drm/msm/adreno: Do CX GBIF config before GMU start

GMU lies on the CX domain and accesses CX GBIF. So do CX GBIF
configurations before GMU wakes up. This was not a problem so far, but
A840 GPU is very sensitive to this requirement. Also, move these
registers to the catalog.

Signed-off-by: Akhil P Oommen <akhilpo@oss.qualcomm.com>
Patchwork: https://patchwork.freedesktop.org/patch/689024/
Message-ID: <20251118-kaana-gpu-support-v4-17-86eeb8e93fb6@oss.qualcomm.com>
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
This commit is contained in:
Akhil P Oommen 2025-11-18 14:20:44 +05:30 committed by Rob Clark
parent 0700b9f682
commit 60a4e18e0e
5 changed files with 54 additions and 14 deletions

View File

@ -672,6 +672,14 @@ static const u32 a690_protect_regs[] = {
};
DECLARE_ADRENO_PROTECT(a690_protect, 48);
static const struct adreno_reglist a640_gbif[] = {
{ REG_A6XX_GBIF_QSB_SIDE0, 0x00071620 },
{ REG_A6XX_GBIF_QSB_SIDE1, 0x00071620 },
{ REG_A6XX_GBIF_QSB_SIDE2, 0x00071620 },
{ REG_A6XX_GBIF_QSB_SIDE3, 0x00071620 },
{ },
};
static const struct adreno_info a6xx_gpus[] = {
{
.chip_ids = ADRENO_CHIP_IDS(0x06010000),
@ -688,6 +696,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a612_hwcg,
.protect = &a630_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020202,
.prim_fifo_threshold = 0x00080000,
},
@ -894,6 +903,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a620_hwcg,
.protect = &a650_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020200,
.prim_fifo_threshold = 0x00010000,
},
@ -916,6 +926,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a690_hwcg,
.protect = &a650_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020200,
.prim_fifo_threshold = 0x00010000,
.bcms = (const struct a6xx_bcm[]) {
@ -998,6 +1009,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a650_hwcg,
.protect = &a650_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020202,
.prim_fifo_threshold = 0x00300200,
},
@ -1024,6 +1036,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a660_hwcg,
.protect = &a660_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020000,
.prim_fifo_threshold = 0x00300200,
},
@ -1042,6 +1055,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a690_hwcg,
.protect = &a660_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020200,
.prim_fifo_threshold = 0x00300200,
},
@ -1066,6 +1080,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a660_hwcg,
.protect = &a660_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020202,
.prim_fifo_threshold = 0x00200200,
},
@ -1112,6 +1127,7 @@ static const struct adreno_info a6xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a690_hwcg,
.protect = &a690_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020200,
.prim_fifo_threshold = 0x00800200,
},
@ -1447,6 +1463,7 @@ static const struct adreno_info a7xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.hwcg = a702_hwcg,
.protect = &a650_protect,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020202,
.prim_fifo_threshold = 0x0000c000,
},
@ -1474,6 +1491,7 @@ static const struct adreno_info a7xx_gpus[] = {
.hwcg = a730_hwcg,
.protect = &a730_protect,
.pwrup_reglist = &a7xx_pwrup_reglist,
.gbif_cx = a640_gbif,
.gmu_cgc_mode = 0x00020000,
},
.preempt_record_size = 2860 * SZ_1K,
@ -1495,6 +1513,7 @@ static const struct adreno_info a7xx_gpus[] = {
.hwcg = a740_hwcg,
.protect = &a730_protect,
.pwrup_reglist = &a7xx_pwrup_reglist,
.gbif_cx = a640_gbif,
.gmu_chipid = 0x7020100,
.gmu_cgc_mode = 0x00020202,
.bcms = (const struct a6xx_bcm[]) {
@ -1529,6 +1548,7 @@ static const struct adreno_info a7xx_gpus[] = {
.protect = &a730_protect,
.pwrup_reglist = &a7xx_pwrup_reglist,
.ifpc_reglist = &a750_ifpc_reglist,
.gbif_cx = a640_gbif,
.gmu_chipid = 0x7050001,
.gmu_cgc_mode = 0x00020202,
.bcms = (const struct a6xx_bcm[]) {
@ -1570,6 +1590,7 @@ static const struct adreno_info a7xx_gpus[] = {
.protect = &a730_protect,
.pwrup_reglist = &a7xx_pwrup_reglist,
.ifpc_reglist = &a750_ifpc_reglist,
.gbif_cx = a640_gbif,
.gmu_chipid = 0x7090100,
.gmu_cgc_mode = 0x00020202,
.bcms = (const struct a6xx_bcm[]) {
@ -1602,6 +1623,7 @@ static const struct adreno_info a7xx_gpus[] = {
.hwcg = a740_hwcg,
.protect = &a730_protect,
.pwrup_reglist = &a7xx_pwrup_reglist,
.gbif_cx = a640_gbif,
.gmu_chipid = 0x70f0000,
.gmu_cgc_mode = 0x00020222,
.bcms = (const struct a6xx_bcm[]) {
@ -1749,6 +1771,15 @@ static const u32 a840_protect_regs[] = {
};
DECLARE_ADRENO_PROTECT(a840_protect, 15);
static const struct adreno_reglist a840_gbif[] = {
{ REG_A6XX_GBIF_QSB_SIDE0, 0x00071e20 },
{ REG_A6XX_GBIF_QSB_SIDE1, 0x00071e20 },
{ REG_A6XX_GBIF_QSB_SIDE2, 0x00071e20 },
{ REG_A6XX_GBIF_QSB_SIDE3, 0x00071e20 },
{ REG_A8XX_GBIF_CX_CONFIG, 0x20023000 },
{ },
};
static const struct adreno_info a8xx_gpus[] = {
{
.chip_ids = ADRENO_CHIP_IDS(0x44050a01),
@ -1766,6 +1797,7 @@ static const struct adreno_info a8xx_gpus[] = {
.a6xx = &(const struct a6xx_info) {
.protect = &a840_protect,
.nonctxt_reglist = a840_nonctxt_regs,
.gbif_cx = a840_gbif,
.max_slices = 3,
.gmu_chipid = 0x8020100,
.bcms = (const struct a6xx_bcm[]) {

View File

@ -894,7 +894,9 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
{
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;
const struct a6xx_info *a6xx_info = adreno_gpu->info->a6xx;
const struct adreno_reglist *gbif_cx = a6xx_info->gbif_cx;
u32 fence_range_lower, fence_range_upper;
u32 chipid = 0;
int ret;
@ -990,6 +992,15 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
gmu->log.iova | (gmu->log.size / SZ_4K - 1));
}
/* For A7x and newer, do the CX GBIF configurations before GMU wake up */
for (int i = 0; (gbif_cx && gbif_cx[i].offset); i++)
gpu_write(gpu, gbif_cx[i].offset, gbif_cx[i].value);
if (adreno_is_a8xx(adreno_gpu)) {
gpu_write(gpu, REG_A8XX_GBIF_CX_CONFIG, 0x20023000);
gmu_write(gmu, REG_A6XX_GMU_MRC_GBIF_QOS_CTRL, 0x33);
}
/* Set up the lowest idle level on the GMU */
a6xx_gmu_power_config(gmu);

View File

@ -1279,17 +1279,20 @@ static int hw_init(struct msm_gpu *gpu)
/* enable hardware clockgating */
a6xx_set_hwcg(gpu, true);
/* VBIF/GBIF start*/
if (adreno_is_a610_family(adreno_gpu) ||
adreno_is_a640_family(adreno_gpu) ||
adreno_is_a650_family(adreno_gpu) ||
adreno_is_a7xx(adreno_gpu)) {
/* For gmuwrapper implementations, do the VBIF/GBIF CX configuration here */
if (adreno_is_a610_family(adreno_gpu)) {
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE0, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE1, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE2, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE3, 0x00071620);
gpu_write(gpu, REG_A6XX_RBBM_GBIF_CLIENT_QOS_CNTL,
adreno_is_a7xx(adreno_gpu) ? 0x2120212 : 0x3);
}
if (adreno_is_a610_family(adreno_gpu) ||
adreno_is_a640_family(adreno_gpu) ||
adreno_is_a650_family(adreno_gpu)) {
gpu_write(gpu, REG_A6XX_RBBM_GBIF_CLIENT_QOS_CNTL, 0x3);
} else if (adreno_is_a7xx(adreno_gpu)) {
gpu_write(gpu, REG_A6XX_RBBM_GBIF_CLIENT_QOS_CNTL, 0x2120212);
} else {
gpu_write(gpu, REG_A6XX_RBBM_VBIF_CLIENT_QOS_CNTL, 0x3);
}

View File

@ -46,6 +46,7 @@ struct a6xx_info {
const struct adreno_protect *protect;
const struct adreno_reglist_list *pwrup_reglist;
const struct adreno_reglist_list *ifpc_reglist;
const struct adreno_reglist *gbif_cx;
const struct adreno_reglist_pipe *nonctxt_reglist;
u32 max_slices;
u32 gmu_chipid;

View File

@ -519,13 +519,6 @@ static int hw_init(struct msm_gpu *gpu)
gpu_write64(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE, 0x00000000);
gpu_write(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_SIZE, 0x00000000);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE0, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE1, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE2, 0x00071620);
gpu_write(gpu, REG_A6XX_GBIF_QSB_SIDE3, 0x00071620);
gpu_write(gpu, REG_A8XX_GBIF_CX_CONFIG, 0x20023000);
gmu_write(gmu, REG_A6XX_GMU_MRC_GBIF_QOS_CTRL, 0x33);
/* Make all blocks contribute to the GPU BUSY perf counter */
gpu_write(gpu, REG_A8XX_RBBM_PERFCTR_GPU_BUSY_MASKED, 0xffffffff);