static void uvd_v6_0_set_uvd_dynamic_clock_mode(struct amdgpu_device *adev,
bool swmode)
{
u32 data, data1 = 0, data2;
/* Always un-gate UVD REGS bit */
data = RREG32(mmUVD_CGC_GATE);
data &= ~(UVD_CGC_GATE__REGS_MASK);
WREG32(mmUVD_CGC_GATE, data);
data = RREG32(mmUVD_CGC_CTRL);
data &= ~(UVD_CGC_CTRL__CLK_OFF_DELAY_MASK |
UVD_CGC_CTRL__CLK_GATE_DLY_TIMER_MASK);
data |= UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK |
1 << REG_FIELD_SHIFT(UVD_CGC_CTRL, CLK_GATE_DLY_TIMER) |
4 << REG_FIELD_SHIFT(UVD_CGC_CTRL, CLK_OFF_DELAY);
data2 = RREG32(mmUVD_SUVD_CGC_CTRL);
if (swmode) {
data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
UVD_CGC_CTRL__SYS_MODE_MASK |
UVD_CGC_CTRL__UDEC_MODE_MASK |
UVD_CGC_CTRL__MPEG2_MODE_MASK |
UVD_CGC_CTRL__REGS_MODE_MASK |
UVD_CGC_CTRL__RBC_MODE_MASK |
UVD_CGC_CTRL__LMI_MC_MODE_MASK |
UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
UVD_CGC_CTRL__IDCT_MODE_MASK |
UVD_CGC_CTRL__MPRD_MODE_MASK |
UVD_CGC_CTRL__MPC_MODE_MASK |
UVD_CGC_CTRL__LBSI_MODE_MASK |
UVD_CGC_CTRL__LRBBM_MODE_MASK |
UVD_CGC_CTRL__WCB_MODE_MASK |
UVD_CGC_CTRL__VCPU_MODE_MASK |
UVD_CGC_CTRL__JPEG_MODE_MASK |
UVD_CGC_CTRL__SCPU_MODE_MASK);
data1 |= UVD_CGC_CTRL2__DYN_OCLK_RAMP_EN_MASK |
UVD_CGC_CTRL2__DYN_RCLK_RAMP_EN_MASK;
data1 &= ~UVD_CGC_CTRL2__GATER_DIV_ID_MASK;
data1 |= 7 << REG_FIELD_SHIFT(UVD_CGC_CTRL2, GATER_DIV_ID);
data2 &= ~(UVD_SUVD_CGC_CTRL__SRE_MODE_MASK |
UVD_SUVD_CGC_CTRL__SIT_MODE_MASK |
UVD_SUVD_CGC_CTRL__SMP_MODE_MASK |
UVD_SUVD_CGC_CTRL__SCM_MODE_MASK |
UVD_SUVD_CGC_CTRL__SDB_MODE_MASK);
} else {
data |= UVD_CGC_CTRL__UDEC_RE_MODE_MASK |
UVD_CGC_CTRL__UDEC_CM_MODE_MASK |
UVD_CGC_CTRL__UDEC_IT_MODE_MASK |
UVD_CGC_CTRL__UDEC_DB_MODE_MASK |
UVD_CGC_CTRL__UDEC_MP_MODE_MASK |
UVD_CGC_CTRL__SYS_MODE_MASK |
UVD_CGC_CTRL__UDEC_MODE_MASK |
UVD_CGC_CTRL__MPEG2_MODE_MASK |
UVD_CGC_CTRL__REGS_MODE_MASK |
UVD_CGC_CTRL__RBC_MODE_MASK |
UVD_CGC_CTRL__LMI_MC_MODE_MASK |
UVD_CGC_CTRL__LMI_UMC_MODE_MASK |
UVD_CGC_CTRL__IDCT_MODE_MASK |
UVD_CGC_CTRL__MPRD_MODE_MASK |
UVD_CGC_CTRL__MPC_MODE_MASK |
UVD_CGC_CTRL__LBSI_MODE_MASK |
UVD_CGC_CTRL__LRBBM_MODE_MASK |
UVD_CGC_CTRL__WCB_MODE_MASK |
UVD_CGC_CTRL__VCPU_MODE_MASK |
UVD_CGC_CTRL__SCPU_MODE_MASK;
data2 |= UVD_SUVD_CGC_CTRL__SRE_MODE_MASK |
UVD_SUVD_CGC_CTRL__SIT_MODE_MASK |
UVD_SUVD_CGC_CTRL__SMP_MODE_MASK |
UVD_SUVD_CGC_CTRL__SCM_MODE_MASK |
UVD_SUVD_CGC_CTRL__SDB_MODE_MASK;
}
WREG32(mmUVD_CGC_CTRL, data);
WREG32(mmUVD_SUVD_CGC_CTRL, data2);
data = RREG32_UVD_CTX(ixUVD_CGC_CTRL2);
data &= ~(REG_FIELD_MASK(UVD_CGC_CTRL2, DYN_OCLK_RAMP_EN) |
REG_FIELD_MASK(UVD_CGC_CTRL2, DYN_RCLK_RAMP_EN) |
REG_FIELD_MASK(UVD_CGC_CTRL2, GATER_DIV_ID));
data1 &= (REG_FIELD_MASK(UVD_CGC_CTRL2, DYN_OCLK_RAMP_EN) |
REG_FIELD_MASK(UVD_CGC_CTRL2, DYN_RCLK_RAMP_EN) |
REG_FIELD_MASK(UVD_CGC_CTRL2, GATER_DIV_ID));
data |= data1;
WREG32_UVD_CTX(ixUVD_CGC_CTRL2, data);
}
int vce_v2_0_resume(struct radeon_device *rdev)
{
uint64_t addr = rdev->vce.gpu_addr;
uint32_t size;
WREG32_P(VCE_CLOCK_GATING_A, 0, ~(1 << 16));
WREG32_P(VCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000);
WREG32_P(VCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F);
WREG32(VCE_CLOCK_GATING_B, 0xf7);
WREG32(VCE_LMI_CTRL, 0x00398000);
WREG32_P(VCE_LMI_CACHE_CTRL, 0x0, ~0x1);
WREG32(VCE_LMI_SWAP_CNTL, 0);
WREG32(VCE_LMI_SWAP_CNTL1, 0);
WREG32(VCE_LMI_VM_CTRL, 0);
size = RADEON_GPU_PAGE_ALIGN(rdev->vce_fw->datasize);
WREG32(VCE_VCPU_CACHE_OFFSET0, addr & 0x7fffffff);
WREG32(VCE_VCPU_CACHE_SIZE0, size);
addr += size;
size = RADEON_VCE_STACK_SIZE;
WREG32(VCE_VCPU_CACHE_OFFSET1, addr & 0x7fffffff);
WREG32(VCE_VCPU_CACHE_SIZE1, size);
addr += size;
size = RADEON_VCE_HEAP_SIZE;
WREG32(VCE_VCPU_CACHE_OFFSET2, addr & 0x7fffffff);
WREG32(VCE_VCPU_CACHE_SIZE2, size);
WREG32_P(VCE_LMI_CTRL2, 0x0, ~0x100);
WREG32_P(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN,
~VCE_SYS_INT_TRAP_INTERRUPT_EN);
vce_v2_0_init_cg(rdev);
return 0;
}
int ni_mc_load_microcode(struct radeon_device *rdev)
{
const __be32 *fw_data;
u32 mem_type, running, blackout = 0;
u32 *io_mc_regs;
int i, ucode_size, regs_size;
if (!rdev->mc_fw)
return -EINVAL;
switch (rdev->family) {
case CHIP_BARTS:
io_mc_regs = (u32 *)&barts_io_mc_regs;
ucode_size = BTC_MC_UCODE_SIZE;
regs_size = BTC_IO_MC_REGS_SIZE;
break;
case CHIP_TURKS:
io_mc_regs = (u32 *)&turks_io_mc_regs;
ucode_size = BTC_MC_UCODE_SIZE;
regs_size = BTC_IO_MC_REGS_SIZE;
break;
case CHIP_CAICOS:
default:
io_mc_regs = (u32 *)&caicos_io_mc_regs;
ucode_size = BTC_MC_UCODE_SIZE;
regs_size = BTC_IO_MC_REGS_SIZE;
break;
case CHIP_CAYMAN:
io_mc_regs = (u32 *)&cayman_io_mc_regs;
ucode_size = CAYMAN_MC_UCODE_SIZE;
regs_size = BTC_IO_MC_REGS_SIZE;
break;
}
mem_type = (RREG32(MC_SEQ_MISC0) & MC_SEQ_MISC0_GDDR5_MASK) >> MC_SEQ_MISC0_GDDR5_SHIFT;
running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK;
if ((mem_type == MC_SEQ_MISC0_GDDR5_VALUE) && (running == 0)) {
if (running) {
blackout = RREG32(MC_SHARED_BLACKOUT_CNTL);
WREG32(MC_SHARED_BLACKOUT_CNTL, 1);
}
/* reset the engine and set to writable */
WREG32(MC_SEQ_SUP_CNTL, 0x00000008);
WREG32(MC_SEQ_SUP_CNTL, 0x00000010);
/* load mc io regs */
for (i = 0; i < regs_size; i++) {
WREG32(MC_SEQ_IO_DEBUG_INDEX, io_mc_regs[(i << 1)]);
WREG32(MC_SEQ_IO_DEBUG_DATA, io_mc_regs[(i << 1) + 1]);
}
/* load the MC ucode */
fw_data = (const __be32 *)rdev->mc_fw->data;
for (i = 0; i < ucode_size; i++)
WREG32(MC_SEQ_SUP_PGM, be32_to_cpup(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);
/* wait for training to complete */
for (i = 0; i < rdev->usec_timeout; i++) {
if (RREG32(MC_IO_PAD_CNTL_D0) & MEM_FALL_OUT_CMD)
break;
udelay(1);
}
if (running)
WREG32(MC_SHARED_BLACKOUT_CNTL, blackout);
}
return 0;
}
int cayman_pcie_gart_enable(struct radeon_device *rdev)
{
int i, r;
if (rdev->gart.robj == NULL) {
dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
return -EINVAL;
}
r = radeon_gart_table_vram_pin(rdev);
if (r)
return r;
radeon_gart_restore(rdev);
/* Setup TLB control */
WREG32(MC_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);
/* Setup L2 cache */
WREG32(VM_L2_CNTL, ENABLE_L2_CACHE |
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 |
L2_CACHE_BIGK_FRAGMENT_SIZE(6));
/* setup context0 */
WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
(u32)(rdev->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(0x15D4, 0);
WREG32(0x15D8, 0);
WREG32(0x15DC, 0);
/* empty context1-7 */
for (i = 1; i < 8; i++) {
WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR + (i << 2), 0);
WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR + (i << 2), 0);
WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (i << 2),
rdev->gart.table_addr >> 12);
}
/* enable context1-7 */
WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
(u32)(rdev->dummy_page.addr >> 12));
WREG32(VM_CONTEXT1_CNTL2, 0);
WREG32(VM_CONTEXT1_CNTL, 0);
WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
cayman_pcie_gart_tlb_flush(rdev);
DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
(unsigned)(rdev->mc.gtt_size >> 20),
(unsigned long long)rdev->gart.table_addr);
rdev->gart.ready = true;
return 0;
}
static bool r700_read_disabled_bios(struct radeon_device *rdev)
{
uint32_t viph_control;
uint32_t bus_cntl;
uint32_t d1vga_control;
uint32_t d2vga_control;
uint32_t vga_render_control;
uint32_t rom_cntl;
uint32_t cg_spll_func_cntl = 0;
uint32_t cg_spll_status;
bool r;
viph_control = RREG32(RADEON_VIPH_CONTROL);
bus_cntl = RREG32(R600_BUS_CNTL);
d1vga_control = RREG32(AVIVO_D1VGA_CONTROL);
d2vga_control = RREG32(AVIVO_D2VGA_CONTROL);
vga_render_control = RREG32(AVIVO_VGA_RENDER_CONTROL);
rom_cntl = RREG32(R600_ROM_CNTL);
/* disable VIP */
WREG32(RADEON_VIPH_CONTROL, (viph_control & ~RADEON_VIPH_EN));
/* enable the rom */
WREG32(R600_BUS_CNTL, (bus_cntl & ~R600_BIOS_ROM_DIS));
/* Disable VGA mode */
WREG32(AVIVO_D1VGA_CONTROL,
(d1vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_D2VGA_CONTROL,
(d2vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_VGA_RENDER_CONTROL,
(vga_render_control & ~AVIVO_VGA_VSTATUS_CNTL_MASK));
if (rdev->family == CHIP_RV730) {
cg_spll_func_cntl = RREG32(R600_CG_SPLL_FUNC_CNTL);
/* enable bypass mode */
WREG32(R600_CG_SPLL_FUNC_CNTL, (cg_spll_func_cntl |
R600_SPLL_BYPASS_EN));
/* wait for SPLL_CHG_STATUS to change to 1 */
cg_spll_status = 0;
while (!(cg_spll_status & R600_SPLL_CHG_STATUS))
cg_spll_status = RREG32(R600_CG_SPLL_STATUS);
WREG32(R600_ROM_CNTL, (rom_cntl & ~R600_SCK_OVERWRITE));
} else
WREG32(R600_ROM_CNTL, (rom_cntl | R600_SCK_OVERWRITE));
r = radeon_read_bios(rdev);
/* restore regs */
if (rdev->family == CHIP_RV730) {
WREG32(R600_CG_SPLL_FUNC_CNTL, cg_spll_func_cntl);
/* wait for SPLL_CHG_STATUS to change to 1 */
cg_spll_status = 0;
while (!(cg_spll_status & R600_SPLL_CHG_STATUS))
cg_spll_status = RREG32(R600_CG_SPLL_STATUS);
}
WREG32(RADEON_VIPH_CONTROL, viph_control);
WREG32(R600_BUS_CNTL, bus_cntl);
WREG32(AVIVO_D1VGA_CONTROL, d1vga_control);
WREG32(AVIVO_D2VGA_CONTROL, d2vga_control);
WREG32(AVIVO_VGA_RENDER_CONTROL, vga_render_control);
WREG32(R600_ROM_CNTL, rom_cntl);
return r;
}
static bool avivo_read_disabled_bios(struct radeon_device *rdev)
{
uint32_t seprom_cntl1;
uint32_t viph_control;
uint32_t bus_cntl;
uint32_t d1vga_control;
uint32_t d2vga_control;
uint32_t vga_render_control;
uint32_t gpiopad_a;
uint32_t gpiopad_en;
uint32_t gpiopad_mask;
bool r;
seprom_cntl1 = RREG32(RADEON_SEPROM_CNTL1);
viph_control = RREG32(RADEON_VIPH_CONTROL);
bus_cntl = RREG32(RV370_BUS_CNTL);
d1vga_control = RREG32(AVIVO_D1VGA_CONTROL);
d2vga_control = RREG32(AVIVO_D2VGA_CONTROL);
vga_render_control = RREG32(AVIVO_VGA_RENDER_CONTROL);
gpiopad_a = RREG32(RADEON_GPIOPAD_A);
gpiopad_en = RREG32(RADEON_GPIOPAD_EN);
gpiopad_mask = RREG32(RADEON_GPIOPAD_MASK);
WREG32(RADEON_SEPROM_CNTL1,
((seprom_cntl1 & ~RADEON_SCK_PRESCALE_MASK) |
(0xc << RADEON_SCK_PRESCALE_SHIFT)));
WREG32(RADEON_GPIOPAD_A, 0);
WREG32(RADEON_GPIOPAD_EN, 0);
WREG32(RADEON_GPIOPAD_MASK, 0);
/* disable VIP */
WREG32(RADEON_VIPH_CONTROL, (viph_control & ~RADEON_VIPH_EN));
/* enable the rom */
WREG32(RV370_BUS_CNTL, (bus_cntl & ~RV370_BUS_BIOS_DIS_ROM));
/* Disable VGA mode */
WREG32(AVIVO_D1VGA_CONTROL,
(d1vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_D2VGA_CONTROL,
(d2vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_VGA_RENDER_CONTROL,
(vga_render_control & ~AVIVO_VGA_VSTATUS_CNTL_MASK));
r = radeon_read_bios(rdev);
/* restore regs */
WREG32(RADEON_SEPROM_CNTL1, seprom_cntl1);
WREG32(RADEON_VIPH_CONTROL, viph_control);
WREG32(RV370_BUS_CNTL, bus_cntl);
WREG32(AVIVO_D1VGA_CONTROL, d1vga_control);
WREG32(AVIVO_D2VGA_CONTROL, d2vga_control);
WREG32(AVIVO_VGA_RENDER_CONTROL, vga_render_control);
WREG32(RADEON_GPIOPAD_A, gpiopad_a);
WREG32(RADEON_GPIOPAD_EN, gpiopad_en);
WREG32(RADEON_GPIOPAD_MASK, gpiopad_mask);
return r;
}
/**
* uvd_v5_0_start - start UVD block
*
* @adev: amdgpu_device pointer
*
* Setup and start the UVD block
*/
static int uvd_v5_0_start(struct amdgpu_device *adev)
{
struct amdgpu_ring *ring = &adev->uvd.inst->ring;
uint32_t rb_bufsz, tmp;
uint32_t lmi_swap_cntl;
uint32_t mp_swap_cntl;
int i, j, r;
/*disable DPG */
WREG32_P(mmUVD_POWER_STATUS, 0, ~(1 << 2));
/* disable byte swapping */
lmi_swap_cntl = 0;
mp_swap_cntl = 0;
uvd_v5_0_mc_resume(adev);
/* disable interupt */
WREG32_P(mmUVD_MASTINT_EN, 0, ~(1 << 1));
/* stall UMC and register bus before resetting VCPU */
WREG32_P(mmUVD_LMI_CTRL2, 1 << 8, ~(1 << 8));
mdelay(1);
/* put LMI, VCPU, RBC etc... into reset */
WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__LMI_SOFT_RESET_MASK |
UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK |
UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | UVD_SOFT_RESET__CSM_SOFT_RESET_MASK |
UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | UVD_SOFT_RESET__TAP_SOFT_RESET_MASK |
UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
mdelay(5);
/* take UVD block out of reset */
WREG32_P(mmSRBM_SOFT_RESET, 0, ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK);
mdelay(5);
/* initialize UVD memory controller */
WREG32(mmUVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) |
(1 << 21) | (1 << 9) | (1 << 20));
#ifdef __BIG_ENDIAN
/* swap (8 in 32) RB and IB */
lmi_swap_cntl = 0xa;
mp_swap_cntl = 0;
#endif
WREG32(mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
WREG32(mmUVD_MP_SWAP_CNTL, mp_swap_cntl);
WREG32(mmUVD_MPC_SET_MUXA0, 0x40c2040);
WREG32(mmUVD_MPC_SET_MUXA1, 0x0);
WREG32(mmUVD_MPC_SET_MUXB0, 0x40c2040);
WREG32(mmUVD_MPC_SET_MUXB1, 0x0);
WREG32(mmUVD_MPC_SET_ALU, 0);
WREG32(mmUVD_MPC_SET_MUX, 0x88);
/* take all subblocks out of reset, except VCPU */
WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(5);
/* enable VCPU clock */
WREG32(mmUVD_VCPU_CNTL, 1 << 9);
/* enable UMC */
WREG32_P(mmUVD_LMI_CTRL2, 0, ~(1 << 8));
/* boot up the VCPU */
WREG32(mmUVD_SOFT_RESET, 0);
mdelay(10);
for (i = 0; i < 10; ++i) {
uint32_t status;
for (j = 0; j < 100; ++j) {
status = RREG32(mmUVD_STATUS);
if (status & 2)
break;
mdelay(10);
}
r = 0;
if (status & 2)
break;
DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n");
WREG32_P(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(10);
WREG32_P(mmUVD_SOFT_RESET, 0, ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(10);
r = -1;
}
if (r) {
DRM_ERROR("UVD not responding, giving up!!!\n");
return r;
}
/* enable master interrupt */
WREG32_P(mmUVD_MASTINT_EN, 3 << 1, ~(3 << 1));
/* clear the bit 4 of UVD_STATUS */
WREG32_P(mmUVD_STATUS, 0, ~(2 << 1));
rb_bufsz = order_base_2(ring->ring_size);
tmp = 0;
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
/* force RBC into idle state */
WREG32(mmUVD_RBC_RB_CNTL, tmp);
/* set the write pointer delay */
WREG32(mmUVD_RBC_RB_WPTR_CNTL, 0);
/* set the wb address */
WREG32(mmUVD_RBC_RB_RPTR_ADDR, (upper_32_bits(ring->gpu_addr) >> 2));
/* programm the RB_BASE for ring buffer */
WREG32(mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
lower_32_bits(ring->gpu_addr));
WREG32(mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
upper_32_bits(ring->gpu_addr));
/* Initialize the ring buffer's read and write pointers */
WREG32(mmUVD_RBC_RB_RPTR, 0);
ring->wptr = RREG32(mmUVD_RBC_RB_RPTR);
WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
WREG32_P(mmUVD_RBC_RB_CNTL, 0, ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
return 0;
}
/**
* r600_dma_set_wptr - commit the write pointer
*
* @rdev: radeon_device pointer
* @ring: radeon ring pointer
*
* Write the wptr back to the hardware (r6xx+).
*/
void r600_dma_set_wptr(struct radeon_device *rdev,
struct radeon_ring *ring)
{
WREG32(DMA_RB_WPTR, (ring->wptr << 2) & 0x3fffc);
}
static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid)
{
struct amdgpu_device *adev = (struct amdgpu_device *) kgd;
WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
}
int radeon_agp_init(struct radeon_device *rdev)
{
#if __OS_HAS_AGP
struct radeon_agpmode_quirk *p = radeon_agpmode_quirk_list;
struct drm_agp_mode mode;
struct drm_agp_info info;
uint32_t agp_status;
int default_mode;
bool is_v3;
int ret;
/* Acquire AGP. */
if (!rdev->ddev->agp->acquired) {
ret = drm_agp_acquire(rdev->ddev);
if (ret) {
DRM_ERROR("Unable to acquire AGP: %d\n", ret);
return ret;
}
}
ret = drm_agp_info(rdev->ddev, &info);
if (ret) {
DRM_ERROR("Unable to get AGP info: %d\n", ret);
return ret;
}
mode.mode = info.mode;
agp_status = (RREG32(RADEON_AGP_STATUS) | RADEON_AGPv3_MODE) & mode.mode;
is_v3 = !!(agp_status & RADEON_AGPv3_MODE);
if (is_v3) {
default_mode = (agp_status & RADEON_AGPv3_8X_MODE) ? 8 : 4;
} else {
if (agp_status & RADEON_AGP_4X_MODE) {
default_mode = 4;
} else if (agp_status & RADEON_AGP_2X_MODE) {
default_mode = 2;
} else {
default_mode = 1;
}
}
/* Apply AGPMode Quirks */
while (p && p->chip_device != 0) {
if (info.id_vendor == p->hostbridge_vendor &&
info.id_device == p->hostbridge_device &&
rdev->pdev->vendor == p->chip_vendor &&
rdev->pdev->device == p->chip_device &&
rdev->pdev->subsystem_vendor == p->subsys_vendor &&
rdev->pdev->subsystem_device == p->subsys_device) {
default_mode = p->default_mode;
}
++p;
}
if (radeon_agpmode > 0) {
if ((radeon_agpmode < (is_v3 ? 4 : 1)) ||
(radeon_agpmode > (is_v3 ? 8 : 4)) ||
(radeon_agpmode & (radeon_agpmode - 1))) {
DRM_ERROR("Illegal AGP Mode: %d (valid %s), leaving at %d\n",
radeon_agpmode, is_v3 ? "4, 8" : "1, 2, 4",
default_mode);
radeon_agpmode = default_mode;
} else {
DRM_INFO("AGP mode requested: %d\n", radeon_agpmode);
}
} else {
radeon_agpmode = default_mode;
}
mode.mode &= ~RADEON_AGP_MODE_MASK;
if (is_v3) {
switch (radeon_agpmode) {
case 8:
mode.mode |= RADEON_AGPv3_8X_MODE;
break;
case 4:
default:
mode.mode |= RADEON_AGPv3_4X_MODE;
break;
}
} else {
switch (radeon_agpmode) {
case 4:
mode.mode |= RADEON_AGP_4X_MODE;
break;
case 2:
mode.mode |= RADEON_AGP_2X_MODE;
break;
case 1:
default:
mode.mode |= RADEON_AGP_1X_MODE;
break;
}
}
mode.mode &= ~RADEON_AGP_FW_MODE; /* disable fw */
ret = drm_agp_enable(rdev->ddev, mode);
if (ret) {
DRM_ERROR("Unable to enable AGP (mode = 0x%lx)\n", mode.mode);
return ret;
}
rdev->mc.agp_base = rdev->ddev->agp->agp_info.aper_base;
rdev->mc.gtt_size = rdev->ddev->agp->agp_info.aper_size << 20;
/* workaround some hw issues */
if (rdev->family < CHIP_R200) {
WREG32(RADEON_AGP_CNTL, RREG32(RADEON_AGP_CNTL) | 0x000e0000);
}
return 0;
#else
return 0;
#endif
}
static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
uint32_t queue_id, uint32_t __user *wptr)
{
struct vi_mqd *m;
uint32_t shadow_wptr, valid_wptr;
struct amdgpu_device *adev = get_amdgpu_device(kgd);
m = get_mqd(mqd);
valid_wptr = copy_from_user(&shadow_wptr, wptr, sizeof(shadow_wptr));
acquire_queue(kgd, pipe_id, queue_id);
WREG32(mmCP_MQD_CONTROL, m->cp_mqd_control);
WREG32(mmCP_MQD_BASE_ADDR, m->cp_mqd_base_addr_lo);
WREG32(mmCP_MQD_BASE_ADDR_HI, m->cp_mqd_base_addr_hi);
WREG32(mmCP_HQD_VMID, m->cp_hqd_vmid);
WREG32(mmCP_HQD_PERSISTENT_STATE, m->cp_hqd_persistent_state);
WREG32(mmCP_HQD_PIPE_PRIORITY, m->cp_hqd_pipe_priority);
WREG32(mmCP_HQD_QUEUE_PRIORITY, m->cp_hqd_queue_priority);
WREG32(mmCP_HQD_QUANTUM, m->cp_hqd_quantum);
WREG32(mmCP_HQD_PQ_BASE, m->cp_hqd_pq_base_lo);
WREG32(mmCP_HQD_PQ_BASE_HI, m->cp_hqd_pq_base_hi);
WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR, m->cp_hqd_pq_rptr_report_addr_lo);
WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
m->cp_hqd_pq_rptr_report_addr_hi);
if (valid_wptr > 0)
WREG32(mmCP_HQD_PQ_WPTR, shadow_wptr);
WREG32(mmCP_HQD_PQ_CONTROL, m->cp_hqd_pq_control);
WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, m->cp_hqd_pq_doorbell_control);
WREG32(mmCP_HQD_EOP_BASE_ADDR, m->cp_hqd_eop_base_addr_lo);
WREG32(mmCP_HQD_EOP_BASE_ADDR_HI, m->cp_hqd_eop_base_addr_hi);
WREG32(mmCP_HQD_EOP_CONTROL, m->cp_hqd_eop_control);
WREG32(mmCP_HQD_EOP_RPTR, m->cp_hqd_eop_rptr);
WREG32(mmCP_HQD_EOP_WPTR, m->cp_hqd_eop_wptr);
WREG32(mmCP_HQD_EOP_EVENTS, m->cp_hqd_eop_done_events);
WREG32(mmCP_HQD_CTX_SAVE_BASE_ADDR_LO, m->cp_hqd_ctx_save_base_addr_lo);
WREG32(mmCP_HQD_CTX_SAVE_BASE_ADDR_HI, m->cp_hqd_ctx_save_base_addr_hi);
WREG32(mmCP_HQD_CTX_SAVE_CONTROL, m->cp_hqd_ctx_save_control);
WREG32(mmCP_HQD_CNTL_STACK_OFFSET, m->cp_hqd_cntl_stack_offset);
WREG32(mmCP_HQD_CNTL_STACK_SIZE, m->cp_hqd_cntl_stack_size);
WREG32(mmCP_HQD_WG_STATE_OFFSET, m->cp_hqd_wg_state_offset);
WREG32(mmCP_HQD_CTX_SAVE_SIZE, m->cp_hqd_ctx_save_size);
WREG32(mmCP_HQD_IB_CONTROL, m->cp_hqd_ib_control);
WREG32(mmCP_HQD_DEQUEUE_REQUEST, m->cp_hqd_dequeue_request);
WREG32(mmCP_HQD_ERROR, m->cp_hqd_error);
WREG32(mmCP_HQD_EOP_WPTR_MEM, m->cp_hqd_eop_wptr_mem);
WREG32(mmCP_HQD_EOP_DONES, m->cp_hqd_eop_dones);
WREG32(mmCP_HQD_ACTIVE, m->cp_hqd_active);
release_queue(kgd);
return 0;
}
/**
* vce_v1_0_start - start VCE block
*
* @rdev: radeon_device pointer
*
* Setup and start the VCE block
*/
int vce_v1_0_start(struct radeon_device *rdev)
{
struct radeon_ring *ring;
int i, j, r;
/* set BUSY flag */
WREG32_P(VCE_STATUS, 1, ~1);
ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX];
WREG32(VCE_RB_RPTR, ring->wptr);
WREG32(VCE_RB_WPTR, ring->wptr);
WREG32(VCE_RB_BASE_LO, ring->gpu_addr);
WREG32(VCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
WREG32(VCE_RB_SIZE, ring->ring_size / 4);
ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX];
WREG32(VCE_RB_RPTR2, ring->wptr);
WREG32(VCE_RB_WPTR2, ring->wptr);
WREG32(VCE_RB_BASE_LO2, ring->gpu_addr);
WREG32(VCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
WREG32(VCE_RB_SIZE2, ring->ring_size / 4);
WREG32_P(VCE_VCPU_CNTL, VCE_CLK_EN, ~VCE_CLK_EN);
WREG32_P(VCE_SOFT_RESET,
VCE_ECPU_SOFT_RESET |
VCE_FME_SOFT_RESET, ~(
VCE_ECPU_SOFT_RESET |
VCE_FME_SOFT_RESET));
mdelay(100);
WREG32_P(VCE_SOFT_RESET, 0, ~(
VCE_ECPU_SOFT_RESET |
VCE_FME_SOFT_RESET));
for (i = 0; i < 10; ++i) {
uint32_t status;
for (j = 0; j < 100; ++j) {
status = RREG32(VCE_STATUS);
if (status & 2)
break;
mdelay(10);
}
r = 0;
if (status & 2)
break;
DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
WREG32_P(VCE_SOFT_RESET, VCE_ECPU_SOFT_RESET, ~VCE_ECPU_SOFT_RESET);
mdelay(10);
WREG32_P(VCE_SOFT_RESET, 0, ~VCE_ECPU_SOFT_RESET);
mdelay(10);
r = -1;
}
/* clear BUSY flag */
WREG32_P(VCE_STATUS, 0, ~1);
if (r) {
DRM_ERROR("VCE not responding, giving up!!!\n");
return r;
}
return 0;
}
int rv770_copy_bytes_to_smc(struct radeon_device *rdev,
u16 smc_start_address, const u8 *src,
u16 byte_count, u16 limit)
{
u32 data, original_data, extra_shift;
u16 addr;
int ret = 0;
if (smc_start_address & 3)
return -EINVAL;
if ((smc_start_address + byte_count) > limit)
return -EINVAL;
addr = smc_start_address;
spin_lock(&rdev->smc_idx_lock);
while (byte_count >= 4) {
/* SMC address space is BE */
data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3];
ret = rv770_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
WREG32(SMC_SRAM_DATA, data);
src += 4;
byte_count -= 4;
addr += 4;
}
/* RMW for final bytes */
if (byte_count > 0) {
data = 0;
ret = rv770_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
original_data = RREG32(SMC_SRAM_DATA);
extra_shift = 8 * (4 - byte_count);
while (byte_count > 0) {
/* SMC address space is BE */
data = (data << 8) + *src++;
byte_count--;
}
data <<= extra_shift;
data |= (original_data & ~((~0UL) << extra_shift));
ret = rv770_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
WREG32(SMC_SRAM_DATA, data);
}
done:
spin_unlock(&rdev->smc_idx_lock);
return ret;
}
static void vce_v2_0_disable_cg(struct radeon_device *rdev)
{
WREG32(VCE_CGTT_CLK_OVERRIDE, 7);
}
/**
* uvd_v5_0_ring_set_wptr - set write pointer
*
* @ring: amdgpu_ring pointer
*
* Commits the write pointer to the hardware
*/
static void uvd_v5_0_ring_set_wptr(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
WREG32(mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr));
}
/**
* r600_dma_resume - setup and start the async dma engine
*
* @rdev: radeon_device pointer
*
* Set up the DMA ring buffer and enable it. (r6xx-evergreen).
* Returns 0 for success, error for failure.
*/
int r600_dma_resume(struct radeon_device *rdev)
{
struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
u32 rb_cntl, dma_cntl, ib_cntl;
u32 rb_bufsz;
int r;
WREG32(DMA_SEM_INCOMPLETE_TIMER_CNTL, 0);
WREG32(DMA_SEM_WAIT_FAIL_TIMER_CNTL, 0);
/* Set ring buffer size in dwords */
rb_bufsz = order_base_2(ring->ring_size / 4);
rb_cntl = rb_bufsz << 1;
#ifdef __BIG_ENDIAN
rb_cntl |= DMA_RB_SWAP_ENABLE | DMA_RPTR_WRITEBACK_SWAP_ENABLE;
#endif
WREG32(DMA_RB_CNTL, rb_cntl);
/* Initialize the ring buffer's read and write pointers */
WREG32(DMA_RB_RPTR, 0);
WREG32(DMA_RB_WPTR, 0);
/* set the wb address whether it's enabled or not */
WREG32(DMA_RB_RPTR_ADDR_HI,
upper_32_bits(rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFF);
WREG32(DMA_RB_RPTR_ADDR_LO,
((rdev->wb.gpu_addr + R600_WB_DMA_RPTR_OFFSET) & 0xFFFFFFFC));
if (rdev->wb.enabled)
rb_cntl |= DMA_RPTR_WRITEBACK_ENABLE;
WREG32(DMA_RB_BASE, ring->gpu_addr >> 8);
/* enable DMA IBs */
ib_cntl = DMA_IB_ENABLE;
#ifdef __BIG_ENDIAN
ib_cntl |= DMA_IB_SWAP_ENABLE;
#endif
WREG32(DMA_IB_CNTL, ib_cntl);
dma_cntl = RREG32(DMA_CNTL);
dma_cntl &= ~CTXEMPTY_INT_ENABLE;
WREG32(DMA_CNTL, dma_cntl);
if (rdev->family >= CHIP_RV770)
WREG32(DMA_MODE, 1);
ring->wptr = 0;
WREG32(DMA_RB_WPTR, ring->wptr << 2);
WREG32(DMA_RB_CNTL, rb_cntl | DMA_RB_ENABLE);
ring->ready = true;
r = radeon_ring_test(rdev, R600_RING_TYPE_DMA_INDEX, ring);
if (r) {
ring->ready = false;
return r;
}
if (rdev->asic->copy.copy_ring_index == R600_RING_TYPE_DMA_INDEX)
radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
return 0;
}
int kv_copy_bytes_to_smc(struct radeon_device *rdev,
u32 smc_start_address,
const u8 *src, u32 byte_count, u32 limit)
{
int ret;
u32 data, original_data, addr, extra_shift, t_byte, count, mask;
if ((smc_start_address + byte_count) > limit)
return -EINVAL;
addr = smc_start_address;
t_byte = addr & 3;
/* RMW for the initial bytes */
if (t_byte != 0) {
addr -= t_byte;
ret = kv_set_smc_sram_address(rdev, addr, limit);
if (ret)
return ret;
original_data = RREG32(SMC_IND_DATA_0);
data = 0;
mask = 0;
count = 4;
while (count > 0) {
if (t_byte > 0) {
mask = (mask << 8) | 0xff;
t_byte--;
} else if (byte_count > 0) {
data = (data << 8) + *src++;
byte_count--;
mask <<= 8;
} else {
data <<= 8;
mask = (mask << 8) | 0xff;
}
count--;
}
data |= original_data & mask;
ret = kv_set_smc_sram_address(rdev, addr, limit);
if (ret)
return ret;
WREG32(SMC_IND_DATA_0, data);
addr += 4;
}
while (byte_count >= 4) {
/* SMC address space is BE */
data = (src[0] << 24) + (src[1] << 16) + (src[2] << 8) + src[3];
ret = kv_set_smc_sram_address(rdev, addr, limit);
if (ret)
return ret;
WREG32(SMC_IND_DATA_0, data);
src += 4;
byte_count -= 4;
addr += 4;
}
/* RMW for the final bytes */
if (byte_count > 0) {
data = 0;
ret = kv_set_smc_sram_address(rdev, addr, limit);
if (ret)
return ret;
original_data= RREG32(SMC_IND_DATA_0);
extra_shift = 8 * (4 - byte_count);
while (byte_count > 0) {
/* SMC address space is BE */
data = (data << 8) + *src++;
byte_count--;
}
data <<= extra_shift;
data |= (original_data & ~((~0UL) << extra_shift));
ret = kv_set_smc_sram_address(rdev, addr, limit);
if (ret)
return ret;
WREG32(SMC_IND_DATA_0, data);
}
return 0;
}
int ci_copy_bytes_to_smc(struct radeon_device *rdev,
u32 smc_start_address,
const u8 *src, u32 byte_count, u32 limit)
{
unsigned long flags;
u32 data, original_data;
u32 addr;
u32 extra_shift;
int ret = 0;
if (smc_start_address & 3)
return -EINVAL;
if ((smc_start_address + byte_count) > limit)
return -EINVAL;
addr = smc_start_address;
spin_lock_irqsave(&rdev->smc_idx_lock, flags);
while (byte_count >= 4) {
/* SMC address space is BE */
data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3];
ret = ci_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
WREG32(SMC_IND_DATA_0, data);
src += 4;
byte_count -= 4;
addr += 4;
}
/* RMW for the final bytes */
if (byte_count > 0) {
data = 0;
ret = ci_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
original_data = RREG32(SMC_IND_DATA_0);
extra_shift = 8 * (4 - byte_count);
while (byte_count > 0) {
data = (data << 8) + *src++;
byte_count--;
}
data <<= extra_shift;
data |= (original_data & ~((~0UL) << extra_shift));
ret = ci_set_smc_sram_address(rdev, addr, limit);
if (ret)
goto done;
WREG32(SMC_IND_DATA_0, data);
}
done:
spin_unlock_irqrestore(&rdev->smc_idx_lock, flags);
return ret;
}
static void cz_set_uvd_clock_gating_branches(struct amdgpu_device *adev,
bool enable)
{
u32 data, data1;
data = RREG32(mmUVD_CGC_GATE);
data1 = RREG32(mmUVD_SUVD_CGC_GATE);
if (enable) {
data |= UVD_CGC_GATE__SYS_MASK |
UVD_CGC_GATE__UDEC_MASK |
UVD_CGC_GATE__MPEG2_MASK |
UVD_CGC_GATE__RBC_MASK |
UVD_CGC_GATE__LMI_MC_MASK |
UVD_CGC_GATE__IDCT_MASK |
UVD_CGC_GATE__MPRD_MASK |
UVD_CGC_GATE__MPC_MASK |
UVD_CGC_GATE__LBSI_MASK |
UVD_CGC_GATE__LRBBM_MASK |
UVD_CGC_GATE__UDEC_RE_MASK |
UVD_CGC_GATE__UDEC_CM_MASK |
UVD_CGC_GATE__UDEC_IT_MASK |
UVD_CGC_GATE__UDEC_DB_MASK |
UVD_CGC_GATE__UDEC_MP_MASK |
UVD_CGC_GATE__WCB_MASK |
UVD_CGC_GATE__VCPU_MASK |
UVD_CGC_GATE__SCPU_MASK;
data1 |= UVD_SUVD_CGC_GATE__SRE_MASK |
UVD_SUVD_CGC_GATE__SIT_MASK |
UVD_SUVD_CGC_GATE__SMP_MASK |
UVD_SUVD_CGC_GATE__SCM_MASK |
UVD_SUVD_CGC_GATE__SDB_MASK |
UVD_SUVD_CGC_GATE__SRE_H264_MASK |
UVD_SUVD_CGC_GATE__SRE_HEVC_MASK |
UVD_SUVD_CGC_GATE__SIT_H264_MASK |
UVD_SUVD_CGC_GATE__SIT_HEVC_MASK |
UVD_SUVD_CGC_GATE__SCM_H264_MASK |
UVD_SUVD_CGC_GATE__SCM_HEVC_MASK |
UVD_SUVD_CGC_GATE__SDB_H264_MASK |
UVD_SUVD_CGC_GATE__SDB_HEVC_MASK;
} else {
data &= ~(UVD_CGC_GATE__SYS_MASK |
UVD_CGC_GATE__UDEC_MASK |
UVD_CGC_GATE__MPEG2_MASK |
UVD_CGC_GATE__RBC_MASK |
UVD_CGC_GATE__LMI_MC_MASK |
UVD_CGC_GATE__LMI_UMC_MASK |
UVD_CGC_GATE__IDCT_MASK |
UVD_CGC_GATE__MPRD_MASK |
UVD_CGC_GATE__MPC_MASK |
UVD_CGC_GATE__LBSI_MASK |
UVD_CGC_GATE__LRBBM_MASK |
UVD_CGC_GATE__UDEC_RE_MASK |
UVD_CGC_GATE__UDEC_CM_MASK |
UVD_CGC_GATE__UDEC_IT_MASK |
UVD_CGC_GATE__UDEC_DB_MASK |
UVD_CGC_GATE__UDEC_MP_MASK |
UVD_CGC_GATE__WCB_MASK |
UVD_CGC_GATE__VCPU_MASK |
UVD_CGC_GATE__SCPU_MASK);
data1 &= ~(UVD_SUVD_CGC_GATE__SRE_MASK |
UVD_SUVD_CGC_GATE__SIT_MASK |
UVD_SUVD_CGC_GATE__SMP_MASK |
UVD_SUVD_CGC_GATE__SCM_MASK |
UVD_SUVD_CGC_GATE__SDB_MASK |
UVD_SUVD_CGC_GATE__SRE_H264_MASK |
UVD_SUVD_CGC_GATE__SRE_HEVC_MASK |
UVD_SUVD_CGC_GATE__SIT_H264_MASK |
UVD_SUVD_CGC_GATE__SIT_HEVC_MASK |
UVD_SUVD_CGC_GATE__SCM_H264_MASK |
UVD_SUVD_CGC_GATE__SCM_HEVC_MASK |
UVD_SUVD_CGC_GATE__SDB_H264_MASK |
UVD_SUVD_CGC_GATE__SDB_HEVC_MASK);
}
WREG32(mmUVD_CGC_GATE, data);
WREG32(mmUVD_SUVD_CGC_GATE, data1);
}
static bool r600_read_disabled_bios(struct radeon_device *rdev)
{
uint32_t viph_control;
uint32_t bus_cntl;
uint32_t d1vga_control;
uint32_t d2vga_control;
uint32_t vga_render_control;
uint32_t rom_cntl;
uint32_t general_pwrmgt;
uint32_t low_vid_lower_gpio_cntl;
uint32_t medium_vid_lower_gpio_cntl;
uint32_t high_vid_lower_gpio_cntl;
uint32_t ctxsw_vid_lower_gpio_cntl;
uint32_t lower_gpio_enable;
bool r;
viph_control = RREG32(RADEON_VIPH_CONTROL);
bus_cntl = RREG32(R600_BUS_CNTL);
d1vga_control = RREG32(AVIVO_D1VGA_CONTROL);
d2vga_control = RREG32(AVIVO_D2VGA_CONTROL);
vga_render_control = RREG32(AVIVO_VGA_RENDER_CONTROL);
rom_cntl = RREG32(R600_ROM_CNTL);
general_pwrmgt = RREG32(R600_GENERAL_PWRMGT);
low_vid_lower_gpio_cntl = RREG32(R600_LOW_VID_LOWER_GPIO_CNTL);
medium_vid_lower_gpio_cntl = RREG32(R600_MEDIUM_VID_LOWER_GPIO_CNTL);
high_vid_lower_gpio_cntl = RREG32(R600_HIGH_VID_LOWER_GPIO_CNTL);
ctxsw_vid_lower_gpio_cntl = RREG32(R600_CTXSW_VID_LOWER_GPIO_CNTL);
lower_gpio_enable = RREG32(R600_LOWER_GPIO_ENABLE);
/* disable VIP */
WREG32(RADEON_VIPH_CONTROL, (viph_control & ~RADEON_VIPH_EN));
/* enable the rom */
WREG32(R600_BUS_CNTL, (bus_cntl & ~R600_BIOS_ROM_DIS));
/* Disable VGA mode */
WREG32(AVIVO_D1VGA_CONTROL,
(d1vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_D2VGA_CONTROL,
(d2vga_control & ~(AVIVO_DVGA_CONTROL_MODE_ENABLE |
AVIVO_DVGA_CONTROL_TIMING_SELECT)));
WREG32(AVIVO_VGA_RENDER_CONTROL,
(vga_render_control & ~AVIVO_VGA_VSTATUS_CNTL_MASK));
WREG32(R600_ROM_CNTL,
((rom_cntl & ~R600_SCK_PRESCALE_CRYSTAL_CLK_MASK) |
(1 << R600_SCK_PRESCALE_CRYSTAL_CLK_SHIFT) |
R600_SCK_OVERWRITE));
WREG32(R600_GENERAL_PWRMGT, (general_pwrmgt & ~R600_OPEN_DRAIN_PADS));
WREG32(R600_LOW_VID_LOWER_GPIO_CNTL,
(low_vid_lower_gpio_cntl & ~0x400));
WREG32(R600_MEDIUM_VID_LOWER_GPIO_CNTL,
(medium_vid_lower_gpio_cntl & ~0x400));
WREG32(R600_HIGH_VID_LOWER_GPIO_CNTL,
(high_vid_lower_gpio_cntl & ~0x400));
WREG32(R600_CTXSW_VID_LOWER_GPIO_CNTL,
(ctxsw_vid_lower_gpio_cntl & ~0x400));
WREG32(R600_LOWER_GPIO_ENABLE, (lower_gpio_enable | 0x400));
r = radeon_read_bios(rdev);
/* restore regs */
WREG32(RADEON_VIPH_CONTROL, viph_control);
WREG32(R600_BUS_CNTL, bus_cntl);
WREG32(AVIVO_D1VGA_CONTROL, d1vga_control);
WREG32(AVIVO_D2VGA_CONTROL, d2vga_control);
WREG32(AVIVO_VGA_RENDER_CONTROL, vga_render_control);
WREG32(R600_ROM_CNTL, rom_cntl);
WREG32(R600_GENERAL_PWRMGT, general_pwrmgt);
WREG32(R600_LOW_VID_LOWER_GPIO_CNTL, low_vid_lower_gpio_cntl);
WREG32(R600_MEDIUM_VID_LOWER_GPIO_CNTL, medium_vid_lower_gpio_cntl);
WREG32(R600_HIGH_VID_LOWER_GPIO_CNTL, high_vid_lower_gpio_cntl);
WREG32(R600_CTXSW_VID_LOWER_GPIO_CNTL, ctxsw_vid_lower_gpio_cntl);
WREG32(R600_LOWER_GPIO_ENABLE, lower_gpio_enable);
return r;
}
/**
* uvd_v6_0_start - start UVD block
*
* @adev: amdgpu_device pointer
*
* Setup and start the UVD block
*/
static int uvd_v6_0_start(struct amdgpu_device *adev)
{
struct amdgpu_ring *ring = &adev->uvd.ring;
uint32_t rb_bufsz, tmp;
uint32_t lmi_swap_cntl;
uint32_t mp_swap_cntl;
int i, j, r;
/* disable DPG */
WREG32_P(mmUVD_POWER_STATUS, 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK);
/* disable byte swapping */
lmi_swap_cntl = 0;
mp_swap_cntl = 0;
uvd_v6_0_mc_resume(adev);
/* Set dynamic clock gating in S/W control mode */
if (adev->cg_flags & AMD_CG_SUPPORT_UVD_MGCG) {
uvd_v6_0_set_sw_clock_gating(adev);
} else {
/* disable clock gating */
uint32_t data = RREG32(mmUVD_CGC_CTRL);
data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
WREG32(mmUVD_CGC_CTRL, data);
}
/* disable interupt */
WREG32_P(mmUVD_MASTINT_EN, 0, ~UVD_MASTINT_EN__VCPU_EN_MASK);
/* stall UMC and register bus before resetting VCPU */
WREG32_P(mmUVD_LMI_CTRL2, UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
mdelay(1);
/* put LMI, VCPU, RBC etc... into reset */
WREG32(mmUVD_SOFT_RESET,
UVD_SOFT_RESET__LMI_SOFT_RESET_MASK |
UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK |
UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK |
UVD_SOFT_RESET__RBC_SOFT_RESET_MASK |
UVD_SOFT_RESET__CSM_SOFT_RESET_MASK |
UVD_SOFT_RESET__CXW_SOFT_RESET_MASK |
UVD_SOFT_RESET__TAP_SOFT_RESET_MASK |
UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
mdelay(5);
/* take UVD block out of reset */
WREG32_P(mmSRBM_SOFT_RESET, 0, ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK);
mdelay(5);
/* initialize UVD memory controller */
WREG32(mmUVD_LMI_CTRL,
(0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) |
UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK |
UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK |
UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK |
UVD_LMI_CTRL__REQ_MODE_MASK |
UVD_LMI_CTRL__DISABLE_ON_FWV_FAIL_MASK);
#ifdef __BIG_ENDIAN
/* swap (8 in 32) RB and IB */
lmi_swap_cntl = 0xa;
mp_swap_cntl = 0;
#endif
WREG32(mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
WREG32(mmUVD_MP_SWAP_CNTL, mp_swap_cntl);
WREG32(mmUVD_MPC_SET_MUXA0, 0x40c2040);
WREG32(mmUVD_MPC_SET_MUXA1, 0x0);
WREG32(mmUVD_MPC_SET_MUXB0, 0x40c2040);
WREG32(mmUVD_MPC_SET_MUXB1, 0x0);
WREG32(mmUVD_MPC_SET_ALU, 0);
WREG32(mmUVD_MPC_SET_MUX, 0x88);
/* take all subblocks out of reset, except VCPU */
WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(5);
/* enable VCPU clock */
WREG32(mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK);
/* enable UMC */
WREG32_P(mmUVD_LMI_CTRL2, 0, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK);
/* boot up the VCPU */
WREG32(mmUVD_SOFT_RESET, 0);
mdelay(10);
for (i = 0; i < 10; ++i) {
uint32_t status;
for (j = 0; j < 100; ++j) {
status = RREG32(mmUVD_STATUS);
if (status & 2)
break;
mdelay(10);
}
r = 0;
if (status & 2)
break;
DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n");
WREG32_P(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(10);
WREG32_P(mmUVD_SOFT_RESET, 0,
~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
mdelay(10);
r = -1;
}
if (r) {
DRM_ERROR("UVD not responding, giving up!!!\n");
return r;
}
/* enable master interrupt */
WREG32_P(mmUVD_MASTINT_EN,
(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK),
~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK));
/* clear the bit 4 of UVD_STATUS */
WREG32_P(mmUVD_STATUS, 0, ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT));
rb_bufsz = order_base_2(ring->ring_size);
tmp = 0;
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1);
tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1);
/* force RBC into idle state */
WREG32(mmUVD_RBC_RB_CNTL, tmp);
/* set the write pointer delay */
WREG32(mmUVD_RBC_RB_WPTR_CNTL, 0);
/* set the wb address */
WREG32(mmUVD_RBC_RB_RPTR_ADDR, (upper_32_bits(ring->gpu_addr) >> 2));
/* programm the RB_BASE for ring buffer */
WREG32(mmUVD_LMI_RBC_RB_64BIT_BAR_LOW,
lower_32_bits(ring->gpu_addr));
WREG32(mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH,
upper_32_bits(ring->gpu_addr));
/* Initialize the ring buffer's read and write pointers */
WREG32(mmUVD_RBC_RB_RPTR, 0);
ring->wptr = RREG32(mmUVD_RBC_RB_RPTR);
WREG32(mmUVD_RBC_RB_WPTR, ring->wptr);
WREG32_P(mmUVD_RBC_RB_CNTL, 0, ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK);
return 0;
}
static bool legacy_read_disabled_bios(struct radeon_device *rdev)
{
uint32_t seprom_cntl1;
uint32_t viph_control;
uint32_t bus_cntl;
uint32_t crtc_gen_cntl;
uint32_t crtc2_gen_cntl;
uint32_t crtc_ext_cntl;
uint32_t fp2_gen_cntl;
bool r;
seprom_cntl1 = RREG32(RADEON_SEPROM_CNTL1);
viph_control = RREG32(RADEON_VIPH_CONTROL);
if (rdev->flags & RADEON_IS_PCIE)
bus_cntl = RREG32(RV370_BUS_CNTL);
else
bus_cntl = RREG32(RADEON_BUS_CNTL);
crtc_gen_cntl = RREG32(RADEON_CRTC_GEN_CNTL);
crtc2_gen_cntl = 0;
crtc_ext_cntl = RREG32(RADEON_CRTC_EXT_CNTL);
fp2_gen_cntl = 0;
if (rdev->ddev->pci_device == PCI_DEVICE_ID_ATI_RADEON_QY) {
fp2_gen_cntl = RREG32(RADEON_FP2_GEN_CNTL);
}
if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
crtc2_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL);
}
WREG32(RADEON_SEPROM_CNTL1,
((seprom_cntl1 & ~RADEON_SCK_PRESCALE_MASK) |
(0xc << RADEON_SCK_PRESCALE_SHIFT)));
/* disable VIP */
WREG32(RADEON_VIPH_CONTROL, (viph_control & ~RADEON_VIPH_EN));
/* enable the rom */
if (rdev->flags & RADEON_IS_PCIE)
WREG32(RV370_BUS_CNTL, (bus_cntl & ~RV370_BUS_BIOS_DIS_ROM));
else
WREG32(RADEON_BUS_CNTL, (bus_cntl & ~RADEON_BUS_BIOS_DIS_ROM));
/* Turn off mem requests and CRTC for both controllers */
WREG32(RADEON_CRTC_GEN_CNTL,
((crtc_gen_cntl & ~RADEON_CRTC_EN) |
(RADEON_CRTC_DISP_REQ_EN_B |
RADEON_CRTC_EXT_DISP_EN)));
if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
WREG32(RADEON_CRTC2_GEN_CNTL,
((crtc2_gen_cntl & ~RADEON_CRTC2_EN) |
RADEON_CRTC2_DISP_REQ_EN_B));
}
/* Turn off CRTC */
WREG32(RADEON_CRTC_EXT_CNTL,
((crtc_ext_cntl & ~RADEON_CRTC_CRT_ON) |
(RADEON_CRTC_SYNC_TRISTAT |
RADEON_CRTC_DISPLAY_DIS)));
if (rdev->ddev->pci_device == PCI_DEVICE_ID_ATI_RADEON_QY) {
WREG32(RADEON_FP2_GEN_CNTL, (fp2_gen_cntl & ~RADEON_FP2_ON));
}
r = radeon_read_bios(rdev);
/* restore regs */
WREG32(RADEON_SEPROM_CNTL1, seprom_cntl1);
WREG32(RADEON_VIPH_CONTROL, viph_control);
if (rdev->flags & RADEON_IS_PCIE)
WREG32(RV370_BUS_CNTL, bus_cntl);
else
WREG32(RADEON_BUS_CNTL, bus_cntl);
WREG32(RADEON_CRTC_GEN_CNTL, crtc_gen_cntl);
if (!(rdev->flags & RADEON_SINGLE_CRTC)) {
WREG32(RADEON_CRTC2_GEN_CNTL, crtc2_gen_cntl);
}
WREG32(RADEON_CRTC_EXT_CNTL, crtc_ext_cntl);
if (rdev->ddev->pci_device == PCI_DEVICE_ID_ATI_RADEON_QY) {
WREG32(RADEON_FP2_GEN_CNTL, fp2_gen_cntl);
}
return r;
}
/**
* uvd_v6_0_ring_set_wptr - set write pointer
*
* @ring: amdgpu_ring pointer
*
* Commits the write pointer to the hardware
*/
static void uvd_v6_0_ring_set_wptr(struct amdgpu_ring *ring)
{
struct amdgpu_device *adev = ring->adev;
WREG32(mmUVD_RBC_RB_WPTR, ring->wptr);
}
/*
* Core functions
*/
static void cayman_gpu_init(struct radeon_device *rdev)
{
u32 gb_addr_config = 0;
u32 mc_shared_chmap, mc_arb_ramcfg;
u32 cgts_tcc_disable;
u32 sx_debug_1;
u32 smx_dc_ctl0;
u32 cgts_sm_ctrl_reg;
u32 hdp_host_path_cntl;
u32 tmp;
u32 disabled_rb_mask;
int i, j;
switch (rdev->family) {
case CHIP_CAYMAN:
rdev->config.cayman.max_shader_engines = 2;
rdev->config.cayman.max_pipes_per_simd = 4;
rdev->config.cayman.max_tile_pipes = 8;
rdev->config.cayman.max_simds_per_se = 12;
rdev->config.cayman.max_backends_per_se = 4;
rdev->config.cayman.max_texture_channel_caches = 8;
rdev->config.cayman.max_gprs = 256;
rdev->config.cayman.max_threads = 256;
rdev->config.cayman.max_gs_threads = 32;
rdev->config.cayman.max_stack_entries = 512;
rdev->config.cayman.sx_num_of_sets = 8;
rdev->config.cayman.sx_max_export_size = 256;
rdev->config.cayman.sx_max_export_pos_size = 64;
rdev->config.cayman.sx_max_export_smx_size = 192;
rdev->config.cayman.max_hw_contexts = 8;
rdev->config.cayman.sq_num_cf_insts = 2;
rdev->config.cayman.sc_prim_fifo_size = 0x100;
rdev->config.cayman.sc_hiz_tile_fifo_size = 0x30;
rdev->config.cayman.sc_earlyz_tile_fifo_size = 0x130;
gb_addr_config = CAYMAN_GB_ADDR_CONFIG_GOLDEN;
break;
case CHIP_ARUBA:
default:
rdev->config.cayman.max_shader_engines = 1;
rdev->config.cayman.max_pipes_per_simd = 4;
rdev->config.cayman.max_tile_pipes = 2;
if ((rdev->pdev->device == 0x9900) ||
(rdev->pdev->device == 0x9901) ||
(rdev->pdev->device == 0x9905) ||
(rdev->pdev->device == 0x9906) ||
(rdev->pdev->device == 0x9907) ||
(rdev->pdev->device == 0x9908) ||
(rdev->pdev->device == 0x9909) ||
(rdev->pdev->device == 0x9910) ||
(rdev->pdev->device == 0x9917)) {
rdev->config.cayman.max_simds_per_se = 6;
rdev->config.cayman.max_backends_per_se = 2;
} else if ((rdev->pdev->device == 0x9903) ||
(rdev->pdev->device == 0x9904) ||
(rdev->pdev->device == 0x990A) ||
(rdev->pdev->device == 0x9913) ||
(rdev->pdev->device == 0x9918)) {
rdev->config.cayman.max_simds_per_se = 4;
rdev->config.cayman.max_backends_per_se = 2;
} else if ((rdev->pdev->device == 0x9919) ||
(rdev->pdev->device == 0x9990) ||
(rdev->pdev->device == 0x9991) ||
(rdev->pdev->device == 0x9994) ||
(rdev->pdev->device == 0x99A0)) {
rdev->config.cayman.max_simds_per_se = 3;
rdev->config.cayman.max_backends_per_se = 1;
} else {
rdev->config.cayman.max_simds_per_se = 2;
rdev->config.cayman.max_backends_per_se = 1;
}
rdev->config.cayman.max_texture_channel_caches = 2;
rdev->config.cayman.max_gprs = 256;
rdev->config.cayman.max_threads = 256;
rdev->config.cayman.max_gs_threads = 32;
rdev->config.cayman.max_stack_entries = 512;
rdev->config.cayman.sx_num_of_sets = 8;
rdev->config.cayman.sx_max_export_size = 256;
rdev->config.cayman.sx_max_export_pos_size = 64;
rdev->config.cayman.sx_max_export_smx_size = 192;
rdev->config.cayman.max_hw_contexts = 8;
rdev->config.cayman.sq_num_cf_insts = 2;
rdev->config.cayman.sc_prim_fifo_size = 0x40;
rdev->config.cayman.sc_hiz_tile_fifo_size = 0x30;
rdev->config.cayman.sc_earlyz_tile_fifo_size = 0x130;
gb_addr_config = ARUBA_GB_ADDR_CONFIG_GOLDEN;
break;
}
/* Initialize HDP */
for (i = 0, j = 0; i < 32; i++, j += 0x18) {
WREG32((0x2c14 + j), 0x00000000);
WREG32((0x2c18 + j), 0x00000000);
WREG32((0x2c1c + j), 0x00000000);
WREG32((0x2c20 + j), 0x00000000);
WREG32((0x2c24 + j), 0x00000000);
}
WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff));
evergreen_fix_pci_max_read_req_size(rdev);
mc_shared_chmap = RREG32(MC_SHARED_CHMAP);
mc_arb_ramcfg = RREG32(MC_ARB_RAMCFG);
tmp = (mc_arb_ramcfg & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT;
rdev->config.cayman.mem_row_size_in_kb = (4 * (1 << (8 + tmp))) / 1024;
if (rdev->config.cayman.mem_row_size_in_kb > 4)
rdev->config.cayman.mem_row_size_in_kb = 4;
/* XXX use MC settings? */
rdev->config.cayman.shader_engine_tile_size = 32;
rdev->config.cayman.num_gpus = 1;
rdev->config.cayman.multi_gpu_tile_size = 64;
tmp = (gb_addr_config & NUM_PIPES_MASK) >> NUM_PIPES_SHIFT;
rdev->config.cayman.num_tile_pipes = (1 << tmp);
tmp = (gb_addr_config & PIPE_INTERLEAVE_SIZE_MASK) >> PIPE_INTERLEAVE_SIZE_SHIFT;
rdev->config.cayman.mem_max_burst_length_bytes = (tmp + 1) * 256;
tmp = (gb_addr_config & NUM_SHADER_ENGINES_MASK) >> NUM_SHADER_ENGINES_SHIFT;
rdev->config.cayman.num_shader_engines = tmp + 1;
tmp = (gb_addr_config & NUM_GPUS_MASK) >> NUM_GPUS_SHIFT;
rdev->config.cayman.num_gpus = tmp + 1;
tmp = (gb_addr_config & MULTI_GPU_TILE_SIZE_MASK) >> MULTI_GPU_TILE_SIZE_SHIFT;
rdev->config.cayman.multi_gpu_tile_size = 1 << tmp;
tmp = (gb_addr_config & ROW_SIZE_MASK) >> ROW_SIZE_SHIFT;
rdev->config.cayman.mem_row_size_in_kb = 1 << tmp;
/* setup tiling info dword. gb_addr_config is not adequate since it does
* not have bank info, so create a custom tiling dword.
* bits 3:0 num_pipes
* bits 7:4 num_banks
* bits 11:8 group_size
* bits 15:12 row_size
*/
rdev->config.cayman.tile_config = 0;
switch (rdev->config.cayman.num_tile_pipes) {
case 1:
default:
rdev->config.cayman.tile_config |= (0 << 0);
break;
case 2:
rdev->config.cayman.tile_config |= (1 << 0);
break;
case 4:
rdev->config.cayman.tile_config |= (2 << 0);
break;
case 8:
rdev->config.cayman.tile_config |= (3 << 0);
break;
}
/* num banks is 8 on all fusion asics. 0 = 4, 1 = 8, 2 = 16 */
if (rdev->flags & RADEON_IS_IGP)
rdev->config.cayman.tile_config |= 1 << 4;
else {
switch ((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) {
case 0: /* four banks */
rdev->config.cayman.tile_config |= 0 << 4;
break;
case 1: /* eight banks */
rdev->config.cayman.tile_config |= 1 << 4;
break;
case 2: /* sixteen banks */
default:
rdev->config.cayman.tile_config |= 2 << 4;
break;
}
}
rdev->config.cayman.tile_config |=
((gb_addr_config & PIPE_INTERLEAVE_SIZE_MASK) >> PIPE_INTERLEAVE_SIZE_SHIFT) << 8;
rdev->config.cayman.tile_config |=
((gb_addr_config & ROW_SIZE_MASK) >> ROW_SIZE_SHIFT) << 12;
tmp = 0;
for (i = (rdev->config.cayman.max_shader_engines - 1); i >= 0; i--) {
u32 rb_disable_bitmap;
WREG32(GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES | SE_INDEX(i));
WREG32(RLC_GFX_INDEX, INSTANCE_BROADCAST_WRITES | SE_INDEX(i));
rb_disable_bitmap = (RREG32(CC_RB_BACKEND_DISABLE) & 0x00ff0000) >> 16;
tmp <<= 4;
tmp |= rb_disable_bitmap;
}
/* enabled rb are just the one not disabled :) */
disabled_rb_mask = tmp;
WREG32(GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES | SE_BROADCAST_WRITES);
WREG32(RLC_GFX_INDEX, INSTANCE_BROADCAST_WRITES | SE_BROADCAST_WRITES);
WREG32(GB_ADDR_CONFIG, gb_addr_config);
WREG32(DMIF_ADDR_CONFIG, gb_addr_config);
WREG32(HDP_ADDR_CONFIG, gb_addr_config);
tmp = gb_addr_config & NUM_PIPES_MASK;
tmp = r6xx_remap_render_backend(rdev, tmp,
rdev->config.cayman.max_backends_per_se *
rdev->config.cayman.max_shader_engines,
CAYMAN_MAX_BACKENDS, disabled_rb_mask);
WREG32(GB_BACKEND_MAP, tmp);
cgts_tcc_disable = 0xffff0000;
for (i = 0; i < rdev->config.cayman.max_texture_channel_caches; i++)
cgts_tcc_disable &= ~(1 << (16 + i));
WREG32(CGTS_TCC_DISABLE, cgts_tcc_disable);
WREG32(CGTS_SYS_TCC_DISABLE, cgts_tcc_disable);
WREG32(CGTS_USER_SYS_TCC_DISABLE, cgts_tcc_disable);
WREG32(CGTS_USER_TCC_DISABLE, cgts_tcc_disable);
/* reprogram the shader complex */
cgts_sm_ctrl_reg = RREG32(CGTS_SM_CTRL_REG);
for (i = 0; i < 16; i++)
WREG32(CGTS_SM_CTRL_REG, OVERRIDE);
WREG32(CGTS_SM_CTRL_REG, cgts_sm_ctrl_reg);
/* set HW defaults for 3D engine */
WREG32(CP_MEQ_THRESHOLDS, MEQ1_START(0x30) | MEQ2_START(0x60));
sx_debug_1 = RREG32(SX_DEBUG_1);
sx_debug_1 |= ENABLE_NEW_SMX_ADDRESS;
WREG32(SX_DEBUG_1, sx_debug_1);
smx_dc_ctl0 = RREG32(SMX_DC_CTL0);
smx_dc_ctl0 &= ~NUMBER_OF_SETS(0x1ff);
smx_dc_ctl0 |= NUMBER_OF_SETS(rdev->config.cayman.sx_num_of_sets);
WREG32(SMX_DC_CTL0, smx_dc_ctl0);
WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(4) | CRC_SIMD_ID_WADDR_DISABLE);
/* need to be explicitly zero-ed */
WREG32(VGT_OFFCHIP_LDS_BASE, 0);
WREG32(SQ_LSTMP_RING_BASE, 0);
WREG32(SQ_HSTMP_RING_BASE, 0);
WREG32(SQ_ESTMP_RING_BASE, 0);
WREG32(SQ_GSTMP_RING_BASE, 0);
WREG32(SQ_VSTMP_RING_BASE, 0);
WREG32(SQ_PSTMP_RING_BASE, 0);
WREG32(TA_CNTL_AUX, DISABLE_CUBE_ANISO);
WREG32(SX_EXPORT_BUFFER_SIZES, (COLOR_BUFFER_SIZE((rdev->config.cayman.sx_max_export_size / 4) - 1) |
POSITION_BUFFER_SIZE((rdev->config.cayman.sx_max_export_pos_size / 4) - 1) |
SMX_BUFFER_SIZE((rdev->config.cayman.sx_max_export_smx_size / 4) - 1)));
WREG32(PA_SC_FIFO_SIZE, (SC_PRIM_FIFO_SIZE(rdev->config.cayman.sc_prim_fifo_size) |
SC_HIZ_TILE_FIFO_SIZE(rdev->config.cayman.sc_hiz_tile_fifo_size) |
SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.cayman.sc_earlyz_tile_fifo_size)));
WREG32(VGT_NUM_INSTANCES, 1);
WREG32(CP_PERFMON_CNTL, 0);
WREG32(SQ_MS_FIFO_SIZES, (CACHE_FIFO_SIZE(16 * rdev->config.cayman.sq_num_cf_insts) |
FETCH_FIFO_HIWATER(0x4) |
DONE_FIFO_HIWATER(0xe0) |
ALU_UPDATE_FIFO_HIWATER(0x8)));
WREG32(SQ_GPR_RESOURCE_MGMT_1, NUM_CLAUSE_TEMP_GPRS(4));
WREG32(SQ_CONFIG, (VC_ENABLE |
EXPORT_SRC_C |
GFX_PRIO(0) |
CS1_PRIO(0) |
CS2_PRIO(1)));
WREG32(SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, DYN_GPR_ENABLE);
WREG32(PA_SC_FORCE_EOV_MAX_CNTS, (FORCE_EOV_MAX_CLK_CNT(4095) |
FORCE_EOV_MAX_REZ_CNT(255)));
WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC) |
AUTO_INVLD_EN(ES_AND_GS_AUTO));
WREG32(VGT_GS_VERTEX_REUSE, 16);
WREG32(PA_SC_LINE_STIPPLE_STATE, 0);
WREG32(CB_PERF_CTR0_SEL_0, 0);
WREG32(CB_PERF_CTR0_SEL_1, 0);
WREG32(CB_PERF_CTR1_SEL_0, 0);
WREG32(CB_PERF_CTR1_SEL_1, 0);
WREG32(CB_PERF_CTR2_SEL_0, 0);
WREG32(CB_PERF_CTR2_SEL_1, 0);
WREG32(CB_PERF_CTR3_SEL_0, 0);
WREG32(CB_PERF_CTR3_SEL_1, 0);
tmp = RREG32(HDP_MISC_CNTL);
tmp |= HDP_FLUSH_INVALIDATE_CACHE;
WREG32(HDP_MISC_CNTL, tmp);
hdp_host_path_cntl = RREG32(HDP_HOST_PATH_CNTL);
WREG32(HDP_HOST_PATH_CNTL, hdp_host_path_cntl);
WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3));
udelay(50);
}
/**
* cik_sdma_gfx_resume - setup and start the async dma engines
*
* @adev: amdgpu_device pointer
*
* Set up the gfx DMA ring buffers and enable them (CIK).
* Returns 0 for success, error for failure.
*/
static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
{
struct amdgpu_ring *ring;
u32 rb_cntl, ib_cntl;
u32 rb_bufsz;
u32 wb_offset;
int i, j, r;
for (i = 0; i < adev->sdma.num_instances; i++) {
ring = &adev->sdma.instance[i].ring;
wb_offset = (ring->rptr_offs * 4);
mutex_lock(&adev->srbm_mutex);
for (j = 0; j < 16; j++) {
cik_srbm_select(adev, 0, 0, 0, j);
/* SDMA GFX */
WREG32(mmSDMA0_GFX_VIRTUAL_ADDR + sdma_offsets[i], 0);
WREG32(mmSDMA0_GFX_APE1_CNTL + sdma_offsets[i], 0);
/* XXX SDMA RLC - todo */
}
cik_srbm_select(adev, 0, 0, 0, 0);
mutex_unlock(&adev->srbm_mutex);
WREG32(mmSDMA0_SEM_INCOMPLETE_TIMER_CNTL + sdma_offsets[i], 0);
WREG32(mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL + sdma_offsets[i], 0);
/* Set ring buffer size in dwords */
rb_bufsz = order_base_2(ring->ring_size / 4);
rb_cntl = rb_bufsz << 1;
#ifdef __BIG_ENDIAN
rb_cntl |= SDMA0_GFX_RB_CNTL__RB_SWAP_ENABLE_MASK |
SDMA0_GFX_RB_CNTL__RPTR_WRITEBACK_SWAP_ENABLE_MASK;
#endif
WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
/* Initialize the ring buffer's read and write pointers */
WREG32(mmSDMA0_GFX_RB_RPTR + sdma_offsets[i], 0);
WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], 0);
/* set the wb address whether it's enabled or not */
WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i],
upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i],
((adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC));
rb_cntl |= SDMA0_GFX_RB_CNTL__RPTR_WRITEBACK_ENABLE_MASK;
WREG32(mmSDMA0_GFX_RB_BASE + sdma_offsets[i], ring->gpu_addr >> 8);
WREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i], ring->gpu_addr >> 40);
ring->wptr = 0;
WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
/* enable DMA RB */
WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i],
rb_cntl | SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK);
ib_cntl = SDMA0_GFX_IB_CNTL__IB_ENABLE_MASK;
#ifdef __BIG_ENDIAN
ib_cntl |= SDMA0_GFX_IB_CNTL__IB_SWAP_ENABLE_MASK;
#endif
/* enable DMA IBs */
WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
ring->ready = true;
r = amdgpu_ring_test_ring(ring);
if (r) {
ring->ready = false;
return r;
}
if (adev->mman.buffer_funcs_ring == ring)
amdgpu_ttm_set_active_vram_size(adev, adev->mc.real_vram_size);
}
return 0;
}
/**
* cik_sdma_gfx_resume - setup and start the async dma engines
*
* @rdev: radeon_device pointer
*
* Set up the gfx DMA ring buffers and enable them (CIK).
* Returns 0 for success, error for failure.
*/
static int cik_sdma_gfx_resume(struct radeon_device *rdev)
{
struct radeon_ring *ring;
u32 rb_cntl, ib_cntl;
u32 rb_bufsz;
u32 reg_offset, wb_offset;
int i, r;
for (i = 0; i < 2; i++) {
if (i == 0) {
ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
reg_offset = SDMA0_REGISTER_OFFSET;
wb_offset = R600_WB_DMA_RPTR_OFFSET;
} else {
ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX];
reg_offset = SDMA1_REGISTER_OFFSET;
wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET;
}
WREG32(SDMA0_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0);
WREG32(SDMA0_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0);
/* Set ring buffer size in dwords */
rb_bufsz = order_base_2(ring->ring_size / 4);
rb_cntl = rb_bufsz << 1;
#ifdef __BIG_ENDIAN
rb_cntl |= SDMA_RB_SWAP_ENABLE | SDMA_RPTR_WRITEBACK_SWAP_ENABLE;
#endif
WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl);
/* Initialize the ring buffer's read and write pointers */
WREG32(SDMA0_GFX_RB_RPTR + reg_offset, 0);
WREG32(SDMA0_GFX_RB_WPTR + reg_offset, 0);
/* set the wb address whether it's enabled or not */
WREG32(SDMA0_GFX_RB_RPTR_ADDR_HI + reg_offset,
upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
WREG32(SDMA0_GFX_RB_RPTR_ADDR_LO + reg_offset,
((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC));
if (rdev->wb.enabled)
rb_cntl |= SDMA_RPTR_WRITEBACK_ENABLE;
WREG32(SDMA0_GFX_RB_BASE + reg_offset, ring->gpu_addr >> 8);
WREG32(SDMA0_GFX_RB_BASE_HI + reg_offset, ring->gpu_addr >> 40);
ring->wptr = 0;
WREG32(SDMA0_GFX_RB_WPTR + reg_offset, ring->wptr << 2);
/* enable DMA RB */
WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl | SDMA_RB_ENABLE);
ib_cntl = SDMA_IB_ENABLE;
#ifdef __BIG_ENDIAN
ib_cntl |= SDMA_IB_SWAP_ENABLE;
#endif
/* enable DMA IBs */
WREG32(SDMA0_GFX_IB_CNTL + reg_offset, ib_cntl);
ring->ready = true;
r = radeon_ring_test(rdev, ring->idx, ring);
if (r) {
ring->ready = false;
return r;
}
}
if ((rdev->asic->copy.copy_ring_index == R600_RING_TYPE_DMA_INDEX) ||
(rdev->asic->copy.copy_ring_index == CAYMAN_RING_TYPE_DMA1_INDEX))
radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
return 0;
}
/**
* gmc_v7_0_mc_program - program the GPU memory controller
*
* @adev: amdgpu_device pointer
*
* Set the location of vram, gart, and AGP in the GPU's
* physical address space (CIK).
*/
static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
{
u32 tmp;
int i, j;
/* Initialize HDP */
for (i = 0, j = 0; i < 32; i++, j += 0x6) {
WREG32((0xb05 + j), 0x00000000);
WREG32((0xb06 + j), 0x00000000);
WREG32((0xb07 + j), 0x00000000);
WREG32((0xb08 + j), 0x00000000);
WREG32((0xb09 + j), 0x00000000);
}
WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
if (gmc_v7_0_wait_for_idle((void *)adev)) {
dev_warn(adev->dev, "Wait for MC idle timedout !\n");
}
if (adev->mode_info.num_crtc) {
/* Lockout access through VGA aperture*/
tmp = RREG32(mmVGA_HDP_CONTROL);
tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
WREG32(mmVGA_HDP_CONTROL, tmp);
/* disable VGA render */
tmp = RREG32(mmVGA_RENDER_CONTROL);
tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
WREG32(mmVGA_RENDER_CONTROL, tmp);
}
/* Update configuration */
WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
adev->mc.vram_start >> 12);
WREG32(mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
adev->mc.vram_end >> 12);
WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
adev->vram_scratch.gpu_addr >> 12);
WREG32(mmMC_VM_AGP_BASE, 0);
WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
if (gmc_v7_0_wait_for_idle((void *)adev)) {
dev_warn(adev->dev, "Wait for MC idle timedout !\n");
}
WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
tmp = RREG32(mmHDP_MISC_CNTL);
tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 0);
WREG32(mmHDP_MISC_CNTL, tmp);
tmp = RREG32(mmHDP_HOST_PATH_CNTL);
WREG32(mmHDP_HOST_PATH_CNTL, tmp);
}
