/* Gracefully try to shut down the GMU and by extension the GPU */
static void a6xx_gmu_shutdown(struct a6xx_gmu *gmu)
{
struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
struct msm_gpu *gpu = &adreno_gpu->base;
u32 val;
/*
* The GMU may still be in slumber unless the GPU started so check and
* skip putting it back into slumber if so
*/
val = gmu_read(gmu, REG_A6XX_GPU_GMU_CX_GMU_RPMH_POWER_STATE);
if (val != 0xf) {
int ret = a6xx_gmu_wait_for_idle(gmu);
/* If the GMU isn't responding assume it is hung */
if (ret) {
a6xx_gmu_force_off(gmu);
return;
}
/* Clear the VBIF pipe before shutting down */
gpu_write(gpu, REG_A6XX_VBIF_XIN_HALT_CTRL0, 0xf);
spin_until((gpu_read(gpu, REG_A6XX_VBIF_XIN_HALT_CTRL1) & 0xf)
== 0xf);
gpu_write(gpu, REG_A6XX_VBIF_XIN_HALT_CTRL0, 0);
/* tell the GMU we want to slumber */
a6xx_gmu_notify_slumber(gmu);
ret = gmu_poll_timeout(gmu,
REG_A6XX_GPU_GMU_AO_GPU_CX_BUSY_STATUS, val,
!(val & A6XX_GPU_GMU_AO_GPU_CX_BUSY_STATUS_GPUBUSYIGNAHB),
100, 10000);
/*
* Let the user know we failed to slumber but don't worry too
* much because we are powering down anyway
*/
if (ret)
DRM_DEV_ERROR(gmu->dev,
"Unable to slumber GMU: status = 0%x/0%x\n",
gmu_read(gmu,
REG_A6XX_GPU_GMU_AO_GPU_CX_BUSY_STATUS),
gmu_read(gmu,
REG_A6XX_GPU_GMU_AO_GPU_CX_BUSY_STATUS2));
}
/* Turn off HFI */
a6xx_hfi_stop(gmu);
/* Stop the interrupts and mask the hardware */
a6xx_gmu_irq_disable(gmu);
/* Tell RPMh to power off the GPU */
a6xx_rpmh_stop(gmu);
}
static void a4xx_recover(struct msm_gpu *gpu)
{
adreno_dump_info(gpu);
/* dump registers before resetting gpu, if enabled: */
if (hang_debug)
a4xx_dump(gpu);
gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 1);
gpu_read(gpu, REG_A4XX_RBBM_SW_RESET_CMD);
gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 0);
adreno_recover(gpu);
}
void memorymap_registers_write(uint32 Addr, uint8 Value)
{
#ifdef DEBUG
printf("Addr 0x%.2x, Value 0x%.2x\n", Addr, Value);
#endif
memorymap_regs[Addr&0x1fff] = Value;
switch (Addr&0x1fff)
{
case 0x00:
case 0x01:
case 0x02:
case 0x03: gpu_write(Addr, Value);
break;
case 0x22: io_write(Addr, Value);
break;
case 0x23:
case 0x24: timer_write(Addr, Value);
break;
case 0x26:
//fprintf(log_get(), "memorymap: writing 0x%.2x to rom bank register\n", Value);
memorymap_lowerRomBank = memorymap_programRom + ((((uint32)Value) & 0x60) << 9);
memorymap_upperRomBank = memorymap_programRom + (memorymap_programRomSize==0x10000?0xc000:0x4000);
return;
case 0x27: //fprintf(log_get(), "regs: writing 0x%.2x from 0x%.4x\n", Value, Addr);
break;
case 0x10:
case 0x11:
case 0x12:
case 0x13: // ALEK
case 0x14:
case 0x15:
case 0x16:
case 0x17: // ALEK
#ifndef NOAUDIO
soundport_w(((Addr&0x4)>>2), Addr&3, Value); break;
#endif
//sound_write(Addr&7, Value); break;
case 0x28:
case 0x29:
case 0x2a:
#ifndef NOAUDIO
svision_noise_w(Addr&0x07, Value); break;
#endif
//sound_noise_write(Addr&0x07, Value); break;
case 0x18:
case 0x19:
case 0x1a:
case 0x1b:
case 0x1c:
#ifdef GP2X
if(currentConfig.enable_sound) sound_audio_dma(Addr&0x07, Value); break;
#else
svision_sounddma_w(Addr&0x07, Value); break;
#endif
break;
}
}
static void a4xx_recover(struct msm_gpu *gpu)
{
int i;
adreno_dump_info(gpu);
for (i = 0; i < 8; i++) {
printk("CP_SCRATCH_REG%d: %u\n", i,
gpu_read(gpu, REG_AXXX_CP_SCRATCH_REG0 + i));
}
/* dump registers before resetting gpu, if enabled: */
if (hang_debug)
a4xx_dump(gpu);
gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 1);
gpu_read(gpu, REG_A4XX_RBBM_SW_RESET_CMD);
gpu_write(gpu, REG_A4XX_RBBM_SW_RESET_CMD, 0);
adreno_recover(gpu);
}
/*
* a4xx_enable_hwcg() - Program the clock control registers
* @device: The adreno device pointer
*/
static void a4xx_enable_hwcg(struct msm_gpu *gpu)
{
struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
unsigned int i;
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TP(i), 0x02222202);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_TP(i), 0x00002222);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TP(i), 0x0E739CE7);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TP(i), 0x00111111);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_SP(i), 0x22222222);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_SP(i), 0x00222222);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_SP(i), 0x00000104);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_SP(i), 0x00000081);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_UCHE, 0x22222222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_UCHE, 0x02222222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL3_UCHE, 0x00000000);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL4_UCHE, 0x00000000);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_UCHE, 0x00004444);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_UCHE, 0x00001112);
for (i = 0; i < 4; i++)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_RB(i), 0x22222222);
/* Disable L1 clocking in A420 due to CCU issues with it */
for (i = 0; i < 4; i++) {
if (adreno_is_a420(adreno_gpu)) {
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i),
0x00002020);
} else {
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2_RB(i),
0x00022020);
}
}
for (i = 0; i < 4; i++) {
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_MARB_CCU(i),
0x00000922);
}
for (i = 0; i < 4; i++) {
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_RB_MARB_CCU(i),
0x00000000);
}
for (i = 0; i < 4; i++) {
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_RB_MARB_CCU_L1(i),
0x00000001);
}
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_MODE_GPC, 0x02222222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_GPC, 0x04100104);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_GPC, 0x00022222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_COM_DCOM, 0x00000022);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_COM_DCOM, 0x0000010F);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_COM_DCOM, 0x00000022);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_TSE_RAS_RBBM, 0x00222222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_TSE_RAS_RBBM, 0x00004104);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_TSE_RAS_RBBM, 0x00000222);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL_HLSQ , 0x00000000);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_HYST_HLSQ, 0x00000000);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, 0x00220000);
/* Early A430's have a timing issue with SP/TP power collapse;
disabling HW clock gating prevents it. */
if (adreno_is_a430(adreno_gpu) && adreno_gpu->rev.patchid < 2)
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL, 0);
else
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL, 0xAAAAAAAA);
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_CTL2, 0);
}
static int a4xx_hw_init(struct msm_gpu *gpu)
{
struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
struct a4xx_gpu *a4xx_gpu = to_a4xx_gpu(adreno_gpu);
uint32_t *ptr, len;
int i, ret;
if (adreno_is_a420(adreno_gpu)) {
gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT, 0x0001001F);
gpu_write(gpu, REG_A4XX_VBIF_ABIT_SORT_CONF, 0x000000A4);
gpu_write(gpu, REG_A4XX_VBIF_GATE_OFF_WRREQ_EN, 0x00000001);
gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF1, 0x00000018);
gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF1, 0x00000018);
gpu_write(gpu, REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x00000003);
} else if (adreno_is_a430(adreno_gpu)) {
gpu_write(gpu, REG_A4XX_VBIF_GATE_OFF_WRREQ_EN, 0x00000001);
gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A4XX_VBIF_IN_RD_LIM_CONF1, 0x00000018);
gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A4XX_VBIF_IN_WR_LIM_CONF1, 0x00000018);
gpu_write(gpu, REG_A4XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x00000003);
} else {
BUG();
}
/* Make all blocks contribute to the GPU BUSY perf counter */
gpu_write(gpu, REG_A4XX_RBBM_GPU_BUSY_MASKED, 0xffffffff);
/* Tune the hystersis counters for SP and CP idle detection */
gpu_write(gpu, REG_A4XX_RBBM_SP_HYST_CNT, 0x10);
gpu_write(gpu, REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL, 0x10);
if (adreno_is_a430(adreno_gpu)) {
gpu_write(gpu, REG_A4XX_RBBM_WAIT_IDLE_CLOCKS_CTL2, 0x30);
}
/* Enable the RBBM error reporting bits */
gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL0, 0x00000001);
/* Enable AHB error reporting*/
gpu_write(gpu, REG_A4XX_RBBM_AHB_CTL1, 0xa6ffffff);
/* Enable power counters*/
gpu_write(gpu, REG_A4XX_RBBM_RBBM_CTL, 0x00000030);
/*
* Turn on hang detection - this spews a lot of useful information
* into the RBBM registers on a hang:
*/
gpu_write(gpu, REG_A4XX_RBBM_INTERFACE_HANG_INT_CTL,
(1 << 30) | 0xFFFF);
gpu_write(gpu, REG_A4XX_RB_GMEM_BASE_ADDR,
(unsigned int)(a4xx_gpu->ocmem_base >> 14));
/* Turn on performance counters: */
gpu_write(gpu, REG_A4XX_RBBM_PERFCTR_CTL, 0x01);
/* use the first CP counter for timestamp queries.. userspace may set
* this as well but it selects the same counter/countable:
*/
gpu_write(gpu, REG_A4XX_CP_PERFCTR_CP_SEL_0, CP_ALWAYS_COUNT);
if (adreno_is_a430(adreno_gpu))
gpu_write(gpu, REG_A4XX_UCHE_CACHE_WAYS_VFD, 0x07);
/* Disable L2 bypass to avoid UCHE out of bounds errors */
gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_LO, 0xffff0000);
gpu_write(gpu, REG_A4XX_UCHE_TRAP_BASE_HI, 0xffff0000);
gpu_write(gpu, REG_A4XX_CP_DEBUG, (1 << 25) |
(adreno_is_a420(adreno_gpu) ? (1 << 29) : 0));
/* On A430 enable SP regfile sleep for power savings */
/* TODO downstream does this for !420, so maybe applies for 405 too? */
if (!adreno_is_a420(adreno_gpu)) {
gpu_write(gpu, REG_A4XX_RBBM_SP_REGFILE_SLEEP_CNTL_0,
0x00000441);
gpu_write(gpu, REG_A4XX_RBBM_SP_REGFILE_SLEEP_CNTL_1,
0x00000441);
}
a4xx_enable_hwcg(gpu);
/*
* For A420 set RBBM_CLOCK_DELAY_HLSQ.CGC_HLSQ_TP_EARLY_CYC >= 2
* due to timing issue with HLSQ_TP_CLK_EN
*/
if (adreno_is_a420(adreno_gpu)) {
unsigned int val;
val = gpu_read(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ);
val &= ~A4XX_CGC_HLSQ_EARLY_CYC__MASK;
val |= 2 << A4XX_CGC_HLSQ_EARLY_CYC__SHIFT;
gpu_write(gpu, REG_A4XX_RBBM_CLOCK_DELAY_HLSQ, val);
}
/* setup access protection: */
gpu_write(gpu, REG_A4XX_CP_PROTECT_CTRL, 0x00000007);
/* RBBM registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(0), 0x62000010);
gpu_write(gpu, REG_A4XX_CP_PROTECT(1), 0x63000020);
gpu_write(gpu, REG_A4XX_CP_PROTECT(2), 0x64000040);
gpu_write(gpu, REG_A4XX_CP_PROTECT(3), 0x65000080);
gpu_write(gpu, REG_A4XX_CP_PROTECT(4), 0x66000100);
gpu_write(gpu, REG_A4XX_CP_PROTECT(5), 0x64000200);
/* CP registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(6), 0x67000800);
gpu_write(gpu, REG_A4XX_CP_PROTECT(7), 0x64001600);
/* RB registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(8), 0x60003300);
/* HLSQ registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(9), 0x60003800);
/* VPC registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(10), 0x61003980);
/* SMMU registers */
gpu_write(gpu, REG_A4XX_CP_PROTECT(11), 0x6e010000);
gpu_write(gpu, REG_A4XX_RBBM_INT_0_MASK, A4XX_INT0_MASK);
ret = adreno_hw_init(gpu);
if (ret)
return ret;
/* Load PM4: */
ptr = (uint32_t *)(adreno_gpu->pm4->data);
len = adreno_gpu->pm4->size / 4;
DBG("loading PM4 ucode version: %u", ptr[0]);
gpu_write(gpu, REG_A4XX_CP_ME_RAM_WADDR, 0);
for (i = 1; i < len; i++)
gpu_write(gpu, REG_A4XX_CP_ME_RAM_DATA, ptr[i]);
/* Load PFP: */
ptr = (uint32_t *)(adreno_gpu->pfp->data);
len = adreno_gpu->pfp->size / 4;
DBG("loading PFP ucode version: %u", ptr[0]);
gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_ADDR, 0);
for (i = 1; i < len; i++)
gpu_write(gpu, REG_A4XX_CP_PFP_UCODE_DATA, ptr[i]);
/* clear ME_HALT to start micro engine */
gpu_write(gpu, REG_A4XX_CP_ME_CNTL, 0);
return a4xx_me_init(gpu) ? 0 : -EINVAL;
}
static int a3xx_hw_init(struct msm_gpu *gpu)
{
struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
struct a3xx_gpu *a3xx_gpu = to_a3xx_gpu(adreno_gpu);
uint32_t *ptr, len;
int i, ret;
DBG("%s", gpu->name);
if (adreno_is_a305(adreno_gpu)) {
/* Set up 16 deep read/write request queues: */
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x10101010);
/* Enable WR-REQ: */
gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x0000ff);
/* Set up round robin arbitration between both AXI ports: */
gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
/* Set up AOOO: */
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003c);
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003c003c);
} else if (adreno_is_a320(adreno_gpu)) {
/* Set up 16 deep read/write request queues: */
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x10101010);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x10101010);
/* Enable WR-REQ: */
gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x0000ff);
/* Set up round robin arbitration between both AXI ports: */
gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
/* Set up AOOO: */
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003c);
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003c003c);
/* Enable 1K sort: */
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x000000ff);
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);
} else if (adreno_is_a330v2(adreno_gpu)) {
/*
* Most of the VBIF registers on 8974v2 have the correct
* values at power on, so we won't modify those if we don't
* need to
*/
/* Enable 1k sort: */
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x0001003f);
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);
/* Enable WR-REQ: */
gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x00003f);
gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
/* Set up VBIF_ROUND_ROBIN_QOS_ARB: */
gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0003);
} else if (adreno_is_a330(adreno_gpu)) {
/* Set up 16 deep read/write request queues: */
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x18181818);
gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x18181818);
gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x18181818);
/* Enable WR-REQ: */
gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x00003f);
/* Set up round robin arbitration between both AXI ports: */
gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
/* Set up VBIF_ROUND_ROBIN_QOS_ARB: */
gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0001);
/* Set up AOOO: */
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003f);
gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003f003f);
/* Enable 1K sort: */
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x0001003f);
gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);
/* Disable VBIF clock gating. This is to enable AXI running
* higher frequency than GPU:
*/
gpu_write(gpu, REG_A3XX_VBIF_CLKON, 0x00000001);
} else {
BUG();
}
/* Make all blocks contribute to the GPU BUSY perf counter: */
gpu_write(gpu, REG_A3XX_RBBM_GPU_BUSY_MASKED, 0xffffffff);
/* Tune the hystersis counters for SP and CP idle detection: */
gpu_write(gpu, REG_A3XX_RBBM_SP_HYST_CNT, 0x10);
gpu_write(gpu, REG_A3XX_RBBM_WAIT_IDLE_CLOCKS_CTL, 0x10);
/* Enable the RBBM error reporting bits. This lets us get
* useful information on failure:
*/
gpu_write(gpu, REG_A3XX_RBBM_AHB_CTL0, 0x00000001);
/* Enable AHB error reporting: */
gpu_write(gpu, REG_A3XX_RBBM_AHB_CTL1, 0xa6ffffff);
/* Turn on the power counters: */
gpu_write(gpu, REG_A3XX_RBBM_RBBM_CTL, 0x00030000);
/* Turn on hang detection - this spews a lot of useful information
* into the RBBM registers on a hang:
*/
gpu_write(gpu, REG_A3XX_RBBM_INTERFACE_HANG_INT_CTL, 0x00010fff);
/* Enable 64-byte cacheline size. HW Default is 32-byte (0x000000E0): */
gpu_write(gpu, REG_A3XX_UCHE_CACHE_MODE_CONTROL_REG, 0x00000001);
/* Enable Clock gating: */
if (adreno_is_a320(adreno_gpu))
gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xbfffffff);
else if (adreno_is_a330v2(adreno_gpu))
gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xaaaaaaaa);
else if (adreno_is_a330(adreno_gpu))
gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xbffcffff);
if (adreno_is_a330v2(adreno_gpu))
gpu_write(gpu, REG_A3XX_RBBM_GPR0_CTL, 0x05515455);
else if (adreno_is_a330(adreno_gpu))
gpu_write(gpu, REG_A3XX_RBBM_GPR0_CTL, 0x00000000);
/* Set the OCMEM base address for A330, etc */
if (a3xx_gpu->ocmem_hdl) {
gpu_write(gpu, REG_A3XX_RB_GMEM_BASE_ADDR,
(unsigned int)(a3xx_gpu->ocmem_base >> 14));
}
static void etnaviv_gpu_hw_init(struct etnaviv_gpu *gpu)
{
u16 prefetch;
if ((etnaviv_is_model_rev(gpu, GC320, 0x5007) ||
etnaviv_is_model_rev(gpu, GC320, 0x5220)) &&
gpu_read(gpu, VIVS_HI_CHIP_TIME) != 0x2062400) {
u32 mc_memory_debug;
mc_memory_debug = gpu_read(gpu, VIVS_MC_DEBUG_MEMORY) & ~0xff;
if (gpu->identity.revision == 0x5007)
mc_memory_debug |= 0x0c;
else
mc_memory_debug |= 0x08;
gpu_write(gpu, VIVS_MC_DEBUG_MEMORY, mc_memory_debug);
}
/*
* Update GPU AXI cache atttribute to "cacheable, no allocate".
* This is necessary to prevent the iMX6 SoC locking up.
*/
gpu_write(gpu, VIVS_HI_AXI_CONFIG,
VIVS_HI_AXI_CONFIG_AWCACHE(2) |
VIVS_HI_AXI_CONFIG_ARCACHE(2));
/* GC2000 rev 5108 needs a special bus config */
if (etnaviv_is_model_rev(gpu, GC2000, 0x5108)) {
u32 bus_config = gpu_read(gpu, VIVS_MC_BUS_CONFIG);
bus_config &= ~(VIVS_MC_BUS_CONFIG_FE_BUS_CONFIG__MASK |
VIVS_MC_BUS_CONFIG_TX_BUS_CONFIG__MASK);
bus_config |= VIVS_MC_BUS_CONFIG_FE_BUS_CONFIG(1) |
VIVS_MC_BUS_CONFIG_TX_BUS_CONFIG(0);
gpu_write(gpu, VIVS_MC_BUS_CONFIG, bus_config);
}
/* set base addresses */
gpu_write(gpu, VIVS_MC_MEMORY_BASE_ADDR_RA, gpu->memory_base);
gpu_write(gpu, VIVS_MC_MEMORY_BASE_ADDR_FE, gpu->memory_base);
gpu_write(gpu, VIVS_MC_MEMORY_BASE_ADDR_TX, gpu->memory_base);
gpu_write(gpu, VIVS_MC_MEMORY_BASE_ADDR_PEZ, gpu->memory_base);
gpu_write(gpu, VIVS_MC_MEMORY_BASE_ADDR_PE, gpu->memory_base);
/* setup the MMU page table pointers */
etnaviv_iommu_domain_restore(gpu, gpu->mmu->domain);
/* Start command processor */
prefetch = etnaviv_buffer_init(gpu);
gpu_write(gpu, VIVS_HI_INTR_ENBL, ~0U);
gpu_write(gpu, VIVS_FE_COMMAND_ADDRESS,
gpu->buffer->paddr - gpu->memory_base);
gpu_write(gpu, VIVS_FE_COMMAND_CONTROL,
VIVS_FE_COMMAND_CONTROL_ENABLE |
VIVS_FE_COMMAND_CONTROL_PREFETCH(prefetch));
}
static int etnaviv_hw_reset(struct etnaviv_gpu *gpu)
{
u32 control, idle;
unsigned long timeout;
bool failed = true;
/* TODO
*
* - clock gating
* - puls eater
* - what about VG?
*/
/* We hope that the GPU resets in under one second */
timeout = jiffies + msecs_to_jiffies(1000);
while (time_is_after_jiffies(timeout)) {
control = VIVS_HI_CLOCK_CONTROL_DISABLE_DEBUG_REGISTERS |
VIVS_HI_CLOCK_CONTROL_FSCALE_VAL(0x40);
/* enable clock */
etnaviv_gpu_load_clock(gpu, control);
/* Wait for stable clock. Vivante's code waited for 1ms */
usleep_range(1000, 10000);
/* isolate the GPU. */
control |= VIVS_HI_CLOCK_CONTROL_ISOLATE_GPU;
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, control);
/* set soft reset. */
control |= VIVS_HI_CLOCK_CONTROL_SOFT_RESET;
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, control);
/* wait for reset. */
msleep(1);
/* reset soft reset bit. */
control &= ~VIVS_HI_CLOCK_CONTROL_SOFT_RESET;
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, control);
/* reset GPU isolation. */
control &= ~VIVS_HI_CLOCK_CONTROL_ISOLATE_GPU;
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, control);
/* read idle register. */
idle = gpu_read(gpu, VIVS_HI_IDLE_STATE);
/* try reseting again if FE it not idle */
if ((idle & VIVS_HI_IDLE_STATE_FE) == 0) {
dev_dbg(gpu->dev, "FE is not idle\n");
continue;
}
/* read reset register. */
control = gpu_read(gpu, VIVS_HI_CLOCK_CONTROL);
/* is the GPU idle? */
if (((control & VIVS_HI_CLOCK_CONTROL_IDLE_3D) == 0) ||
((control & VIVS_HI_CLOCK_CONTROL_IDLE_2D) == 0)) {
dev_dbg(gpu->dev, "GPU is not idle\n");
continue;
}
failed = false;
break;
}
if (failed) {
idle = gpu_read(gpu, VIVS_HI_IDLE_STATE);
control = gpu_read(gpu, VIVS_HI_CLOCK_CONTROL);
dev_err(gpu->dev, "GPU failed to reset: FE %sidle, 3D %sidle, 2D %sidle\n",
idle & VIVS_HI_IDLE_STATE_FE ? "" : "not ",
control & VIVS_HI_CLOCK_CONTROL_IDLE_3D ? "" : "not ",
control & VIVS_HI_CLOCK_CONTROL_IDLE_2D ? "" : "not ");
return -EBUSY;
}
/* We rely on the GPU running, so program the clock */
control = VIVS_HI_CLOCK_CONTROL_DISABLE_DEBUG_REGISTERS |
VIVS_HI_CLOCK_CONTROL_FSCALE_VAL(0x40);
/* enable clock */
etnaviv_gpu_load_clock(gpu, control);
return 0;
}
static void etnaviv_gpu_load_clock(struct etnaviv_gpu *gpu, u32 clock)
{
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, clock |
VIVS_HI_CLOCK_CONTROL_FSCALE_CMD_LOAD);
gpu_write(gpu, VIVS_HI_CLOCK_CONTROL, clock);
}
