 for 1-XCC */
case 1:
WREG32_SOC15(GC, xcc_id, regCP_HYP_XCP_CTL, 0x8);
@@ -1288,9 +1297,10 @@ static void gfx_v9_4_3_enable_gui_idle_interrupt(struct amdgpu_device *adev,
static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
{
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
WREG32_FIELD15_PREREG(GC, i, RLC_CNTL, RLC_ENABLE_F32, 0);
gfx_v9_4_3_enable_gui_idle_interrupt(adev, false, i);
gfx_v9_4_3_wait_for_rlc_serdes(adev, i);
@@ -1299,9 +1309,10 @@ static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
static void gfx_v9_4_3_rlc_reset(struct amdgpu_device *adev)
{
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
WREG32_FIELD15_PREREG(GC, i, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
udelay(50);
WREG32_FIELD15_PREREG(GC, i, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
@@ -1314,9 +1325,10 @@ static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
#ifdef AMDGPU_RLC_DEBUG_RETRY
u32 rlc_ucode_ver;
#endif
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
WREG32_FIELD15_PREREG(GC, i, RLC_CNTL, RLC_ENABLE_F32, 1);
udelay(50);
@@ -1377,11 +1389,12 @@ static int gfx_v9_4_3_rlc_load_microcode(struct amdgpu_device *adev, int xcc_id)
static int gfx_v9_4_3_rlc_resume(struct amdgpu_device *adev)
{
- int r, i;
+ int r, i, num_xcc;
adev->gfx.rlc.funcs->stop(adev);
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
/* disable CG */
WREG32_SOC15(GC, i, regRLC_CGCG_CGLS_CTRL, 0);
@@ -1954,10 +1967,11 @@ done:
static int gfx_v9_4_3_cp_resume(struct amdgpu_device *adev)
{
- int r, i, j;
+ int r, i, j, num_xcc;
struct amdgpu_ring *ring;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
gfx_v9_4_3_enable_gui_idle_interrupt(adev, false, i);
if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
@@ -2021,12 +2035,13 @@ static int gfx_v9_4_3_hw_init(void *handle)
static int gfx_v9_4_3_hw_fini(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- int i;
+ int i, num_xcc;
amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
if (amdgpu_gfx_disable_kcq(adev, i))
DRM_ERROR("XCD %d KCQ disable failed\n", i);
@@ -2069,9 +2084,10 @@ static int gfx_v9_4_3_resume(void *handle)
static bool gfx_v9_4_3_is_idle(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
if (REG_GET_FIELD(RREG32_SOC15(GC, i, regGRBM_STATUS),
GRBM_STATUS, GUI_ACTIVE))
return false;
@@ -2183,30 +2199,30 @@ static void gfx_v9_4_3_ring_emit_gds_switch(struct amdgpu_ring *ring,
static int gfx_v9_4_3_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ int num_xcc;
- /* hardcode in emulation phase */
- adev->gfx.num_xcd = 1;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
adev->gfx.partition_mode = amdgpu_user_partt_mode;
/* calculate the num_xcc_in_xcp for the partition mode*/
switch (amdgpu_user_partt_mode) {
case AMDGPU_SPX_PARTITION_MODE:
- adev->gfx.num_xcc_per_xcp = adev->gfx.num_xcd;
+ adev->gfx.num_xcc_per_xcp = num_xcc;
break;
case AMDGPU_DPX_PARTITION_MODE:
- adev->gfx.num_xcc_per_xcp = adev->gfx.num_xcd / 2;
+ adev->gfx.num_xcc_per_xcp = num_xcc / 2;
break;
case AMDGPU_TPX_PARTITION_MODE:
- adev->gfx.num_xcc_per_xcp = adev->gfx.num_xcd / 3;
+ adev->gfx.num_xcc_per_xcp = num_xcc / 3;
break;
case AMDGPU_QPX_PARTITION_MODE:
- adev->gfx.num_xcc_per_xcp = adev->gfx.num_xcd / 4;
+ adev->gfx.num_xcc_per_xcp = num_xcc / 4;
break;
case AMDGPU_CPX_PARTITION_MODE:
adev->gfx.num_xcc_per_xcp = 1;
break;
default:
- adev->gfx.num_xcc_per_xcp = adev->gfx.num_xcd;
+ adev->gfx.num_xcc_per_xcp = num_xcc;
break;
}
@@ -2404,14 +2420,15 @@ static int gfx_v9_4_3_set_clockgating_state(void *handle,
enum amd_clockgating_state state)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
- int i;
+ int i, num_xcc;
if (amdgpu_sriov_vf(adev))
return 0;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
switch (adev->ip_versions[GC_HWIP][0]) {
case IP_VERSION(9, 4, 3):
- for (i = 0; i < adev->gfx.num_xcd; i++)
+ for (i = 0; i < num_xcc; i++)
gfx_v9_4_3_update_gfx_clock_gating(adev,
state == AMD_CG_STATE_GATE, i);
break;
@@ -2739,12 +2756,13 @@ static int gfx_v9_4_3_set_priv_reg_fault_state(struct amdgpu_device *adev,
unsigned type,
enum amdgpu_interrupt_state state)
{
- int i;
+ int i, num_xcc;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
case AMDGPU_IRQ_STATE_ENABLE:
- for (i = 0; i < adev->gfx.num_xcd; i++)
+ for (i = 0; i < num_xcc; i++)
WREG32_FIELD15_PREREG(GC, i, CP_INT_CNTL_RING0,
PRIV_REG_INT_ENABLE,
state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
@@ -2761,12 +2779,13 @@ static int gfx_v9_4_3_set_priv_inst_fault_state(struct amdgpu_device *adev,
unsigned type,
enum amdgpu_interrupt_state state)
{
- int i;
+ int i, num_xcc;
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
case AMDGPU_IRQ_STATE_ENABLE:
- for (i = 0; i < adev->gfx.num_xcd; i++)
+ for (i = 0; i < num_xcc; i++)
WREG32_FIELD15_PREREG(GC, i, CP_INT_CNTL_RING0,
PRIV_INSTR_INT_ENABLE,
state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
@@ -2783,8 +2802,10 @@ static int gfx_v9_4_3_set_eop_interrupt_state(struct amdgpu_device *adev,
unsigned type,
enum amdgpu_interrupt_state state)
{
- int i;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ int i, num_xcc;
+
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
switch (type) {
case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
gfx_v9_4_3_set_compute_eop_interrupt_state(adev, 1, 0, state, i);
@@ -2842,6 +2863,7 @@ static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
/* Per-queue interrupt is supported for MEC starting from VI.
* The interrupt can only be enabled/disabled per pipe instead of per queue.
*/
+
if ((ring->me == me_id) && (ring->pipe == pipe_id) && (ring->queue == queue_id))
amdgpu_fence_process(ring);
}
@@ -3056,9 +3078,10 @@ static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_kiq = {
static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev)
{
- int i, j;
+ int i, j, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
adev->gfx.kiq[i].ring.funcs = &gfx_v9_4_3_ring_funcs_kiq;
for (j = 0; j < adev->gfx.num_compute_rings; j++)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 1bb17d95f720..e35365ab3f1f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -43,9 +43,10 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
uint64_t page_table_base)
{
struct amdgpu_vmhub *hub;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
WREG32_SOC15_OFFSET(GC, i,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
@@ -56,13 +57,14 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
hub->ctx_addr_distance * vmid,
upper_32_bits(page_table_base));
+
}
}
static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
{
uint64_t pt_base;
- int i;
+ int i, num_xcc;
if (adev->gmc.pdb0_bo)
pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
@@ -74,7 +76,8 @@ static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
/* If use GART for FB translation, vmid0 page table covers both
* vram and system memory (gart)
*/
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
if (adev->gmc.pdb0_bo) {
WREG32_SOC15(GC, i,
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
@@ -111,9 +114,10 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
{
uint64_t value;
uint32_t tmp;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
/* Program the AGP BAR */
WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BASE, 0);
WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
@@ -177,9 +181,10 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
{
uint32_t tmp;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
/* Setup TLB control */
tmp = RREG32_SOC15(GC, i, regMC_VM_MX_L1_TLB_CNTL);
@@ -202,9 +207,10 @@ static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
{
uint32_t tmp;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
/* Setup L2 cache */
tmp = RREG32_SOC15(GC, i, regVM_L2_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
@@ -249,9 +255,10 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
{
uint32_t tmp;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
tmp = RREG32_SOC15(GC, i, regVM_CONTEXT0_CNTL);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
@@ -266,9 +273,10 @@ static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
{
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
WREG32_SOC15(GC, i,
regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
0XFFFFFFFF);
@@ -295,7 +303,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
struct amdgpu_vmhub *hub;
unsigned num_level, block_size;
uint32_t tmp;
- int i, j;
+ int i, j, num_xcc;
num_level = adev->vm_manager.num_level;
block_size = adev->vm_manager.block_size;
@@ -304,7 +312,8 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
else
block_size -= 9;
- for (j = 0; j < adev->gfx.num_xcd; j++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (j = 0; j < num_xcc; j++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0; i <= 14; i++) {
tmp = RREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL, i);
@@ -362,10 +371,12 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
{
struct amdgpu_vmhub *hub;
- unsigned i, j;
+ unsigned i, j, num_xcc;
- for (j = 0; j < adev->gfx.num_xcd; j++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (j = 0; j < num_xcc; j++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+
for (i = 0 ; i < 18; ++i) {
WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
i * hub->eng_addr_distance, 0xffffffff);
@@ -377,9 +388,10 @@ static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
{
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
if (amdgpu_sriov_vf(adev)) {
/*
* MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
@@ -413,9 +425,10 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
{
struct amdgpu_vmhub *hub;
u32 tmp;
- u32 i, j;
+ u32 i, j, num_xcc;
- for (j = 0; j < adev->gfx.num_xcd; j++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (j = 0; j < num_xcc; j++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
/* Disable all tables */
for (i = 0; i < 16; i++)
@@ -449,9 +462,10 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
bool value)
{
u32 tmp;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
tmp = RREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -490,9 +504,10 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
static void gfxhub_v1_2_init(struct amdgpu_device *adev)
{
struct amdgpu_vmhub *hub;
- int i;
+ int i, num_xcc;
- for (i = 0; i < adev->gfx.num_xcd; i++) {
+ num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+ for (i = 0; i < num_xcc; i++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
hub->ctx0_ptb_addr_lo32 =
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 4b2c4ecd7253..2c322a25bf1c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -1733,7 +1733,8 @@ static int gmc_v9_0_sw_init(void *handle)
adev->gmc.translate_further = adev->vm_manager.num_level > 1;
break;
case IP_VERSION(9, 4, 3):
- bitmap_set(adev->vmhubs_mask, AMDGPU_GFXHUB(0), adev->gfx.num_xcd);
+ bitmap_set(adev->vmhubs_mask, AMDGPU_GFXHUB(0),
+ NUM_XCC(adev->gfx.xcc_mask));
bitmap_set(adev->vmhubs_mask, AMDGPU_MMHUB0(0), adev->num_aid);
amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 829e32433faf..df96c4c508a0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -592,6 +592,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
struct kfd_node *node;
uint32_t first_vmid_kfd, last_vmid_kfd, vmid_num_kfd;
unsigned int max_proc_per_quantum;
+ int num_xcd;
kfd->mec_fw_version = amdgpu_amdkfd_get_fw_version(kfd->adev,
KGD_ENGINE_MEC1);
@@ -601,16 +602,15 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
KGD_ENGINE_SDMA1);
kfd->shared_resources = *gpu_resources;
- if (kfd->adev->gfx.num_xcd == 0 || kfd->adev->gfx.num_xcd == 1 ||
- kfd->adev->gfx.num_xcc_per_xcp == 0)
+ num_xcd = NUM_XCC(kfd->adev->gfx.xcc_mask);
+ if (num_xcd == 0 || num_xcd == 1 || kfd->adev->gfx.num_xcc_per_xcp == 0)
kfd->num_nodes = 1;
else
- kfd->num_nodes =
- kfd->adev->gfx.num_xcd/kfd->adev->gfx.num_xcc_per_xcp;
+ kfd->num_nodes = num_xcd / kfd->adev->gfx.num_xcc_per_xcp;
if (kfd->num_nodes == 0) {
dev_err(kfd_device,
"KFD num nodes cannot be 0, GC inst: %d, num_xcc_in_node: %d\n",
- kfd->adev->gfx.num_xcd, kfd->adev->gfx.num_xcc_per_xcp);
+ num_xcd, kfd->adev->gfx.num_xcc_per_xcp);
goto out;
}
--
cgit v1.2.3
From 659a4ab8e27eb39cc61cb74cc714ba1a8f8c9a61 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Tue, 5 Jul 2022 09:56:41 +0530
Subject: drm/amdgpu: Use instance lookup table for GC 9.4.3
Register accesses need to be based on physical instance on bare metal.
Pass the right instance using logical to physical instance lookup
table before accessing registers. Add a macro GET_INST to get the right
physical instance of an IP corresponding to a logical instance.
v2: fix gfx_v9_4_3_check_rlcg_range() (Alex)
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c | 20 +-
drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 381 +++++++++++----------
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 146 ++++----
drivers/gpu/drm/amd/amdgpu/soc15_common.h | 3 +
4 files changed, 279 insertions(+), 271 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
index e81bdca53f42..5daec0b45545 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -279,8 +279,8 @@ static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
/* HQD registers extend to CP_HQD_AQL_DISPATCH_ID_HI */
mqd_hqd = &m->cp_mqd_base_addr_lo;
- hqd_base = SOC15_REG_OFFSET(GC, inst, regCP_MQD_BASE_ADDR);
- hqd_end = SOC15_REG_OFFSET(GC, inst, regCP_HQD_AQL_DISPATCH_ID_HI);
+ hqd_base = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_MQD_BASE_ADDR);
+ hqd_end = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_AQL_DISPATCH_ID_HI);
for (reg = hqd_base; reg <= hqd_end; reg++)
WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
@@ -289,7 +289,7 @@ static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
/* Activate doorbell logic before triggering WPTR poll. */
data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_PQ_DOORBELL_CONTROL),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_DOORBELL_CONTROL),
data);
if (wptr) {
@@ -319,27 +319,27 @@ static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_PQ_WPTR_LO),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_LO),
lower_32_bits(guessed_wptr));
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_PQ_WPTR_HI),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_HI),
upper_32_bits(guessed_wptr));
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_PQ_WPTR_POLL_ADDR),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR),
lower_32_bits((uintptr_t)wptr));
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst,
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
upper_32_bits((uintptr_t)wptr));
- WREG32(SOC15_REG_OFFSET(GC, inst, regCP_PQ_WPTR_POLL_CNTL1),
+ WREG32(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_PQ_WPTR_POLL_CNTL1),
(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id,
queue_id));
}
/* Start the EOP fetcher */
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_EOP_RPTR),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_EOP_RPTR),
REG_SET_FIELD(m->cp_hqd_eop_rptr,
CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
- WREG32_RLC(SOC15_REG_OFFSET(GC, inst, regCP_HQD_ACTIVE), data);
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_ACTIVE), data);
kgd_gfx_v9_release_queue(adev, inst);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 47d8ac64e877..e9c12b4970f9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -190,7 +190,7 @@ static void gfx_v9_4_3_init_golden_registers(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 2; i < num_xcc; i++)
- WREG32_SOC15(GC, i, regGRBM_MCM_ADDR, 0x4);
+ WREG32_SOC15(GC, GET_INST(GC, i), regGRBM_MCM_ADDR, 0x4);
}
static void gfx_v9_4_3_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
@@ -236,7 +236,7 @@ static int gfx_v9_4_3_ring_test_ring(struct amdgpu_ring *ring)
/* scratch_reg0_offset is 32bit even with full XCD config */
uint32_t scratch_reg0_offset;
- scratch_reg0_offset = SOC15_REG_OFFSET(GC, ring->xcc_id, regSCRATCH_REG0);
+ scratch_reg0_offset = SOC15_REG_OFFSET(GC, GET_INST(GC, ring->xcc_id), regSCRATCH_REG0);
WREG32(scratch_reg0_offset, 0xCAFEDEAD);
r = amdgpu_ring_alloc(ring, 3);
@@ -324,9 +324,9 @@ static uint64_t gfx_v9_4_3_get_gpu_clock_counter(struct amdgpu_device *adev)
amdgpu_gfx_off_ctrl(adev, false);
mutex_lock(&adev->gfx.gpu_clock_mutex);
- WREG32_SOC15(GC, 0, regRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
- clock = (uint64_t)RREG32_SOC15(GC, 0, regRLC_GPU_CLOCK_COUNT_LSB) |
- ((uint64_t)RREG32_SOC15(GC, 0, regRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+ WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+ clock = (uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_LSB) |
+ ((uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
mutex_unlock(&adev->gfx.gpu_clock_mutex);
amdgpu_gfx_off_ctrl(adev, true);
@@ -598,24 +598,24 @@ static void gfx_v9_4_3_select_se_sh(struct amdgpu_device *adev,
else
data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num);
- WREG32_SOC15_RLC_SHADOW_EX(reg, GC, xcc_id, regGRBM_GFX_INDEX, data);
+ WREG32_SOC15_RLC_SHADOW_EX(reg, GC, GET_INST(GC, xcc_id), regGRBM_GFX_INDEX, data);
}
static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address)
{
- WREG32_SOC15_RLC(GC, 0, regSQ_IND_INDEX,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, 0), regSQ_IND_INDEX,
(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
(address << SQ_IND_INDEX__INDEX__SHIFT) |
(SQ_IND_INDEX__FORCE_READ_MASK));
- return RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+ return RREG32_SOC15(GC, GET_INST(GC, 0), regSQ_IND_DATA);
}
static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
uint32_t wave, uint32_t thread,
uint32_t regno, uint32_t num, uint32_t *out)
{
- WREG32_SOC15_RLC(GC, 0, regSQ_IND_INDEX,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, 0), regSQ_IND_INDEX,
(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
(regno << SQ_IND_INDEX__INDEX__SHIFT) |
@@ -623,7 +623,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
(SQ_IND_INDEX__FORCE_READ_MASK) |
(SQ_IND_INDEX__AUTO_INCR_MASK));
while (num--)
- *(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+ *(out++) = RREG32_SOC15(GC, GET_INST(GC, 0), regSQ_IND_DATA);
}
static void gfx_v9_4_3_read_wave_data(struct amdgpu_device *adev,
@@ -669,7 +669,7 @@ static void gfx_v9_4_3_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd
static void gfx_v9_4_3_select_me_pipe_q(struct amdgpu_device *adev,
u32 me, u32 pipe, u32 q, u32 vm)
{
- soc15_grbm_select(adev, me, pipe, q, vm, 0);
+ soc15_grbm_select(adev, me, pipe, q, vm, GET_INST(GC, 0));
}
static enum amdgpu_gfx_partition
@@ -719,7 +719,7 @@ static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
num_xcc_per_partition);
tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, VIRTUAL_XCC_ID,
i % num_xcc_per_partition);
- WREG32_SOC15(GC, i, regCP_HYP_XCP_CTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, i), regCP_HYP_XCP_CTL, tmp);
}
if (adev->nbio.funcs->set_compute_partition_mode)
@@ -755,7 +755,7 @@ static int gfx_v9_4_3_gpu_early_init(struct amdgpu_device *adev)
adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
- gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG);
+ gb_addr_config = RREG32_SOC15(GC, GET_INST(GC, 0), regGB_ADDR_CONFIG);
break;
default:
BUG();
@@ -957,8 +957,8 @@ static u32 gfx_v9_4_3_get_rb_active_bitmap(struct amdgpu_device *adev)
{
u32 data, mask;
- data = RREG32_SOC15(GC, 0, regCC_RB_BACKEND_DISABLE);
- data |= RREG32_SOC15(GC, 0, regGC_USER_RB_BACKEND_DISABLE);
+ data = RREG32_SOC15(GC, GET_INST(GC, 0), regCC_RB_BACKEND_DISABLE);
+ data |= RREG32_SOC15(GC, GET_INST(GC, 0), regGC_USER_RB_BACKEND_DISABLE);
data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK;
data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT;
@@ -1014,21 +1014,21 @@ static void gfx_v9_4_3_init_compute_vmid(struct amdgpu_device *adev, int xcc_id)
mutex_lock(&adev->srbm_mutex);
for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
- soc15_grbm_select(adev, 0, 0, 0, i, xcc_id);
+ soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, xcc_id));
/* CP and shaders */
- WREG32_SOC15_RLC(GC, xcc_id, regSH_MEM_CONFIG, sh_mem_config);
- WREG32_SOC15_RLC(GC, xcc_id, regSH_MEM_BASES, sh_mem_bases);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_CONFIG, sh_mem_config);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_BASES, sh_mem_bases);
}
- soc15_grbm_select(adev, 0, 0, 0, 0, xcc_id);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
mutex_unlock(&adev->srbm_mutex);
/* Initialize all compute VMIDs to have no GDS, GWS, or OA
acccess. These should be enabled by FW for target VMIDs. */
for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_VMID0_BASE, 2 * i, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_VMID0_SIZE, 2 * i, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_GWS_VMID0, i, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_OA_VMID0, i, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * i, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * i, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, i, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, i, 0);
}
}
@@ -1043,10 +1043,10 @@ static void gfx_v9_4_3_init_gds_vmid(struct amdgpu_device *adev, int xcc_id)
* access so that HWS firmware can save/restore entries.
*/
for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) {
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_VMID0_BASE, 2 * vmid, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_VMID0_SIZE, 2 * vmid, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_GWS_VMID0, vmid, 0);
- WREG32_SOC15_OFFSET(GC, xcc_id, regGDS_OA_VMID0, vmid, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * vmid, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * vmid, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, vmid, 0);
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, vmid, 0);
}
}
@@ -1057,42 +1057,42 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- WREG32_FIELD15_PREREG(GC, i, GRBM_CNTL, READ_TIMEOUT, 0xff);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), GRBM_CNTL, READ_TIMEOUT, 0xff);
gfx_v9_4_3_setup_rb(adev, i);
}
gfx_v9_4_3_get_cu_info(adev, &adev->gfx.cu_info);
- adev->gfx.config.db_debug2 = RREG32_SOC15(GC, 0, regDB_DEBUG2);
+ adev->gfx.config.db_debug2 = RREG32_SOC15(GC, GET_INST(GC, 0), regDB_DEBUG2);
/* XXX SH_MEM regs */
/* where to put LDS, scratch, GPUVM in FSA64 space */
mutex_lock(&adev->srbm_mutex);
for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
for (j = 0; j < num_xcc; j++) {
- soc15_grbm_select(adev, 0, 0, 0, i, j);
+ soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, j));
/* CP and shaders */
if (i == 0) {
tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
SH_MEM_ALIGNMENT_MODE_UNALIGNED);
tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
!!adev->gmc.noretry);
- WREG32_SOC15_RLC(GC, j, regSH_MEM_CONFIG, tmp);
- WREG32_SOC15_RLC(GC, j, regSH_MEM_BASES, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, j), regSH_MEM_CONFIG, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, j), regSH_MEM_BASES, 0);
} else {
tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
SH_MEM_ALIGNMENT_MODE_UNALIGNED);
tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
!!adev->gmc.noretry);
- WREG32_SOC15_RLC(GC, j, regSH_MEM_CONFIG, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, j), regSH_MEM_CONFIG, tmp);
tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
(adev->gmc.private_aperture_start >> 48));
tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
(adev->gmc.shared_aperture_start >> 48));
- WREG32_SOC15_RLC(GC, j, regSH_MEM_BASES, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, j), regSH_MEM_BASES, tmp);
}
}
}
- soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, 0));
mutex_unlock(&adev->srbm_mutex);
@@ -1105,18 +1105,18 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
static void gfx_v9_4_3_enable_save_restore_machine(struct amdgpu_device *adev,
int xcc_id)
{
- WREG32_FIELD15_PREREG(GC, xcc_id, RLC_SRM_CNTL, SRM_ENABLE, 1);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_SRM_CNTL, SRM_ENABLE, 1);
}
static void gfx_v9_4_3_init_csb(struct amdgpu_device *adev, int xcc_id)
{
adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
/* csib */
- WREG32_RLC(SOC15_REG_OFFSET(GC, xcc_id, regRLC_CSIB_ADDR_HI),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regRLC_CSIB_ADDR_HI),
adev->gfx.rlc.clear_state_gpu_addr >> 32);
- WREG32_RLC(SOC15_REG_OFFSET(GC, xcc_id, regRLC_CSIB_ADDR_LO),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regRLC_CSIB_ADDR_LO),
adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc);
- WREG32_RLC(SOC15_REG_OFFSET(GC, xcc_id, regRLC_CSIB_LENGTH),
+ WREG32_RLC(SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regRLC_CSIB_LENGTH),
adev->gfx.rlc.clear_state_size);
}
@@ -1137,7 +1137,7 @@ static void gfx_v9_4_3_init_pg(struct amdgpu_device *adev, int xcc_id)
AMD_PG_SUPPORT_CP |
AMD_PG_SUPPORT_GDS |
AMD_PG_SUPPORT_RLC_SMU_HS)) {
- WREG32_SOC15(GC, 0, regRLC_JUMP_TABLE_RESTORE,
+ WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_JUMP_TABLE_RESTORE,
adev->gfx.rlc.cp_table_gpu_addr >> 8);
}
}
@@ -1146,9 +1146,9 @@ static void gfx_v9_4_3_disable_gpa_mode(struct amdgpu_device *adev, int xcc_id)
{
uint32_t data;
- data = RREG32_SOC15(GC, xcc_id, regCPC_PSP_DEBUG);
+ data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG);
data |= CPC_PSP_DEBUG__UTCL2IUGPAOVERRIDE_MASK;
- WREG32_SOC15(GC, xcc_id, regCPC_PSP_DEBUG, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG, data);
}
static void gfx_v9_4_3_program_xcc_id(struct amdgpu_device *adev, int xcc_id)
@@ -1160,16 +1160,16 @@ static void gfx_v9_4_3_program_xcc_id(struct amdgpu_device *adev, int xcc_id)
switch (num_xcc) {
/* directly config VIRTUAL_XCC_ID to 0 for 1-XCC */
case 1:
- WREG32_SOC15(GC, xcc_id, regCP_HYP_XCP_CTL, 0x8);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HYP_XCP_CTL, 0x8);
break;
case 2:
tmp = (xcc_id % adev->gfx.num_xcc_per_xcp) << REG_FIELD_SHIFT(CP_HYP_XCP_CTL, VIRTUAL_XCC_ID);
tmp = tmp | (adev->gfx.num_xcc_per_xcp << REG_FIELD_SHIFT(CP_HYP_XCP_CTL, NUM_XCC_IN_XCP));
- WREG32_SOC15(GC, xcc_id, regCP_HYP_XCP_CTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HYP_XCP_CTL, tmp);
tmp = xcc_id << REG_FIELD_SHIFT(CP_PSP_XCP_CTL, PHYSICAL_XCC_ID);
tmp = tmp | (xcc_id << REG_FIELD_SHIFT(CP_PSP_XCP_CTL, XCC_DIE_ID));
- WREG32_SOC15(GC, xcc_id, regCP_PSP_XCP_CTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_PSP_XCP_CTL, tmp);
break;
default:
break;
@@ -1181,7 +1181,7 @@ static bool gfx_v9_4_3_is_rlc_enabled(struct amdgpu_device *adev)
uint32_t rlc_setting;
/* if RLC is not enabled, do nothing */
- rlc_setting = RREG32_SOC15(GC, 0, regRLC_CNTL);
+ rlc_setting = RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CNTL);
if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
return false;
@@ -1195,11 +1195,11 @@ static void gfx_v9_4_3_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
data = RLC_SAFE_MODE__CMD_MASK;
data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
- WREG32_SOC15(GC, xcc_id, regRLC_SAFE_MODE, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
/* wait for RLC_SAFE_MODE */
for (i = 0; i < adev->usec_timeout; i++) {
- if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, regRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
+ if (!REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
break;
udelay(1);
}
@@ -1210,7 +1210,7 @@ static void gfx_v9_4_3_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
uint32_t data;
data = RLC_SAFE_MODE__CMD_MASK;
- WREG32_SOC15(GC, xcc_id, regRLC_SAFE_MODE, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
}
static int gfx_v9_4_3_rlc_init(struct amdgpu_device *adev)
@@ -1247,7 +1247,7 @@ static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev,
for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
gfx_v9_4_3_select_se_sh(adev, i, j, 0xffffffff, xcc_id);
for (k = 0; k < adev->usec_timeout; k++) {
- if (RREG32_SOC15(GC, 0, regRLC_SERDES_CU_MASTER_BUSY) == 0)
+ if (RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SERDES_CU_MASTER_BUSY) == 0)
break;
udelay(1);
}
@@ -1270,7 +1270,7 @@ static void gfx_v9_4_3_wait_for_rlc_serdes(struct amdgpu_device *adev,
RLC_SERDES_NONCU_MASTER_BUSY__TC0_MASTER_BUSY_MASK |
RLC_SERDES_NONCU_MASTER_BUSY__TC1_MASTER_BUSY_MASK;
for (k = 0; k < adev->usec_timeout; k++) {
- if ((RREG32_SOC15(GC, 0, regRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
+ if ((RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
break;
udelay(1);
}
@@ -1283,13 +1283,13 @@ static void gfx_v9_4_3_enable_gui_idle_interrupt(struct amdgpu_device *adev,
/* These interrupts should be enabled to drive DS clock */
- tmp = RREG32_SOC15(GC, xcc_id, regCP_INT_CNTL_RING0);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0);
tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0);
tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0);
tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
- WREG32_SOC15(GC, xcc_id, regCP_INT_CNTL_RING0, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0, tmp);
}
static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
@@ -1298,7 +1298,7 @@ static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- WREG32_FIELD15_PREREG(GC, i, RLC_CNTL, RLC_ENABLE_F32, 0);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), RLC_CNTL, RLC_ENABLE_F32, 0);
gfx_v9_4_3_enable_gui_idle_interrupt(adev, false, i);
gfx_v9_4_3_wait_for_rlc_serdes(adev, i);
}
@@ -1310,9 +1310,9 @@ static void gfx_v9_4_3_rlc_reset(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- WREG32_FIELD15_PREREG(GC, i, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
udelay(50);
- WREG32_FIELD15_PREREG(GC, i, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
udelay(50);
}
}
@@ -1326,7 +1326,7 @@ static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- WREG32_FIELD15_PREREG(GC, i, RLC_CNTL, RLC_ENABLE_F32, 1);
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), RLC_CNTL, RLC_ENABLE_F32, 1);
udelay(50);
/* carrizo do enable cp interrupt after cp inited */
@@ -1337,18 +1337,18 @@ static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
#ifdef AMDGPU_RLC_DEBUG_RETRY
/* RLC_GPM_GENERAL_6 : RLC Ucode version */
- rlc_ucode_ver = RREG32_SOC15(GC, i, regRLC_GPM_GENERAL_6);
+ rlc_ucode_ver = RREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_6);
if (rlc_ucode_ver == 0x108) {
dev_info(adev->dev,
"Using rlc debug ucode. regRLC_GPM_GENERAL_6 ==0x08%x / fw_ver == %i \n",
rlc_ucode_ver, adev->gfx.rlc_fw_version);
/* RLC_GPM_TIMER_INT_3 : Timer interval in RefCLK cycles,
* default is 0x9C4 to create a 100us interval */
- WREG32_SOC15(GC, i, regRLC_GPM_TIMER_INT_3, 0x9C4);
+ WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_TIMER_INT_3, 0x9C4);
/* RLC_GPM_GENERAL_12 : Minimum gap between wptr and rptr
* to disable the page fault retry interrupts, default is
* 0x100 (256) */
- WREG32_SOC15(GC, i, regRLC_GPM_GENERAL_12, 0x100);
+ WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_12, 0x100);
}
#endif
}
@@ -1370,16 +1370,16 @@ static int gfx_v9_4_3_rlc_load_microcode(struct amdgpu_device *adev, int xcc_id)
le32_to_cpu(hdr->header.ucode_array_offset_bytes));
fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
- WREG32_SOC15(GC, xcc_id, regRLC_GPM_UCODE_ADDR,
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR,
RLCG_UCODE_LOADING_START_ADDRESS);
for (i = 0; i < fw_size; i++) {
if (amdgpu_emu_mode == 1 && i % 100 == 0) {
dev_info(adev->dev, "Write RLC ucode data %u DWs\n", i);
msleep(1);
}
- WREG32_SOC15(GC, xcc_id, regRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
}
- WREG32_SOC15(GC, xcc_id, regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
return 0;
}
@@ -1393,7 +1393,7 @@ static int gfx_v9_4_3_rlc_resume(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
/* disable CG */
- WREG32_SOC15(GC, i, regRLC_CGCG_CGLS_CTRL, 0);
+ WREG32_SOC15(GC, GET_INST(GC, i), regRLC_CGCG_CGLS_CTRL, 0);
gfx_v9_4_3_init_pg(adev, i);
@@ -1415,7 +1415,7 @@ static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev,
{
u32 reg, data;
- reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
+ reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL);
if (amdgpu_sriov_is_pp_one_vf(adev))
data = RREG32_NO_KIQ(reg);
else
@@ -1425,9 +1425,9 @@ static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev,
data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
if (amdgpu_sriov_is_pp_one_vf(adev))
- WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
+ WREG32_SOC15_NO_KIQ(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
else
- WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
}
static const struct soc15_reg_rlcg rlcg_access_gc_9_4_3[] = {
@@ -1439,7 +1439,7 @@ static bool gfx_v9_4_3_check_rlcg_range(struct amdgpu_device *adev,
uint32_t offset,
struct soc15_reg_rlcg *entries, int arr_size)
{
- int i;
+ int i, inst;
uint32_t reg;
if (!entries)
@@ -1449,7 +1449,12 @@ static bool gfx_v9_4_3_check_rlcg_range(struct amdgpu_device *adev,
const struct soc15_reg_rlcg *entry;
entry = &entries[i];
- reg = adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg;
+ inst = adev->ip_map.logical_to_dev_inst ?
+ adev->ip_map.logical_to_dev_inst(
+ adev, entry->hwip, entry->instance) :
+ entry->instance;
+ reg = adev->reg_offset[entry->hwip][inst][entry->segment] +
+ entry->reg;
if (offset == reg)
return true;
}
@@ -1468,9 +1473,9 @@ static void gfx_v9_4_3_cp_compute_enable(struct amdgpu_device *adev,
bool enable, int xcc_id)
{
if (enable) {
- WREG32_SOC15_RLC(GC, xcc_id, regCP_MEC_CNTL, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL, 0);
} else {
- WREG32_SOC15_RLC(GC, xcc_id, regCP_MEC_CNTL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL,
(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
adev->gfx.kiq[xcc_id].ring.sched.ready = false;
}
@@ -1501,17 +1506,17 @@ static int gfx_v9_4_3_cp_compute_load_microcode(struct amdgpu_device *adev,
tmp = 0;
tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
- WREG32_SOC15(GC, xcc_id, regCP_CPC_IC_BASE_CNTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_CNTL, tmp);
- WREG32_SOC15(GC, xcc_id, regCP_CPC_IC_BASE_LO,
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_LO,
adev->gfx.mec.mec_fw_gpu_addr & 0xFFFFF000);
- WREG32_SOC15(GC, xcc_id, regCP_CPC_IC_BASE_HI,
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_HI,
upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
mec_ucode_addr_offset =
- SOC15_REG_OFFSET(GC, xcc_id, regCP_MEC_ME1_UCODE_ADDR);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_ADDR);
mec_ucode_data_offset =
- SOC15_REG_OFFSET(GC, xcc_id, regCP_MEC_ME1_UCODE_DATA);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_DATA);
/* MEC1 */
WREG32(mec_ucode_addr_offset, mec_hdr->jt_offset);
@@ -1532,12 +1537,12 @@ static void gfx_v9_4_3_kiq_setting(struct amdgpu_ring *ring, int xcc_id)
struct amdgpu_device *adev = ring->adev;
/* tell RLC which is KIQ queue */
- tmp = RREG32_SOC15(GC, xcc_id, regRLC_CP_SCHEDULERS);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS);
tmp &= 0xffffff00;
tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
- WREG32_SOC15_RLC(GC, xcc_id, regRLC_CP_SCHEDULERS, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS, tmp);
tmp |= 0x80;
- WREG32_SOC15_RLC(GC, xcc_id, regRLC_CP_SCHEDULERS, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS, tmp);
}
static void gfx_v9_4_3_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd)
@@ -1580,14 +1585,14 @@ static int gfx_v9_4_3_mqd_init(struct amdgpu_ring *ring)
mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
- tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_EOP_CONTROL);
tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
(order_base_2(GFX9_MEC_HPD_SIZE / 4) - 1));
mqd->cp_hqd_eop_control = tmp;
/* enable doorbell? */
- tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_PQ_DOORBELL_CONTROL);
if (ring->use_doorbell) {
tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
@@ -1617,7 +1622,7 @@ static int gfx_v9_4_3_mqd_init(struct amdgpu_ring *ring)
mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
/* set MQD vmid to 0 */
- tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_MQD_CONTROL);
tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
mqd->cp_mqd_control = tmp;
@@ -1627,7 +1632,7 @@ static int gfx_v9_4_3_mqd_init(struct amdgpu_ring *ring)
mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
/* set up the HQD, this is similar to CP_RB0_CNTL */
- tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_PQ_CONTROL);
tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
(order_base_2(ring->ring_size / 4) - 1));
tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
@@ -1654,23 +1659,23 @@ static int gfx_v9_4_3_mqd_init(struct amdgpu_ring *ring)
/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
ring->wptr = 0;
- mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR);
+ mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_PQ_RPTR);
/* set the vmid for the queue */
mqd->cp_hqd_vmid = 0;
- tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_PERSISTENT_STATE);
tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53);
mqd->cp_hqd_persistent_state = tmp;
/* set MIN_IB_AVAIL_SIZE */
- tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_IB_CONTROL);
tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
mqd->cp_hqd_ib_control = tmp;
/* set static priority for a queue/ring */
gfx_v9_4_3_mqd_set_priority(ring, mqd);
- mqd->cp_hqd_quantum = RREG32_SOC15(GC, 0, regCP_HQD_QUANTUM);
+ mqd->cp_hqd_quantum = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HQD_QUANTUM);
/* map_queues packet doesn't need activate the queue,
* so only kiq need set this field.
@@ -1690,94 +1695,94 @@ static int gfx_v9_4_3_kiq_init_register(struct amdgpu_ring *ring, int xcc_id)
/* disable wptr polling */
WREG32_FIELD15_PREREG(GC, xcc_id, CP_PQ_WPTR_POLL_CNTL, EN, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_EOP_BASE_ADDR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR,
mqd->cp_hqd_eop_base_addr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_EOP_BASE_ADDR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR_HI,
mqd->cp_hqd_eop_base_addr_hi);
/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_EOP_CONTROL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_CONTROL,
mqd->cp_hqd_eop_control);
/* enable doorbell? */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_DOORBELL_CONTROL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
mqd->cp_hqd_pq_doorbell_control);
/* disable the queue if it's active */
- if (RREG32_SOC15(GC, xcc_id, regCP_HQD_ACTIVE) & 1) {
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_DEQUEUE_REQUEST, 1);
+ if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
for (j = 0; j < adev->usec_timeout; j++) {
- if (!(RREG32_SOC15(GC, xcc_id, regCP_HQD_ACTIVE) & 1))
+ if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
break;
udelay(1);
}
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_DEQUEUE_REQUEST,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
mqd->cp_hqd_dequeue_request);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_RPTR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR,
mqd->cp_hqd_pq_rptr);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_LO,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
mqd->cp_hqd_pq_wptr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
mqd->cp_hqd_pq_wptr_hi);
}
/* set the pointer to the MQD */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_MQD_BASE_ADDR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR,
mqd->cp_mqd_base_addr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_MQD_BASE_ADDR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR_HI,
mqd->cp_mqd_base_addr_hi);
/* set MQD vmid to 0 */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_MQD_CONTROL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_CONTROL,
mqd->cp_mqd_control);
/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_BASE,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE,
mqd->cp_hqd_pq_base_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_BASE_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE_HI,
mqd->cp_hqd_pq_base_hi);
/* set up the HQD, this is similar to CP_RB0_CNTL */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_CONTROL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_CONTROL,
mqd->cp_hqd_pq_control);
/* set the wb address whether it's enabled or not */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_RPTR_REPORT_ADDR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR,
mqd->cp_hqd_pq_rptr_report_addr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
mqd->cp_hqd_pq_rptr_report_addr_hi);
/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_POLL_ADDR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR,
mqd->cp_hqd_pq_wptr_poll_addr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
mqd->cp_hqd_pq_wptr_poll_addr_hi);
/* enable the doorbell if requested */
if (ring->use_doorbell) {
- WREG32_SOC15(GC, xcc_id, regCP_MEC_DOORBELL_RANGE_LOWER,
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEC_DOORBELL_RANGE_LOWER,
(adev->doorbell_index.kiq * 2) << 2);
- WREG32_SOC15(GC, xcc_id, regCP_MEC_DOORBELL_RANGE_UPPER,
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEC_DOORBELL_RANGE_UPPER,
(adev->doorbell_index.userqueue_end * 2) << 2);
}
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_DOORBELL_CONTROL,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
mqd->cp_hqd_pq_doorbell_control);
/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_LO,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
mqd->cp_hqd_pq_wptr_lo);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_HI,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
mqd->cp_hqd_pq_wptr_hi);
/* set the vmid for the queue */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_VMID, mqd->cp_hqd_vmid);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_VMID, mqd->cp_hqd_vmid);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PERSISTENT_STATE,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE,
mqd->cp_hqd_persistent_state);
/* activate the queue */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_ACTIVE,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE,
mqd->cp_hqd_active);
if (ring->use_doorbell)
@@ -1792,12 +1797,12 @@ static int gfx_v9_4_3_kiq_fini_register(struct amdgpu_ring *ring, int xcc_id)
int j;
/* disable the queue if it's active */
- if (RREG32_SOC15(GC, xcc_id, regCP_HQD_ACTIVE) & 1) {
+ if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_DEQUEUE_REQUEST, 1);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
for (j = 0; j < adev->usec_timeout; j++) {
- if (!(RREG32_SOC15(GC, xcc_id, regCP_HQD_ACTIVE) & 1))
+ if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
break;
udelay(1);
}
@@ -1806,21 +1811,21 @@ static int gfx_v9_4_3_kiq_fini_register(struct amdgpu_ring *ring, int xcc_id)
DRM_DEBUG("KIQ dequeue request failed.\n");
/* Manual disable if dequeue request times out */
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_ACTIVE, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE, 0);
}
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_DEQUEUE_REQUEST,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
0);
}
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_IQ_TIMER, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_IB_CONTROL, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PERSISTENT_STATE, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_DOORBELL_CONTROL, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_RPTR, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_HI, 0);
- WREG32_SOC15_RLC(GC, xcc_id, regCP_HQD_PQ_WPTR_LO, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IQ_TIMER, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO, 0);
return 0;
}
@@ -1848,19 +1853,19 @@ static int gfx_v9_4_3_kiq_init_queue(struct amdgpu_ring *ring, int xcc_id)
ring->wptr = 0;
amdgpu_ring_clear_ring(ring);
mutex_lock(&adev->srbm_mutex);
- soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, xcc_id);
+ soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
gfx_v9_4_3_kiq_init_register(ring, xcc_id);
- soc15_grbm_select(adev, 0, 0, 0, 0, xcc_id);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
mutex_unlock(&adev->srbm_mutex);
} else {
memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
mutex_lock(&adev->srbm_mutex);
- soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, xcc_id);
+ soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
gfx_v9_4_3_mqd_init(ring);
gfx_v9_4_3_kiq_init_register(ring, xcc_id);
- soc15_grbm_select(adev, 0, 0, 0, 0, xcc_id);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
mutex_unlock(&adev->srbm_mutex);
if (adev->gfx.kiq[xcc_id].mqd_backup)
@@ -1888,9 +1893,9 @@ static int gfx_v9_4_3_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id)
((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
mutex_lock(&adev->srbm_mutex);
- soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, xcc_id);
+ soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
gfx_v9_4_3_mqd_init(ring);
- soc15_grbm_select(adev, 0, 0, 0, 0, xcc_id);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
mutex_unlock(&adev->srbm_mutex);
if (adev->gfx.mec.mqd_backup[mqd_idx])
@@ -2049,9 +2054,9 @@ static int gfx_v9_4_3_hw_fini(void *handle)
mutex_lock(&adev->srbm_mutex);
soc15_grbm_select(adev, adev->gfx.kiq[i].ring.me,
adev->gfx.kiq[i].ring.pipe,
- adev->gfx.kiq[i].ring.queue, 0, i);
+ adev->gfx.kiq[i].ring.queue, 0, GET_INST(GC, i));
gfx_v9_4_3_kiq_fini_register(&adev->gfx.kiq[i].ring, i);
- soc15_grbm_select(adev, 0, 0, 0, 0, i);
+ soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, i));
mutex_unlock(&adev->srbm_mutex);
}
@@ -2085,7 +2090,7 @@ static bool gfx_v9_4_3_is_idle(void *handle)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- if (REG_GET_FIELD(RREG32_SOC15(GC, i, regGRBM_STATUS),
+ if (REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, i), regGRBM_STATUS),
GRBM_STATUS, GUI_ACTIVE))
return false;
}
@@ -2112,7 +2117,7 @@ static int gfx_v9_4_3_soft_reset(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
/* GRBM_STATUS */
- tmp = RREG32_SOC15(GC, 0, regGRBM_STATUS);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS);
if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
@@ -2131,7 +2136,7 @@ static int gfx_v9_4_3_soft_reset(void *handle)
}
/* GRBM_STATUS2 */
- tmp = RREG32_SOC15(GC, 0, regGRBM_STATUS2);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS2);
if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY))
grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
@@ -2145,17 +2150,17 @@ static int gfx_v9_4_3_soft_reset(void *handle)
gfx_v9_4_3_cp_compute_enable(adev, false, 0);
if (grbm_soft_reset) {
- tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
tmp |= grbm_soft_reset;
dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
- WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, tmp);
- tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET);
+ WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
udelay(50);
tmp &= ~grbm_soft_reset;
- WREG32_SOC15(GC, 0, regGRBM_SOFT_RESET, tmp);
- tmp = RREG32_SOC15(GC, 0, regGRBM_SOFT_RESET);
+ WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
}
/* Wait a little for things to settle down */
@@ -2174,22 +2179,22 @@ static void gfx_v9_4_3_ring_emit_gds_switch(struct amdgpu_ring *ring,
/* GDS Base */
gfx_v9_4_3_write_data_to_reg(ring, 0, false,
- SOC15_REG_OFFSET(GC, 0, regGDS_VMID0_BASE) + 2 * vmid,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_BASE) + 2 * vmid,
gds_base);
/* GDS Size */
gfx_v9_4_3_write_data_to_reg(ring, 0, false,
- SOC15_REG_OFFSET(GC, 0, regGDS_VMID0_SIZE) + 2 * vmid,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_SIZE) + 2 * vmid,
gds_size);
/* GWS */
gfx_v9_4_3_write_data_to_reg(ring, 0, false,
- SOC15_REG_OFFSET(GC, 0, regGDS_GWS_VMID0) + vmid,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_GWS_VMID0) + vmid,
gws_size << GDS_GWS_VMID0__SIZE__SHIFT | gws_base);
/* OA */
gfx_v9_4_3_write_data_to_reg(ring, 0, false,
- SOC15_REG_OFFSET(GC, 0, regGDS_OA_VMID0) + vmid,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_OA_VMID0) + vmid,
(1 << (oa_size + oa_base)) - (1 << oa_base));
}
@@ -2260,7 +2265,7 @@ static void gfx_v9_4_3_update_medium_grain_clock_gating(struct amdgpu_device *ad
/* It is disabled by HW by default */
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
/* 1 - RLC_CGTT_MGCG_OVERRIDE */
- def = data = RREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
@@ -2270,28 +2275,28 @@ static void gfx_v9_4_3_update_medium_grain_clock_gating(struct amdgpu_device *ad
data |= RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK;
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
/* MGLS is a global flag to control all MGLS in GFX */
if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
/* 2 - RLC memory Light sleep */
if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) {
- def = data = RREG32_SOC15(GC, xcc_id, regRLC_MEM_SLP_CNTL);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_MEM_SLP_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
}
/* 3 - CP memory Light sleep */
if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) {
- def = data = RREG32_SOC15(GC, xcc_id, regCP_MEM_SLP_CNTL);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
if (def != data)
- WREG32_SOC15(GC, xcc_id, regCP_MEM_SLP_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
}
}
} else {
/* 1 - MGCG_OVERRIDE */
- def = data = RREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
@@ -2299,20 +2304,20 @@ static void gfx_v9_4_3_update_medium_grain_clock_gating(struct amdgpu_device *ad
RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
/* 2 - disable MGLS in RLC */
- data = RREG32_SOC15(GC, xcc_id, regRLC_MEM_SLP_CNTL);
+ data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) {
data &= ~RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
- WREG32_SOC15(GC, xcc_id, regRLC_MEM_SLP_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
}
/* 3 - disable MGLS in CP */
- data = RREG32_SOC15(GC, xcc_id, regCP_MEM_SLP_CNTL);
+ data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) {
data &= ~CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
- WREG32_SOC15(GC, xcc_id, regCP_MEM_SLP_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
}
}
@@ -2327,7 +2332,7 @@ static void gfx_v9_4_3_update_coarse_grain_clock_gating(struct amdgpu_device *ad
amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
- def = data = RREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
/* unset CGCG override */
data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
@@ -2336,10 +2341,10 @@ static void gfx_v9_4_3_update_coarse_grain_clock_gating(struct amdgpu_device *ad
data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
/* update CGCG and CGLS override bits */
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_CGTT_MGCG_OVERRIDE, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
/* enable cgcg FSM(0x0000363F) */
- def = RREG32_SOC15(GC, xcc_id, regRLC_CGCG_CGLS_CTRL);
+ def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
if (adev->asic_type == CHIP_ARCTURUS)
data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
@@ -2351,21 +2356,21 @@ static void gfx_v9_4_3_update_coarse_grain_clock_gating(struct amdgpu_device *ad
data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_CGCG_CGLS_CTRL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
/* set IDLE_POLL_COUNT(0x00900100) */
- def = RREG32_SOC15(GC, xcc_id, regCP_RB_WPTR_POLL_CNTL);
+ def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL);
data = (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
(0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
if (def != data)
- WREG32_SOC15(GC, xcc_id, regCP_RB_WPTR_POLL_CNTL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL, data);
} else {
- def = data = RREG32_SOC15(GC, xcc_id, regRLC_CGCG_CGLS_CTRL);
+ def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
/* reset CGCG/CGLS bits */
data &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
/* disable cgcg and cgls in FSM */
if (def != data)
- WREG32_SOC15(GC, xcc_id, regRLC_CGCG_CGLS_CTRL, data);
+ WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
}
amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
@@ -2444,12 +2449,12 @@ static void gfx_v9_4_3_get_clockgating_state(void *handle, u64 *flags)
*flags = 0;
/* AMD_CG_SUPPORT_GFX_MGCG */
- data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, regRLC_CGTT_MGCG_OVERRIDE));
+ data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGTT_MGCG_OVERRIDE));
if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
*flags |= AMD_CG_SUPPORT_GFX_MGCG;
/* AMD_CG_SUPPORT_GFX_CGCG */
- data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, regRLC_CGCG_CGLS_CTRL));
+ data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGCG_CGLS_CTRL));
if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
*flags |= AMD_CG_SUPPORT_GFX_CGCG;
@@ -2458,12 +2463,12 @@ static void gfx_v9_4_3_get_clockgating_state(void *handle, u64 *flags)
*flags |= AMD_CG_SUPPORT_GFX_CGLS;
/* AMD_CG_SUPPORT_GFX_RLC_LS */
- data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, regRLC_MEM_SLP_CNTL));
+ data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_MEM_SLP_CNTL));
if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK)
*flags |= AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_MGLS;
/* AMD_CG_SUPPORT_GFX_CP_LS */
- data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, regCP_MEM_SLP_CNTL));
+ data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCP_MEM_SLP_CNTL));
if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
}
@@ -2634,7 +2639,7 @@ static void gfx_v9_4_3_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
WRITE_DATA_DST_SEL(0) | WR_CONFIRM));
- amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, regCPC_INT_STATUS));
+ amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCPC_INT_STATUS));
amdgpu_ring_write(ring, 0);
amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */
}
@@ -2710,16 +2715,16 @@ static void gfx_v9_4_3_set_compute_eop_interrupt_state(struct amdgpu_device *ade
if (me == 1) {
switch (pipe) {
case 0:
- mec_int_cntl_reg = SOC15_REG_OFFSET(GC, xcc_id, regCP_ME1_PIPE0_INT_CNTL);
+ mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL);
break;
case 1:
- mec_int_cntl_reg = SOC15_REG_OFFSET(GC, xcc_id, regCP_ME1_PIPE1_INT_CNTL);
+ mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL);
break;
case 2:
- mec_int_cntl_reg = SOC15_REG_OFFSET(GC, xcc_id, regCP_ME1_PIPE2_INT_CNTL);
+ mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL);
break;
case 3:
- mec_int_cntl_reg = SOC15_REG_OFFSET(GC, xcc_id, regCP_ME1_PIPE3_INT_CNTL);
+ mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL);
break;
default:
DRM_DEBUG("invalid pipe %d\n", pipe);
@@ -2760,7 +2765,7 @@ static int gfx_v9_4_3_set_priv_reg_fault_state(struct amdgpu_device *adev,
case AMDGPU_IRQ_STATE_DISABLE:
case AMDGPU_IRQ_STATE_ENABLE:
for (i = 0; i < num_xcc; i++)
- WREG32_FIELD15_PREREG(GC, i, CP_INT_CNTL_RING0,
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
PRIV_REG_INT_ENABLE,
state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
break;
@@ -2783,7 +2788,7 @@ static int gfx_v9_4_3_set_priv_inst_fault_state(struct amdgpu_device *adev,
case AMDGPU_IRQ_STATE_DISABLE:
case AMDGPU_IRQ_STATE_ENABLE:
for (i = 0; i < num_xcc; i++)
- WREG32_FIELD15_PREREG(GC, i, CP_INT_CNTL_RING0,
+ WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
PRIV_INSTR_INT_ENABLE,
state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
break;
@@ -2943,16 +2948,16 @@ static void gfx_v9_4_3_emit_wave_limit_cs(struct amdgpu_ring *ring,
switch (pipe) {
case 0:
- wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, regSPI_WCL_PIPE_PERCENT_CS0);
+ wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS0);
break;
case 1:
- wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, regSPI_WCL_PIPE_PERCENT_CS1);
+ wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS1);
break;
case 2:
- wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, regSPI_WCL_PIPE_PERCENT_CS2);
+ wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS2);
break;
case 3:
- wcl_cs_reg = SOC15_REG_OFFSET(GC, 0, regSPI_WCL_PIPE_PERCENT_CS3);
+ wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS3);
break;
default:
DRM_DEBUG("invalid pipe %d\n", pipe);
@@ -2974,7 +2979,7 @@ static void gfx_v9_4_3_emit_wave_limit(struct amdgpu_ring *ring, bool enable)
*/
val = enable ? 0x1f : 0x07ffffff;
amdgpu_ring_emit_wreg(ring,
- SOC15_REG_OFFSET(GC, 0, regSPI_WCL_PIPE_PERCENT_GFX),
+ SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_GFX),
val);
/* Restrict waves for normal/low priority compute queues as well
@@ -3161,15 +3166,15 @@ static void gfx_v9_4_3_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
- WREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG, data);
+ WREG32_SOC15(GC, GET_INST(GC, 0), regGC_USER_SHADER_ARRAY_CONFIG, data);
}
static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev)
{
u32 data, mask;
- data = RREG32_SOC15(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG);
- data |= RREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG);
+ data = RREG32_SOC15(GC, GET_INST(GC, 0), regCC_GC_SHADER_ARRAY_CONFIG);
+ data |= RREG32_SOC15(GC, GET_INST(GC, 0), regGC_USER_SHADER_ARRAY_CONFIG);
data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index e35365ab3f1f..c26ac0662c7e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -35,7 +35,7 @@
static u64 gfxhub_v1_2_get_mc_fb_offset(struct amdgpu_device *adev)
{
- return (u64)RREG32_SOC15(GC, 0, regMC_VM_FB_OFFSET) << 24;
+ return (u64)RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_FB_OFFSET) << 24;
}
static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
@@ -48,12 +48,12 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
- WREG32_SOC15_OFFSET(GC, i,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
hub->ctx_addr_distance * vmid,
lower_32_bits(page_table_base));
- WREG32_SOC15_OFFSET(GC, i,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
hub->ctx_addr_distance * vmid,
upper_32_bits(page_table_base));
@@ -79,31 +79,31 @@ static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
if (adev->gmc.pdb0_bo) {
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
(u32)(adev->gmc.fb_start >> 12));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
(u32)(adev->gmc.fb_start >> 44));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
(u32)(adev->gmc.gart_end >> 12));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
(u32)(adev->gmc.gart_end >> 44));
} else {
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
(u32)(adev->gmc.gart_start >> 12));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
(u32)(adev->gmc.gart_start >> 44));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
(u32)(adev->gmc.gart_end >> 12));
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
(u32)(adev->gmc.gart_end >> 44));
}
@@ -119,13 +119,13 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
/* Program the AGP BAR */
- WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BASE, 0);
- WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
- WREG32_SOC15_RLC(GC, i, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BASE, 0);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
/* Program the system aperture low logical page number. */
- WREG32_SOC15_RLC(GC, i, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
if (adev->apu_flags & AMD_APU_IS_RAVEN2)
@@ -136,44 +136,44 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
* aperture high address (add 1) to get rid of the VM
* fault and hardware hang.
*/
- WREG32_SOC15_RLC(GC, i,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i),
regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
max((adev->gmc.fb_end >> 18) + 0x1,
adev->gmc.agp_end >> 18));
else
- WREG32_SOC15_RLC(GC, i,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i),
regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
/* Set default page address. */
value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
- WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
(u32)(value >> 12));
- WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
(u32)(value >> 44));
/* Program "protection fault". */
- WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+ WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
(u32)(adev->dummy_page_addr >> 12));
- WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+ WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
(u32)((u64)adev->dummy_page_addr >> 44));
- tmp = RREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL2);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
- WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
}
/* In the case squeezing vram into GART aperture, we don't use
* FB aperture and AGP aperture. Disable them.
*/
if (adev->gmc.pdb0_bo) {
- WREG32_SOC15(GC, i, regMC_VM_FB_LOCATION_TOP, 0);
- WREG32_SOC15(GC, i, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
- WREG32_SOC15(GC, i, regMC_VM_AGP_TOP, 0);
- WREG32_SOC15(GC, i, regMC_VM_AGP_BOT, 0xFFFFFF);
- WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
- WREG32_SOC15(GC, i, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP, 0);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, 0);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, 0xFFFFFF);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
+ WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
}
}
}
@@ -186,7 +186,7 @@ static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
/* Setup TLB control */
- tmp = RREG32_SOC15(GC, i, regMC_VM_MX_L1_TLB_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL);
tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
ENABLE_L1_TLB, 1);
@@ -200,7 +200,7 @@ static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
MTYPE, MTYPE_UC);/* XXX for emulation. */
tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
- WREG32_SOC15_RLC(GC, i, regMC_VM_MX_L1_TLB_CNTL, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL, tmp);
}
}
@@ -212,7 +212,7 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
/* Setup L2 cache */
- tmp = RREG32_SOC15(GC, i, regVM_L2_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
/* XXX for emulation, Refer to closed source code.*/
@@ -221,12 +221,12 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
- WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL, tmp);
- tmp = RREG32_SOC15(GC, i, regVM_L2_CNTL2);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL2);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
- WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL2, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL2, tmp);
tmp = regVM_L2_CNTL3_DEFAULT;
if (adev->gmc.translate_further) {
@@ -238,7 +238,7 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
}
- WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL3, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL3, tmp);
tmp = regVM_L2_CNTL4_DEFAULT;
if (adev->gmc.xgmi.connected_to_cpu) {
@@ -248,7 +248,7 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
}
- WREG32_SOC15_RLC(GC, i, regVM_L2_CNTL4, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL4, tmp);
}
}
@@ -259,7 +259,7 @@ static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- tmp = RREG32_SOC15(GC, i, regVM_CONTEXT0_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
adev->gmc.vmid0_page_table_depth);
@@ -267,7 +267,7 @@ static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
adev->gmc.vmid0_page_table_block_size);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
- WREG32_SOC15(GC, i, regVM_CONTEXT0_CNTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL, tmp);
}
}
@@ -277,23 +277,23 @@ static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
0XFFFFFFFF);
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
0x0000000F);
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
0);
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
0);
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
- WREG32_SOC15(GC, i,
+ WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
}
}
@@ -316,7 +316,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
for (j = 0; j < num_xcc; j++) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0; i <= 14; i++) {
- tmp = RREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL, i);
+ tmp = RREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL, i);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
num_level);
@@ -348,19 +348,19 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
!adev->gmc.noretry ||
adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3));
- WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT1_CNTL,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL,
i * hub->ctx_distance, tmp);
- WREG32_SOC15_OFFSET(GC, j,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
i * hub->ctx_addr_distance, 0);
- WREG32_SOC15_OFFSET(GC, j,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
i * hub->ctx_addr_distance, 0);
- WREG32_SOC15_OFFSET(GC, j,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
i * hub->ctx_addr_distance,
lower_32_bits(adev->vm_manager.max_pfn - 1));
- WREG32_SOC15_OFFSET(GC, j,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
i * hub->ctx_addr_distance,
upper_32_bits(adev->vm_manager.max_pfn - 1));
@@ -378,9 +378,9 @@ static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0 ; i < 18; ++i) {
- WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
i * hub->eng_addr_distance, 0xffffffff);
- WREG32_SOC15_OFFSET(GC, j, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
i * hub->eng_addr_distance, 0x1f);
}
}
@@ -398,9 +398,9 @@ static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
* VF copy registers so vbios post doesn't program them, for
* SRIOV driver need to program them
*/
- WREG32_SOC15_RLC(GC, i, regMC_VM_FB_LOCATION_BASE,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE,
adev->gmc.vram_start >> 24);
- WREG32_SOC15_RLC(GC, i, regMC_VM_FB_LOCATION_TOP,
+ WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP,
adev->gmc.vram_end >> 24);
}
}
@@ -432,23 +432,23 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
/* Disable all tables */
for (i = 0; i < 16; i++)
- WREG32_SOC15_OFFSET(GC, j, regVM_CONTEXT0_CNTL,
+ WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT0_CNTL,
i * hub->ctx_distance, 0);
/* Setup TLB control */
- tmp = RREG32_SOC15(GC, j, regMC_VM_MX_L1_TLB_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL);
tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
tmp = REG_SET_FIELD(tmp,
MC_VM_MX_L1_TLB_CNTL,
ENABLE_ADVANCED_DRIVER_MODEL,
0);
- WREG32_SOC15_RLC(GC, j, regMC_VM_MX_L1_TLB_CNTL, tmp);
+ WREG32_SOC15_RLC(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL, tmp);
/* Setup L2 cache */
- tmp = RREG32_SOC15(GC, j, regVM_L2_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
- WREG32_SOC15(GC, j, regVM_L2_CNTL, tmp);
- WREG32_SOC15(GC, j, regVM_L2_CNTL3, 0);
+ WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL3, 0);
}
}
@@ -466,7 +466,7 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
num_xcc = NUM_XCC(adev->gfx.xcc_mask);
for (i = 0; i < num_xcc; i++) {
- tmp = RREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL);
+ tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
@@ -497,7 +497,7 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
CRASH_ON_RETRY_FAULT, 1);
}
- WREG32_SOC15(GC, i, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
+ WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL, tmp);
}
}
@@ -511,24 +511,24 @@ static void gfxhub_v1_2_init(struct amdgpu_device *adev)
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
hub->ctx0_ptb_addr_lo32 =
- SOC15_REG_OFFSET(GC, i,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
hub->ctx0_ptb_addr_hi32 =
- SOC15_REG_OFFSET(GC, i,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
hub->vm_inv_eng0_sem =
- SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_SEM);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_SEM);
hub->vm_inv_eng0_req =
- SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_REQ);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_REQ);
hub->vm_inv_eng0_ack =
- SOC15_REG_OFFSET(GC, i, regVM_INVALIDATE_ENG0_ACK);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_ACK);
hub->vm_context0_cntl =
- SOC15_REG_OFFSET(GC, i, regVM_CONTEXT0_CNTL);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
hub->vm_l2_pro_fault_status =
- SOC15_REG_OFFSET(GC, i,
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i),
regVM_L2_PROTECTION_FAULT_STATUS);
hub->vm_l2_pro_fault_cntl =
- SOC15_REG_OFFSET(GC, i, regVM_L2_PROTECTION_FAULT_CNTL);
+ SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
hub->ctx_distance = regVM_CONTEXT1_CNTL -
regVM_CONTEXT0_CNTL;
@@ -551,9 +551,9 @@ static int gfxhub_v1_2_get_xgmi_info(struct amdgpu_device *adev)
u32 max_region;
u64 seg_size;
- xgmi_lfb_cntl = RREG32_SOC15(GC, 0, regMC_VM_XGMI_LFB_CNTL);
+ xgmi_lfb_cntl = RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_CNTL);
seg_size = REG_GET_FIELD(
- RREG32_SOC15(GC, 0, regMC_VM_XGMI_LFB_SIZE),
+ RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_SIZE),
MC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
max_region =
REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
index eb35096756b8..39e4406da4ae 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h
@@ -24,6 +24,9 @@
#ifndef __SOC15_COMMON_H__
#define __SOC15_COMMON_H__
+/* GET_INST returns the physical instance corresponding to a logical instance */
+#define GET_INST(ip, inst) (adev->ip_map.logical_to_dev_inst? adev->ip_map.logical_to_dev_inst(adev, ip##_HWIP, inst): inst)
+
/* Register Access Macros */
#define SOC15_REG_OFFSET(ip, inst, reg) (adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg)
#define SOC15_REG_OFFSET1(ip, inst, reg, offset) \
--
cgit v1.2.3
From dd1a02e280dae1904c8858c8cb572a61a84ba7c0 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Wed, 14 Sep 2022 12:16:48 +0530
Subject: drm/amdgpu: Add xcc specific functions for gfxhub
GFXHUB 1.2 supports multiple XCC instances. Add XCC specific functions
to handle XCC instances separately.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 +
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 202 +++++++++++++++++++------------
2 files changed, 128 insertions(+), 78 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 4c4ce33c8359..c2dd100bbd60 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -1263,6 +1263,10 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter));
+#define for_each_inst(i, inst_mask) \
+ for (i = ffs(inst_mask) - 1; inst_mask; \
+ inst_mask &= ~(1U << i), i = ffs(inst_mask) - 1)
+
#define MIN(X, Y) ((X) < (Y) ? (X) : (Y))
/* Common functions */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index c26ac0662c7e..b9316bfb7263 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -38,15 +38,15 @@ static u64 gfxhub_v1_2_get_mc_fb_offset(struct amdgpu_device *adev)
return (u64)RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_FB_OFFSET) << 24;
}
-static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
- uint32_t vmid,
- uint64_t page_table_base)
+static void gfxhub_v1_2_xcc_setup_vm_pt_regs(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint64_t page_table_base,
+ uint32_t xcc_mask)
{
struct amdgpu_vmhub *hub;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
@@ -57,27 +57,36 @@ static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
hub->ctx_addr_distance * vmid,
upper_32_bits(page_table_base));
-
}
}
-static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
+ uint32_t vmid,
+ uint64_t page_table_base)
+{
+ uint32_t xcc_mask;
+
+ xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+ gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, vmid, page_table_base, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_init_gart_aperture_regs(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
uint64_t pt_base;
- int i, num_xcc;
+ int i;
if (adev->gmc.pdb0_bo)
pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
else
pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
- gfxhub_v1_2_setup_vm_pt_regs(adev, 0, pt_base);
+ gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, 0, pt_base, xcc_mask);
/* If use GART for FB translation, vmid0 page table covers both
* vram and system memory (gart)
*/
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
if (adev->gmc.pdb0_bo) {
WREG32_SOC15(GC, GET_INST(GC, i),
regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
@@ -110,14 +119,15 @@ static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_init_system_aperture_regs(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
uint64_t value;
uint32_t tmp;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
/* Program the AGP BAR */
WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BASE, 0);
WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
@@ -178,13 +188,13 @@ static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_tlb_regs(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
uint32_t tmp;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
/* Setup TLB control */
tmp = RREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL);
@@ -204,13 +214,13 @@ static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_cache_regs(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
uint32_t tmp;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
/* Setup L2 cache */
tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
@@ -252,13 +262,13 @@ static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_enable_system_domain(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
uint32_t tmp;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
@@ -271,12 +281,13 @@ static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_disable_identity_aperture(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
WREG32_SOC15(GC, GET_INST(GC, i),
regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
0XFFFFFFFF);
@@ -298,12 +309,13 @@ static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_setup_vmid_config(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
struct amdgpu_vmhub *hub;
unsigned num_level, block_size;
uint32_t tmp;
- int i, j, num_xcc;
+ int i, j;
num_level = adev->vm_manager.num_level;
block_size = adev->vm_manager.block_size;
@@ -312,8 +324,7 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
else
block_size -= 9;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (j = 0; j < num_xcc; j++) {
+ for_each_inst(j, xcc_mask) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0; i <= 14; i++) {
tmp = RREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL, i);
@@ -368,13 +379,13 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
}
}
-static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_program_invalidation(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
struct amdgpu_vmhub *hub;
- unsigned i, j, num_xcc;
+ unsigned int i, j;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (j = 0; j < num_xcc; j++) {
+ for_each_inst(j, xcc_mask) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
for (i = 0 ; i < 18; ++i) {
@@ -386,18 +397,21 @@ static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
}
}
-static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+static int gfxhub_v1_2_xcc_gart_enable(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
- int i, num_xcc;
-
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
- if (amdgpu_sriov_vf(adev)) {
- /*
- * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
- * VF copy registers so vbios post doesn't program them, for
- * SRIOV driver need to program them
- */
+ uint32_t tmp_mask;
+ int i;
+
+ tmp_mask = xcc_mask;
+ /*
+ * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, because they are
+ * VF copy registers so vbios post doesn't program them, for
+ * SRIOV driver need to program them
+ */
+ if (amdgpu_sriov_vf(adev)) {
+ for_each_inst(i, tmp_mask) {
+ i = ffs(tmp_mask) - 1;
WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE,
adev->gmc.vram_start >> 24);
WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP,
@@ -406,29 +420,37 @@ static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
}
/* GART Enable. */
- gfxhub_v1_2_init_gart_aperture_regs(adev);
- gfxhub_v1_2_init_system_aperture_regs(adev);
- gfxhub_v1_2_init_tlb_regs(adev);
+ gfxhub_v1_2_xcc_init_gart_aperture_regs(adev, xcc_mask);
+ gfxhub_v1_2_xcc_init_system_aperture_regs(adev, xcc_mask);
+ gfxhub_v1_2_xcc_init_tlb_regs(adev, xcc_mask);
if (!amdgpu_sriov_vf(adev))
- gfxhub_v1_2_init_cache_regs(adev);
+ gfxhub_v1_2_xcc_init_cache_regs(adev, xcc_mask);
- gfxhub_v1_2_enable_system_domain(adev);
+ gfxhub_v1_2_xcc_enable_system_domain(adev, xcc_mask);
if (!amdgpu_sriov_vf(adev))
- gfxhub_v1_2_disable_identity_aperture(adev);
- gfxhub_v1_2_setup_vmid_config(adev);
- gfxhub_v1_2_program_invalidation(adev);
+ gfxhub_v1_2_xcc_disable_identity_aperture(adev, xcc_mask);
+ gfxhub_v1_2_xcc_setup_vmid_config(adev, xcc_mask);
+ gfxhub_v1_2_xcc_program_invalidation(adev, xcc_mask);
return 0;
}
-static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
+static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+{
+ uint32_t xcc_mask;
+
+ xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+ return gfxhub_v1_2_xcc_gart_enable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_gart_disable(struct amdgpu_device *adev,
+ uint32_t xcc_mask)
{
struct amdgpu_vmhub *hub;
u32 tmp;
- u32 i, j, num_xcc;
+ u32 i, j;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (j = 0; j < num_xcc; j++) {
+ for_each_inst(j, xcc_mask) {
hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
/* Disable all tables */
for (i = 0; i < 16; i++)
@@ -452,20 +474,22 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
}
}
-/**
- * gfxhub_v1_2_set_fault_enable_default - update GART/VM fault handling
- *
- * @adev: amdgpu_device pointer
- * @value: true redirects VM faults to the default page
- */
-static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
- bool value)
+static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
+{
+ uint32_t xcc_mask;
+
+ xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+ gfxhub_v1_2_xcc_gart_disable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_set_fault_enable_default(struct amdgpu_device *adev,
+ bool value,
+ uint32_t xcc_mask)
{
u32 tmp;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -501,13 +525,27 @@ static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
}
}
-static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+/**
+ * gfxhub_v1_2_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
+ bool value)
+{
+ uint32_t xcc_mask;
+
+ xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+ gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_init(struct amdgpu_device *adev, uint32_t xcc_mask)
{
struct amdgpu_vmhub *hub;
- int i, num_xcc;
+ int i;
- num_xcc = NUM_XCC(adev->gfx.xcc_mask);
- for (i = 0; i < num_xcc; i++) {
+ for_each_inst(i, xcc_mask) {
hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
hub->ctx0_ptb_addr_lo32 =
@@ -543,6 +581,14 @@ static void gfxhub_v1_2_init(struct amdgpu_device *adev)
}
}
+static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+{
+ uint32_t xcc_mask;
+
+ xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+ gfxhub_v1_2_xcc_init(adev, xcc_mask);
+}
+
static int gfxhub_v1_2_get_xgmi_info(struct amdgpu_device *adev)
{
u32 max_num_physical_nodes;
--
cgit v1.2.3
From dfdd6f89c4c7f3315d9cabd9bf80a8174ecb5753 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Fri, 23 Sep 2022 15:20:08 +0530
Subject: drm/amdgpu: Add GFXHUB v1.2 XCP funcs
Add functions required for suspend/resume of GFXHUB instances which are
part of an XCP.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 35 ++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index b9316bfb7263..8ba59ffe0e9f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -21,6 +21,7 @@
*
*/
#include "amdgpu.h"
+#include "amdgpu_xcp.h"
#include "gfxhub_v1_2.h"
#include "gfxhub_v1_1.h"
@@ -638,3 +639,37 @@ const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
.init = gfxhub_v1_2_init,
.get_xgmi_info = gfxhub_v1_2_get_xgmi_info,
};
+
+static int gfxhub_v1_2_xcp_resume(void *handle, uint32_t inst_mask)
+{
+ struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+ bool value;
+ int ret;
+
+ if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
+ value = false;
+ else
+ value = true;
+
+ gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, inst_mask);
+
+ if (!amdgpu_sriov_vf(adev))
+ ret = gfxhub_v1_2_xcc_gart_enable(adev, inst_mask);
+
+ return ret;
+}
+
+static int gfxhub_v1_2_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+ struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+ if (!amdgpu_sriov_vf(adev))
+ gfxhub_v1_2_xcc_gart_disable(adev, inst_mask);
+
+ return 0;
+}
+
+struct amdgpu_xcp_ip_funcs gfxhub_v1_2_xcp_funcs = {
+ .suspend = &gfxhub_v1_2_xcp_suspend,
+ .resume = &gfxhub_v1_2_xcp_resume
+};
--
cgit v1.2.3
From 7a7aaab021a6772b29c81c22db9f4d2d8cd66ecd Mon Sep 17 00:00:00 2001
From: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Date: Tue, 20 Dec 2022 15:37:57 -0500
Subject: drm/amdgpu: Make UTCL2 snoop CPU caches
On AMD APP APUs, to make UTCL2 snoop CPU caches, its not sufficient to
rely on xgmi connected flag so add the logic to use is_app_apu to
program the PDE_REQUEST_PHYSICAL bit correctly for gfxhub and mmhub
both.
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 3 ++-
drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c | 3 ++-
2 files changed, 4 insertions(+), 2 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 8ba59ffe0e9f..8901e73fd700 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -252,7 +252,8 @@ static void gfxhub_v1_2_xcc_init_cache_regs(struct amdgpu_device *adev,
WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL3, tmp);
tmp = regVM_L2_CNTL4_DEFAULT;
- if (adev->gmc.xgmi.connected_to_cpu) {
+ /* For AMD APP APUs setup WC memory */
+ if (adev->gmc.xgmi.connected_to_cpu || adev->gmc.is_app_apu) {
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
} else {
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
index a530e2a3cc28..a8faf66b6878 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_8.c
@@ -257,7 +257,8 @@ static void mmhub_v1_8_init_cache_regs(struct amdgpu_device *adev)
WREG32_SOC15(MMHUB, i, regVM_L2_CNTL3, tmp);
tmp = regVM_L2_CNTL4_DEFAULT;
- if (adev->gmc.xgmi.connected_to_cpu) {
+ /* For AMD APP APUs setup WC memory */
+ if (adev->gmc.xgmi.connected_to_cpu || adev->gmc.is_app_apu) {
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4,
--
cgit v1.2.3
From 1893549af62135c788a66e7ff27c81459c532fb2 Mon Sep 17 00:00:00 2001
From: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Date: Fri, 19 May 2023 10:25:28 +0530
Subject: drm/amdgpu: Fix uninitialized variable in gfxhub_v1_2_xcp_resume
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c:657:6: error: variable 'ret' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
if (!amdgpu_sriov_vf(adev))
^~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c:660:9: note: uninitialized use occurs here
return ret;
^~~
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c:657:2: note: remove the 'if' if its condition is always true
if (!amdgpu_sriov_vf(adev))
^~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c:648:9: note: initialize the variable 'ret' to silence this warning
int ret;
^
= 0
1 error generated.
Cc: Luben Tuikov <luben.tuikov@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Lijo Lazar <lijo.lazar@amd.com>
Cc: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
(limited to 'drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c')
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index 8901e73fd700..4dabf910334b 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -645,7 +645,6 @@ static int gfxhub_v1_2_xcp_resume(void *handle, uint32_t inst_mask)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
bool value;
- int ret;
if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
value = false;
@@ -655,9 +654,9 @@ static int gfxhub_v1_2_xcp_resume(void *handle, uint32_t inst_mask)
gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, inst_mask);
if (!amdgpu_sriov_vf(adev))
- ret = gfxhub_v1_2_xcc_gart_enable(adev, inst_mask);
+ return gfxhub_v1_2_xcc_gart_enable(adev, inst_mask);
- return ret;
+ return 0;
}
static int gfxhub_v1_2_xcp_suspend(void *handle, uint32_t inst_mask)
--
cgit v1.2.3