static int a3xx_snapshot_cp_roq(struct kgsl_device *device, void *snapshot,
int remain, void *priv)
{
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
struct kgsl_snapshot_debug *header = snapshot;
unsigned int *data = snapshot + sizeof(*header);
int i, size;
size = adreno_is_a330(adreno_dev) ?
A330_CP_ROQ_SIZE : A320_CP_ROQ_SIZE;
if (remain < DEBUG_SECTION_SZ(size)) {
SNAPSHOT_ERR_NOMEM(device, "CP ROQ DEBUG");
return 0;
}
header->type = SNAPSHOT_DEBUG_CP_ROQ;
header->size = size;
adreno_regwrite(device, A3XX_CP_ROQ_ADDR, 0x0);
for (i = 0; i < size; i++)
adreno_regread(device, A3XX_CP_ROQ_DATA, &data[i]);
return DEBUG_SECTION_SZ(size);
}
static void _snapshot_hlsq_regs(struct kgsl_snapshot_registers *regs,
struct kgsl_snapshot_registers_list *list,
struct adreno_device *adreno_dev)
{
struct kgsl_device *device = &adreno_dev->dev;
if (adreno_is_a330(adreno_dev)) {
unsigned int stall_context_full = 0;
_rbbm_debug_bus_read(device, RBBM_BLOCK_ID_HLSQ, 49,
&stall_context_full);
stall_context_full &= 0x08000000;
if (stall_context_full)
return;
} else {
unsigned int next_pif = 0;
_rbbm_debug_bus_read(device, RBBM_BLOCK_ID_HLSQ, 4, &next_pif);
next_pif &= 0x1f;
if (next_pif != 0 && next_pif != 1 && next_pif != 28)
return;
_rbbm_debug_bus_read(device, RBBM_BLOCK_ID_HLSQ, 7, &next_pif);
next_pif &= 0x3f;
if (next_pif != 0 && next_pif != 1 && next_pif != 10)
return;
}
regs[list->count].regs = (unsigned int *) a3xx_hlsq_registers;
regs[list->count].count = a3xx_hlsq_registers_count;
list->count++;
}
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));
}
void *a3xx_snapshot(struct adreno_device *adreno_dev, void *snapshot,
int *remain, int hang)
{
struct kgsl_device *device = &adreno_dev->dev;
struct kgsl_snapshot_registers_list list;
struct kgsl_snapshot_registers regs[5];
list.registers = regs;
list.count = 0;
adreno_regwrite(device, A3XX_RBBM_CLOCK_CTL, 0x00);
_snapshot_a3xx_regs(regs, &list);
_snapshot_hlsq_regs(regs, &list, adreno_dev);
if (adreno_is_a330(adreno_dev))
_snapshot_a330_regs(regs, &list);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_REGS, snapshot, remain,
kgsl_snapshot_dump_regs, &list);
snapshot = kgsl_snapshot_indexed_registers(device, snapshot,
remain, REG_CP_STATE_DEBUG_INDEX,
REG_CP_STATE_DEBUG_DATA, 0x0, 0x14);
snapshot = kgsl_snapshot_indexed_registers(device, snapshot,
remain, REG_CP_ME_CNTL, REG_CP_ME_STATUS,
64, 44);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_vpc_memory, NULL);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_meq, NULL);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_shader_memory, NULL);
if (hang) {
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_pfp_ram, NULL);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_pm4_ram, NULL);
}
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_roq, NULL);
if (adreno_is_a330(adreno_dev)) {
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a330_snapshot_cp_merciu, NULL);
}
snapshot = a3xx_snapshot_debugbus(device, snapshot, remain);
adreno_regwrite(device, A3XX_RBBM_CLOCK_CTL,
A3XX_RBBM_CLOCK_CTL_DEFAULT);
return snapshot;
}
void *a3xx_snapshot(struct adreno_device *adreno_dev, void *snapshot,
int *remain, int hang)
{
struct kgsl_device *device = &adreno_dev->dev;
struct kgsl_snapshot_registers_list list;
struct kgsl_snapshot_registers regs[2];
regs[0].regs = (unsigned int *) a3xx_registers;
regs[0].count = a3xx_registers_count;
list.registers = regs;
list.count = 1;
/* For A330, append the additional list of new registers to grab */
if (adreno_is_a330(adreno_dev)) {
regs[1].regs = (unsigned int *) a330_registers;
regs[1].count = a330_registers_count;
list.count++;
}
/* Master set of (non debug) registers */
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_REGS, snapshot, remain,
kgsl_snapshot_dump_regs, &list);
/* CP_STATE_DEBUG indexed registers */
snapshot = kgsl_snapshot_indexed_registers(device, snapshot,
remain, REG_CP_STATE_DEBUG_INDEX,
REG_CP_STATE_DEBUG_DATA, 0x0, 0x14);
/* CP_ME indexed registers */
snapshot = kgsl_snapshot_indexed_registers(device, snapshot,
remain, REG_CP_ME_CNTL, REG_CP_ME_STATUS,
64, 44);
/* Disable Clock gating temporarily for the debug bus to work */
adreno_regwrite(device, A3XX_RBBM_CLOCK_CTL, 0x00);
/* VPC memory */
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_vpc_memory, NULL);
/* CP MEQ */
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_meq, NULL);
/* Shader working/shadow memory */
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_shader_memory, NULL);
/* CP PFP and PM4 */
/* Reading these will hang the GPU if it isn't already hung */
if (hang) {
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_pfp_ram, NULL);
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_pm4_ram, NULL);
}
/* CP ROQ */
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a3xx_snapshot_cp_roq, NULL);
if (adreno_is_a330(adreno_dev)) {
snapshot = kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_DEBUG, snapshot, remain,
a330_snapshot_cp_merciu, NULL);
}
snapshot = a3xx_snapshot_debugbus(device, snapshot, remain);
/* Enable Clock gating */
adreno_regwrite(device, A3XX_RBBM_CLOCK_CTL,
A3XX_RBBM_CLOCK_CTL_DEFAULT);
return snapshot;
}
