drm/amd/amdgpu: Port GMC v6 driver to new SI headers (v2)

Port the GMC v6 driver over to the new SI headers.

Tested with a Tahiti SI ASIC.

(v2) Fixed a couple of typos (in commented code) and moved
     defines to si_enums.h

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Tom St Denis 2016-11-08 11:55:42 -05:00 committed by Alex Deucher
parent cf54d6d9f8
commit 7251826971
2 changed files with 184 additions and 147 deletions

View File

@ -1,4 +1,3 @@
/*
* Copyright 2014 Advanced Micro Devices, Inc.
*
@ -26,7 +25,16 @@
#include "amdgpu.h"
#include "gmc_v6_0.h"
#include "amdgpu_ucode.h"
#include "si/sid.h"
#include "bif/bif_3_0_d.h"
#include "bif/bif_3_0_sh_mask.h"
#include "oss/oss_1_0_d.h"
#include "oss/oss_1_0_sh_mask.h"
#include "gmc/gmc_6_0_d.h"
#include "gmc/gmc_6_0_sh_mask.h"
#include "dce/dce_6_0_d.h"
#include "dce/dce_6_0_sh_mask.h"
#include "si_enums.h"
static void gmc_v6_0_set_gart_funcs(struct amdgpu_device *adev);
static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev);
@ -37,6 +45,16 @@ MODULE_FIRMWARE("radeon/pitcairn_mc.bin");
MODULE_FIRMWARE("radeon/verde_mc.bin");
MODULE_FIRMWARE("radeon/oland_mc.bin");
#define MC_SEQ_MISC0__MT__MASK 0xf0000000
#define MC_SEQ_MISC0__MT__GDDR1 0x10000000
#define MC_SEQ_MISC0__MT__DDR2 0x20000000
#define MC_SEQ_MISC0__MT__GDDR3 0x30000000
#define MC_SEQ_MISC0__MT__GDDR4 0x40000000
#define MC_SEQ_MISC0__MT__GDDR5 0x50000000
#define MC_SEQ_MISC0__MT__HBM 0x60000000
#define MC_SEQ_MISC0__MT__DDR3 0xB0000000
static const u32 crtc_offsets[6] =
{
SI_CRTC0_REGISTER_OFFSET,
@ -57,14 +75,14 @@ static void gmc_v6_0_mc_stop(struct amdgpu_device *adev,
gmc_v6_0_wait_for_idle((void *)adev);
blackout = RREG32(MC_SHARED_BLACKOUT_CNTL);
if (REG_GET_FIELD(blackout, mmMC_SHARED_BLACKOUT_CNTL, xxBLACKOUT_MODE) != 1) {
blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
/* Block CPU access */
WREG32(BIF_FB_EN, 0);
WREG32(mmBIF_FB_EN, 0);
/* blackout the MC */
blackout = REG_SET_FIELD(blackout,
mmMC_SHARED_BLACKOUT_CNTL, xxBLACKOUT_MODE, 0);
WREG32(MC_SHARED_BLACKOUT_CNTL, blackout | 1);
MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE, 0);
WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1);
}
/* wait for the MC to settle */
udelay(100);
@ -77,13 +95,13 @@ static void gmc_v6_0_mc_resume(struct amdgpu_device *adev,
u32 tmp;
/* unblackout the MC */
tmp = RREG32(MC_SHARED_BLACKOUT_CNTL);
tmp = REG_SET_FIELD(tmp, mmMC_SHARED_BLACKOUT_CNTL, xxBLACKOUT_MODE, 0);
WREG32(MC_SHARED_BLACKOUT_CNTL, tmp);
tmp = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
tmp = REG_SET_FIELD(tmp, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE, 0);
WREG32(mmMC_SHARED_BLACKOUT_CNTL, tmp);
/* allow CPU access */
tmp = REG_SET_FIELD(0, mmBIF_FB_EN, xxFB_READ_EN, 1);
tmp = REG_SET_FIELD(tmp, mmBIF_FB_EN, xxFB_WRITE_EN, 1);
WREG32(BIF_FB_EN, tmp);
tmp = REG_SET_FIELD(0, BIF_FB_EN, FB_READ_EN, 1);
tmp = REG_SET_FIELD(tmp, BIF_FB_EN, FB_WRITE_EN, 1);
WREG32(mmBIF_FB_EN, tmp);
if (adev->mode_info.num_crtc)
amdgpu_display_resume_mc_access(adev, save);
@ -158,37 +176,37 @@ static int gmc_v6_0_mc_load_microcode(struct amdgpu_device *adev)
new_fw_data = (const __le32 *)
(adev->mc.fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK;
running = RREG32(mmMC_SEQ_SUP_CNTL) & MC_SEQ_SUP_CNTL__RUN_MASK;
if (running == 0) {
/* reset the engine and set to writable */
WREG32(MC_SEQ_SUP_CNTL, 0x00000008);
WREG32(MC_SEQ_SUP_CNTL, 0x00000010);
WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010);
/* load mc io regs */
for (i = 0; i < regs_size; i++) {
WREG32(MC_SEQ_IO_DEBUG_INDEX, le32_to_cpup(new_io_mc_regs++));
WREG32(MC_SEQ_IO_DEBUG_DATA, le32_to_cpup(new_io_mc_regs++));
WREG32(mmMC_SEQ_IO_DEBUG_INDEX, le32_to_cpup(new_io_mc_regs++));
WREG32(mmMC_SEQ_IO_DEBUG_DATA, le32_to_cpup(new_io_mc_regs++));
}
/* load the MC ucode */
for (i = 0; i < ucode_size; i++) {
WREG32(MC_SEQ_SUP_PGM, le32_to_cpup(new_fw_data++));
WREG32(mmMC_SEQ_SUP_PGM, le32_to_cpup(new_fw_data++));
}
/* put the engine back into the active state */
WREG32(MC_SEQ_SUP_CNTL, 0x00000008);
WREG32(MC_SEQ_SUP_CNTL, 0x00000004);
WREG32(MC_SEQ_SUP_CNTL, 0x00000001);
WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
WREG32(mmMC_SEQ_SUP_CNTL, 0x00000004);
WREG32(mmMC_SEQ_SUP_CNTL, 0x00000001);
/* wait for training to complete */
for (i = 0; i < adev->usec_timeout; i++) {
if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D0)
if (RREG32(mmMC_SEQ_TRAIN_WAKEUP_CNTL) & MC_SEQ_TRAIN_WAKEUP_CNTL__TRAIN_DONE_D0_MASK)
break;
udelay(1);
}
for (i = 0; i < adev->usec_timeout; i++) {
if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D1)
if (RREG32(mmMC_SEQ_TRAIN_WAKEUP_CNTL) & MC_SEQ_TRAIN_WAKEUP_CNTL__TRAIN_DONE_D1_MASK)
break;
udelay(1);
}
@ -225,7 +243,7 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
WREG32((0xb08 + j), 0x00000000);
WREG32((0xb09 + j), 0x00000000);
}
WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);
WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
gmc_v6_0_mc_stop(adev, &save);
@ -233,24 +251,24 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
dev_warn(adev->dev, "Wait for MC idle timedout !\n");
}
WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
WREG32(mmVGA_HDP_CONTROL, VGA_HDP_CONTROL__VGA_MEMORY_DISABLE_MASK);
/* Update configuration */
WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
adev->mc.vram_start >> 12);
WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
WREG32(mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
adev->mc.vram_end >> 12);
WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
adev->vram_scratch.gpu_addr >> 12);
tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16;
tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF);
WREG32(MC_VM_FB_LOCATION, tmp);
WREG32(mmMC_VM_FB_LOCATION, tmp);
/* XXX double check these! */
WREG32(HDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8));
WREG32(HDP_NONSURFACE_INFO, (2 << 7) | (1 << 30));
WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF);
WREG32(MC_VM_AGP_BASE, 0);
WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8));
WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30));
WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF);
WREG32(mmMC_VM_AGP_BASE, 0);
WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
if (gmc_v6_0_wait_for_idle((void *)adev)) {
dev_warn(adev->dev, "Wait for MC idle timedout !\n");
@ -265,16 +283,16 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
u32 tmp;
int chansize, numchan;
tmp = RREG32(MC_ARB_RAMCFG);
if (tmp & CHANSIZE_OVERRIDE) {
tmp = RREG32(mmMC_ARB_RAMCFG);
if (tmp & (1 << 11)) {
chansize = 16;
} else if (tmp & CHANSIZE_MASK) {
} else if (tmp & MC_ARB_RAMCFG__CHANSIZE_MASK) {
chansize = 64;
} else {
chansize = 32;
}
tmp = RREG32(MC_SHARED_CHMAP);
switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
tmp = RREG32(mmMC_SHARED_CHMAP);
switch ((tmp & MC_SHARED_CHMAP__NOOFCHAN_MASK) >> MC_SHARED_CHMAP__NOOFCHAN__SHIFT) {
case 0:
default:
numchan = 1;
@ -309,8 +327,8 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
adev->mc.aper_base = pci_resource_start(adev->pdev, 0);
adev->mc.aper_size = pci_resource_len(adev->pdev, 0);
/* size in MB on si */
adev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024ULL * 1024ULL;
adev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024ULL * 1024ULL;
adev->mc.mc_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL;
adev->mc.real_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL;
adev->mc.visible_vram_size = adev->mc.aper_size;
/* unless the user had overridden it, set the gart
@ -329,9 +347,9 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
static void gmc_v6_0_gart_flush_gpu_tlb(struct amdgpu_device *adev,
uint32_t vmid)
{
WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0);
WREG32(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0);
WREG32(VM_INVALIDATE_REQUEST, 1 << vmid);
WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
}
static int gmc_v6_0_gart_set_pte_pde(struct amdgpu_device *adev,
@ -355,20 +373,20 @@ static void gmc_v6_0_set_fault_enable_default(struct amdgpu_device *adev,
{
u32 tmp;
tmp = RREG32(VM_CONTEXT1_CNTL);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxRANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxDUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxPDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxVALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxREAD_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, mmVM_CONTEXT1_CNTL,
xxWRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
WREG32(VM_CONTEXT1_CNTL, tmp);
tmp = RREG32(mmVM_CONTEXT1_CNTL);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
WREG32(mmVM_CONTEXT1_CNTL, tmp);
}
static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
@ -383,33 +401,39 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
if (r)
return r;
/* Setup TLB control */
WREG32(MC_VM_MX_L1_TLB_CNTL,
WREG32(mmMC_VM_MX_L1_TLB_CNTL,
(0xA << 7) |
ENABLE_L1_TLB |
ENABLE_L1_FRAGMENT_PROCESSING |
SYSTEM_ACCESS_MODE_NOT_IN_SYS |
ENABLE_ADVANCED_DRIVER_MODEL |
SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU);
MC_VM_MX_L1_TLB_CNTL__ENABLE_L1_TLB_MASK |
MC_VM_MX_L1_TLB_CNTL__ENABLE_L1_FRAGMENT_PROCESSING_MASK |
MC_VM_MX_L1_TLB_CNTL__SYSTEM_ACCESS_MODE_MASK |
MC_VM_MX_L1_TLB_CNTL__ENABLE_ADVANCED_DRIVER_MODEL_MASK |
(0UL << MC_VM_MX_L1_TLB_CNTL__SYSTEM_APERTURE_UNMAPPED_ACCESS__SHIFT));
/* Setup L2 cache */
WREG32(VM_L2_CNTL, ENABLE_L2_CACHE |
ENABLE_L2_FRAGMENT_PROCESSING |
ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE |
EFFECTIVE_L2_QUEUE_SIZE(7) |
CONTEXT1_IDENTITY_ACCESS_MODE(1));
WREG32(VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS | INVALIDATE_L2_CACHE);
WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY |
BANK_SELECT(4) |
L2_CACHE_BIGK_FRAGMENT_SIZE(4));
WREG32(mmVM_L2_CNTL,
VM_L2_CNTL__ENABLE_L2_CACHE_MASK |
VM_L2_CNTL__ENABLE_L2_FRAGMENT_PROCESSING_MASK |
VM_L2_CNTL__ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE_MASK |
VM_L2_CNTL__ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE_MASK |
(7UL << VM_L2_CNTL__EFFECTIVE_L2_QUEUE_SIZE__SHIFT) |
(1UL << VM_L2_CNTL__CONTEXT1_IDENTITY_ACCESS_MODE__SHIFT));
WREG32(mmVM_L2_CNTL2,
VM_L2_CNTL2__INVALIDATE_ALL_L1_TLBS_MASK |
VM_L2_CNTL2__INVALIDATE_L2_CACHE_MASK);
WREG32(mmVM_L2_CNTL3,
VM_L2_CNTL3__L2_CACHE_BIGK_ASSOCIATIVITY_MASK |
(4UL << VM_L2_CNTL3__BANK_SELECT__SHIFT) |
(4UL << VM_L2_CNTL3__L2_CACHE_BIGK_FRAGMENT_SIZE__SHIFT));
/* setup context0 */
WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12);
WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12);
WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12);
WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12);
WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12);
WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12);
WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
(u32)(adev->dummy_page.addr >> 12));
WREG32(VM_CONTEXT0_CNTL2, 0);
WREG32(VM_CONTEXT0_CNTL, (ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT));
WREG32(mmVM_CONTEXT0_CNTL2, 0);
WREG32(mmVM_CONTEXT0_CNTL,
VM_CONTEXT0_CNTL__ENABLE_CONTEXT_MASK |
(0UL << VM_CONTEXT0_CNTL__PAGE_TABLE_DEPTH__SHIFT) |
VM_CONTEXT0_CNTL__RANGE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK);
WREG32(0x575, 0);
WREG32(0x576, 0);
@ -417,39 +441,41 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
/* empty context1-15 */
/* set vm size, must be a multiple of 4 */
WREG32(VM_CONTEXT1_PAGE_TABLE_START_ADDR, 0);
WREG32(VM_CONTEXT1_PAGE_TABLE_END_ADDR, adev->vm_manager.max_pfn - 1);
WREG32(mmVM_CONTEXT1_PAGE_TABLE_START_ADDR, 0);
WREG32(mmVM_CONTEXT1_PAGE_TABLE_END_ADDR, adev->vm_manager.max_pfn - 1);
/* Assign the pt base to something valid for now; the pts used for
* the VMs are determined by the application and setup and assigned
* on the fly in the vm part of radeon_gart.c
*/
for (i = 1; i < 16; i++) {
if (i < 8)
WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i,
WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i,
adev->gart.table_addr >> 12);
else
WREG32(VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8,
WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8,
adev->gart.table_addr >> 12);
}
/* enable context1-15 */
WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
WREG32(mmVM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
(u32)(adev->dummy_page.addr >> 12));
WREG32(VM_CONTEXT1_CNTL2, 4);
WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) |
PAGE_TABLE_BLOCK_SIZE(amdgpu_vm_block_size - 9) |
RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT |
DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT |
PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT |
PDE0_PROTECTION_FAULT_ENABLE_DEFAULT |
VALID_PROTECTION_FAULT_ENABLE_INTERRUPT |
VALID_PROTECTION_FAULT_ENABLE_DEFAULT |
READ_PROTECTION_FAULT_ENABLE_INTERRUPT |
READ_PROTECTION_FAULT_ENABLE_DEFAULT |
WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT |
WRITE_PROTECTION_FAULT_ENABLE_DEFAULT);
WREG32(mmVM_CONTEXT1_CNTL2, 4);
WREG32(mmVM_CONTEXT1_CNTL,
VM_CONTEXT1_CNTL__ENABLE_CONTEXT_MASK |
(1UL << VM_CONTEXT1_CNTL__PAGE_TABLE_DEPTH__SHIFT) |
((amdgpu_vm_block_size - 9) << VM_CONTEXT1_CNTL__PAGE_TABLE_BLOCK_SIZE__SHIFT) |
VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK |
VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK |
VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_DEFAULT_MASK |
VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_DEFAULT_MASK |
VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_DEFAULT_MASK |
VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_DEFAULT_MASK);
gmc_v6_0_gart_flush_gpu_tlb(adev, 0);
dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
@ -488,19 +514,22 @@ static void gmc_v6_0_gart_disable(struct amdgpu_device *adev)
}*/
/* Disable all tables */
WREG32(VM_CONTEXT0_CNTL, 0);
WREG32(VM_CONTEXT1_CNTL, 0);
WREG32(mmVM_CONTEXT0_CNTL, 0);
WREG32(mmVM_CONTEXT1_CNTL, 0);
/* Setup TLB control */
WREG32(MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE_NOT_IN_SYS |
SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU);
WREG32(mmMC_VM_MX_L1_TLB_CNTL,
MC_VM_MX_L1_TLB_CNTL__SYSTEM_ACCESS_MODE_MASK |
(0UL << MC_VM_MX_L1_TLB_CNTL__SYSTEM_APERTURE_UNMAPPED_ACCESS__SHIFT));
/* Setup L2 cache */
WREG32(VM_L2_CNTL, ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE |
EFFECTIVE_L2_QUEUE_SIZE(7) |
CONTEXT1_IDENTITY_ACCESS_MODE(1));
WREG32(VM_L2_CNTL2, 0);
WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY |
L2_CACHE_BIGK_FRAGMENT_SIZE(0));
WREG32(mmVM_L2_CNTL,
VM_L2_CNTL__ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE_MASK |
VM_L2_CNTL__ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE_MASK |
(7UL << VM_L2_CNTL__EFFECTIVE_L2_QUEUE_SIZE__SHIFT) |
(1UL << VM_L2_CNTL__CONTEXT1_IDENTITY_ACCESS_MODE__SHIFT));
WREG32(mmVM_L2_CNTL2, 0);
WREG32(mmVM_L2_CNTL3,
VM_L2_CNTL3__L2_CACHE_BIGK_ASSOCIATIVITY_MASK |
(0UL << VM_L2_CNTL3__L2_CACHE_BIGK_FRAGMENT_SIZE__SHIFT));
amdgpu_gart_table_vram_unpin(adev);
}
@ -523,7 +552,7 @@ static int gmc_v6_0_vm_init(struct amdgpu_device *adev)
/* base offset of vram pages */
if (adev->flags & AMD_IS_APU) {
u64 tmp = RREG32(MC_VM_FB_OFFSET);
u64 tmp = RREG32(mmMC_VM_FB_OFFSET);
tmp <<= 22;
adev->vm_manager.vram_base_offset = tmp;
} else
@ -540,19 +569,19 @@ static void gmc_v6_0_vm_decode_fault(struct amdgpu_device *adev,
u32 status, u32 addr, u32 mc_client)
{
u32 mc_id;
u32 vmid = REG_GET_FIELD(status, mmVM_CONTEXT1_PROTECTION_FAULT_STATUS, xxVMID);
u32 protections = REG_GET_FIELD(status, mmVM_CONTEXT1_PROTECTION_FAULT_STATUS,
xxPROTECTIONS);
u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
PROTECTIONS);
char block[5] = { mc_client >> 24, (mc_client >> 16) & 0xff,
(mc_client >> 8) & 0xff, mc_client & 0xff, 0 };
mc_id = REG_GET_FIELD(status, mmVM_CONTEXT1_PROTECTION_FAULT_STATUS,
xxMEMORY_CLIENT_ID);
mc_id = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
MEMORY_CLIENT_ID);
dev_err(adev->dev, "VM fault (0x%02x, vmid %d) at page %u, %s from '%s' (0x%08x) (%d)\n",
protections, vmid, addr,
REG_GET_FIELD(status, mmVM_CONTEXT1_PROTECTION_FAULT_STATUS,
xxMEMORY_CLIENT_RW) ?
REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
MEMORY_CLIENT_RW) ?
"write" : "read", block, mc_client, mc_id);
}
@ -655,7 +684,7 @@ static void gmc_v6_0_enable_hdp_mgcg(struct amdgpu_device *adev,
{
u32 orig, data;
orig = data = RREG32(HDP_HOST_PATH_CNTL);
orig = data = RREG32(mmHDP_HOST_PATH_CNTL);
if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_MGCG))
data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 0);
@ -663,7 +692,7 @@ static void gmc_v6_0_enable_hdp_mgcg(struct amdgpu_device *adev,
data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 1);
if (orig != data)
WREG32(HDP_HOST_PATH_CNTL, data);
WREG32(mmHDP_HOST_PATH_CNTL, data);
}
static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
@ -671,7 +700,7 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
{
u32 orig, data;
orig = data = RREG32(HDP_MEM_POWER_LS);
orig = data = RREG32(mmHDP_MEM_POWER_LS);
if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_LS))
data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 1);
@ -679,7 +708,7 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 0);
if (orig != data)
WREG32(HDP_MEM_POWER_LS, data);
WREG32(mmHDP_MEM_POWER_LS, data);
}
*/
@ -713,7 +742,7 @@ static int gmc_v6_0_early_init(void *handle)
if (adev->flags & AMD_IS_APU) {
adev->mc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
} else {
u32 tmp = RREG32(MC_SEQ_MISC0);
u32 tmp = RREG32(mmMC_SEQ_MISC0);
tmp &= MC_SEQ_MISC0__MT__MASK;
adev->mc.vram_type = gmc_v6_0_convert_vram_type(tmp);
}
@ -879,7 +908,7 @@ static int gmc_v6_0_resume(void *handle)
static bool gmc_v6_0_is_idle(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
u32 tmp = RREG32(SRBM_STATUS);
u32 tmp = RREG32(mmSRBM_STATUS);
if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
SRBM_STATUS__MCC_BUSY_MASK | SRBM_STATUS__MCD_BUSY_MASK | SRBM_STATUS__VMC_BUSY_MASK))
@ -895,7 +924,7 @@ static int gmc_v6_0_wait_for_idle(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
for (i = 0; i < adev->usec_timeout; i++) {
tmp = RREG32(SRBM_STATUS) & (SRBM_STATUS__MCB_BUSY_MASK |
tmp = RREG32(mmSRBM_STATUS) & (SRBM_STATUS__MCB_BUSY_MASK |
SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
SRBM_STATUS__MCC_BUSY_MASK |
SRBM_STATUS__MCD_BUSY_MASK |
@ -913,17 +942,17 @@ static int gmc_v6_0_soft_reset(void *handle)
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
struct amdgpu_mode_mc_save save;
u32 srbm_soft_reset = 0;
u32 tmp = RREG32(SRBM_STATUS);
u32 tmp = RREG32(mmSRBM_STATUS);
if (tmp & SRBM_STATUS__VMC_BUSY_MASK)
srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
mmSRBM_SOFT_RESET, xxSOFT_RESET_VMC, 1);
SRBM_SOFT_RESET, SOFT_RESET_VMC, 1);
if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
SRBM_STATUS__MCC_BUSY_MASK | SRBM_STATUS__MCD_BUSY_MASK)) {
if (!(adev->flags & AMD_IS_APU))
srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
mmSRBM_SOFT_RESET, xxSOFT_RESET_MC, 1);
SRBM_SOFT_RESET, SOFT_RESET_MC, 1);
}
if (srbm_soft_reset) {
@ -933,17 +962,17 @@ static int gmc_v6_0_soft_reset(void *handle)
}
tmp = RREG32(SRBM_SOFT_RESET);
tmp = RREG32(mmSRBM_SOFT_RESET);
tmp |= srbm_soft_reset;
dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
WREG32(SRBM_SOFT_RESET, tmp);
tmp = RREG32(SRBM_SOFT_RESET);
WREG32(mmSRBM_SOFT_RESET, tmp);
tmp = RREG32(mmSRBM_SOFT_RESET);
udelay(50);
tmp &= ~srbm_soft_reset;
WREG32(SRBM_SOFT_RESET, tmp);
tmp = RREG32(SRBM_SOFT_RESET);
WREG32(mmSRBM_SOFT_RESET, tmp);
tmp = RREG32(mmSRBM_SOFT_RESET);
udelay(50);
@ -969,20 +998,20 @@ static int gmc_v6_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
switch (state) {
case AMDGPU_IRQ_STATE_DISABLE:
tmp = RREG32(VM_CONTEXT0_CNTL);
tmp = RREG32(mmVM_CONTEXT0_CNTL);
tmp &= ~bits;
WREG32(VM_CONTEXT0_CNTL, tmp);
tmp = RREG32(VM_CONTEXT1_CNTL);
WREG32(mmVM_CONTEXT0_CNTL, tmp);
tmp = RREG32(mmVM_CONTEXT1_CNTL);
tmp &= ~bits;
WREG32(VM_CONTEXT1_CNTL, tmp);
WREG32(mmVM_CONTEXT1_CNTL, tmp);
break;
case AMDGPU_IRQ_STATE_ENABLE:
tmp = RREG32(VM_CONTEXT0_CNTL);
tmp = RREG32(mmVM_CONTEXT0_CNTL);
tmp |= bits;
WREG32(VM_CONTEXT0_CNTL, tmp);
tmp = RREG32(VM_CONTEXT1_CNTL);
WREG32(mmVM_CONTEXT0_CNTL, tmp);
tmp = RREG32(mmVM_CONTEXT1_CNTL);
tmp |= bits;
WREG32(VM_CONTEXT1_CNTL, tmp);
WREG32(mmVM_CONTEXT1_CNTL, tmp);
break;
default:
break;
@ -997,9 +1026,9 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
{
u32 addr, status;
addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR);
status = RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS);
WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1);
addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
WREG32_P(mmVM_CONTEXT1_CNTL2, 1, ~1);
if (!addr && !status)
return 0;

View File

@ -23,6 +23,14 @@
#ifndef SI_ENUMS_H
#define SI_ENUMS_H
#define AMDGPU_NUM_OF_VMIDS 8
#define SI_CRTC0_REGISTER_OFFSET 0
#define SI_CRTC1_REGISTER_OFFSET 0x300
#define SI_CRTC2_REGISTER_OFFSET 0x2600
#define SI_CRTC3_REGISTER_OFFSET 0x2900
#define SI_CRTC4_REGISTER_OFFSET 0x2c00
#define SI_CRTC5_REGISTER_OFFSET 0x2f00
#define DMA0_REGISTER_OFFSET 0x000
#define DMA1_REGISTER_OFFSET 0x200
#define ES_AND_GS_AUTO 3