The RAS bad page retire flip bits can be set per vram type,
vram vendor and nps mode.
Signed-off-by: Tao Zhou <tao.zh...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 1 -
drivers/gpu/drm/amd/amdgpu/umc_v12_0.c | 82 +++++++++++++++++---------
drivers/gpu/drm/amd/amdgpu/umc_v12_0.h | 2 -
3 files changed, 53 insertions(+), 32 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 3caebe7c25a5..464015fc2012 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -1504,7 +1504,6 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device
*adev)
adev->umc.umc_inst_num = UMC_V12_0_UMC_INSTANCE_NUM;
adev->umc.node_inst_num /= UMC_V12_0_UMC_INSTANCE_NUM;
adev->umc.channel_offs = UMC_V12_0_PER_CHANNEL_OFFSET;
- adev->umc.retire_unit = UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL;
if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu)
adev->umc.ras = &umc_v12_0_ras;
break;
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
index d715cfde6aec..774d3baa62d7 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.c
@@ -174,6 +174,49 @@ static void umc_v12_0_query_ras_error_count(struct
amdgpu_device *adev,
umc_v12_0_reset_error_count(adev);
}
+static void umc_v12_0_get_retire_flip_bits(struct amdgpu_device *adev)
+{
+ enum amdgpu_memory_partition nps = AMDGPU_NPS1_PARTITION_MODE;
+ uint32_t vram_type = adev->gmc.vram_type;
+ struct amdgpu_umc_flip_bits *flip_bits = &(adev->umc.flip_bits);
+
+ if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+ nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+
+ /* default setting */
+ flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_C2_BIT;
+ flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_C3_BIT;
+ flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_C4_BIT;
+ flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R13_BIT;
+ flip_bits->bit_num = 4;
+
+ switch (vram_type) {
+ case AMDGPU_VRAM_TYPE_HBM:
+ /* other nps modes are taken as nps1 */
+ if (nps == AMDGPU_NPS2_PARTITION_MODE) {
+ flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_CH5_BIT;
+ flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_C2_BIT;
+ flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_B1_BIT;
+ flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R12_BIT;
+ }
+
+ if (nps == AMDGPU_NPS4_PARTITION_MODE) {
+ flip_bits->flip_bits_in_pa[0] = UMC_V12_0_PA_CH4_BIT;
+ flip_bits->flip_bits_in_pa[1] = UMC_V12_0_PA_CH5_BIT;
+ flip_bits->flip_bits_in_pa[2] = UMC_V12_0_PA_B0_BIT;
+ flip_bits->flip_bits_in_pa[3] = UMC_V12_0_PA_R11_BIT;
+ }
+
+ break;
+ default:
+ dev_warn(adev->dev,
+ "Unknown HBM type, set RAS retire flip bits to the
value in NPS1 mode.\n");
+ break;
+ }
+
+ adev->umc.retire_unit = 0x1 << flip_bits->bit_num;
+}
+
static int umc_v12_0_convert_error_address(struct amdgpu_device *adev,
struct ras_err_data *err_data,
struct ta_ras_query_address_input
*addr_in,
@@ -182,11 +225,10 @@ static int umc_v12_0_convert_error_address(struct
amdgpu_device *adev,
{
uint32_t col, col_lower, row, row_lower, row_high, bank;
uint32_t channel_index = 0, umc_inst = 0;
- uint32_t i, loop_bits[UMC_V12_0_RETIRE_LOOP_BITS];
+ uint32_t i, bit_num, retire_unit, *flip_bits;
uint64_t soc_pa, column, err_addr;
struct ta_ras_query_address_output addr_out_tmp;
struct ta_ras_query_address_output *paddr_out;
- enum amdgpu_memory_partition nps = AMDGPU_NPS1_PARTITION_MODE;
int ret = 0;
if (!addr_out)
@@ -211,34 +253,15 @@ static int umc_v12_0_convert_error_address(struct
amdgpu_device *adev,
umc_inst = addr_in->ma.umc_inst;
}
- loop_bits[0] = UMC_V12_0_PA_C2_BIT;
- loop_bits[1] = UMC_V12_0_PA_C3_BIT;
- loop_bits[2] = UMC_V12_0_PA_C4_BIT;
- loop_bits[3] = UMC_V12_0_PA_R13_BIT;
-
- if (adev->gmc.gmc_funcs->query_mem_partition_mode)
- nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
-
- /* other nps modes are taken as nps1 */
- if (nps == AMDGPU_NPS2_PARTITION_MODE) {
- loop_bits[0] = UMC_V12_0_PA_CH5_BIT;
- loop_bits[1] = UMC_V12_0_PA_C2_BIT;
- loop_bits[2] = UMC_V12_0_PA_B1_BIT;
- loop_bits[3] = UMC_V12_0_PA_R12_BIT;
- }
-
- if (nps == AMDGPU_NPS4_PARTITION_MODE) {
- loop_bits[0] = UMC_V12_0_PA_CH4_BIT;
- loop_bits[1] = UMC_V12_0_PA_CH5_BIT;
- loop_bits[2] = UMC_V12_0_PA_B0_BIT;
- loop_bits[3] = UMC_V12_0_PA_R11_BIT;
- }
+ flip_bits = adev->umc.flip_bits.flip_bits_in_pa;
+ bit_num = adev->umc.flip_bits.bit_num;
+ retire_unit = adev->umc.retire_unit;
soc_pa = paddr_out->pa.pa;
channel_index = paddr_out->pa.channel_idx;
/* clear loop bits in soc physical address */
- for (i = 0; i < UMC_V12_0_RETIRE_LOOP_BITS; i++)
- soc_pa &= ~BIT_ULL(loop_bits[i]);
+ for (i = 0; i < bit_num; i++)
+ soc_pa &= ~BIT_ULL(flip_bits[i]);
paddr_out->pa.pa = soc_pa;
/* get column bit 0 and 1 in mca address */
@@ -259,10 +282,10 @@ static int umc_v12_0_convert_error_address(struct
amdgpu_device *adev,
goto out;
/* loop for all possibilities of retired bits */
- for (column = 0; column < UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL; column++)
{
+ for (column = 0; column < retire_unit; column++) {
soc_pa = paddr_out->pa.pa;
- for (i = 0; i < UMC_V12_0_RETIRE_LOOP_BITS; i++)
- soc_pa |= (((column >> i) & 0x1ULL) << loop_bits[i]);
+ for (i = 0; i < bit_num; i++)
+ soc_pa |= (((column >> i) & 0x1ULL) << flip_bits[i]);
col = ((column & 0x7) << 2) | col_lower;
/* add row bit 13 */
@@ -684,5 +707,6 @@ struct amdgpu_umc_ras umc_v12_0_ras = {
.update_ecc_status = umc_v12_0_update_ecc_status,
.convert_ras_err_addr = umc_v12_0_convert_error_address,
.get_die_id_from_pa = umc_v12_0_get_die_id,
+ .get_retire_flip_bits = umc_v12_0_get_retire_flip_bits,
};
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
index 056bbc038312..ccdd6cd430f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v12_0.h
@@ -55,8 +55,6 @@
#define UMC_V12_0_NA_MAP_PA_NUM 8
/* R13 bit shift should be considered, double the number */
#define UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL (UMC_V12_0_NA_MAP_PA_NUM * 2)
-/* C2, C3, C4, R13, four bits in MCA address are looped in retirement */
-#define UMC_V12_0_RETIRE_LOOP_BITS 4
/* column bits in SOC physical address */
#define UMC_V12_0_PA_C2_BIT 15
--
2.34.1
