From cecde184ca32ae862c5494a7875d03592c893ab9 Mon Sep 17 00:00:00 2001 From: Tomer Tayar Date: Mon, 5 Sep 2022 16:24:21 +0300 Subject: habanalabs/gaudi2: print RAZWI info upon PCIe access error Add the dump of the RAZWI information when a PCIe access is blocked by RR. Signed-off-by: Tomer Tayar Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- drivers/misc/habanalabs/gaudi2/gaudi2.c | 52 ++++++++++++++++++++++++--- drivers/misc/habanalabs/gaudi2/gaudi2_masks.h | 13 +++++++ 2 files changed, 61 insertions(+), 4 deletions(-) (limited to 'drivers/misc') diff --git a/drivers/misc/habanalabs/gaudi2/gaudi2.c b/drivers/misc/habanalabs/gaudi2/gaudi2.c index 5761ca5d50ae..c040e01adafe 100644 --- a/drivers/misc/habanalabs/gaudi2/gaudi2.c +++ b/drivers/misc/habanalabs/gaudi2/gaudi2.c @@ -7963,14 +7963,58 @@ static void gaudi2_handle_dma_core_event(struct hl_device *hdev, u64 intr_cause_ gaudi2_dma_core_interrupts_cause[i]); } +static void gaudi2_print_pcie_mstr_rr_mstr_if_razwi_info(struct hl_device *hdev) +{ + u32 mstr_if_base_addr = mmPCIE_MSTR_RR_MSTR_IF_RR_SHRD_HBW_BASE, razwi_happened_addr; + + razwi_happened_addr = mstr_if_base_addr + RR_SHRD_HBW_AW_RAZWI_HAPPENED; + if (RREG32(razwi_happened_addr)) { + gaudi2_razwi_rr_hbw_shared_printf_info(hdev, mstr_if_base_addr, true, "PCIE", true, + NULL); + WREG32(razwi_happened_addr, 0x1); + } + + razwi_happened_addr = mstr_if_base_addr + RR_SHRD_HBW_AR_RAZWI_HAPPENED; + if (RREG32(razwi_happened_addr)) { + gaudi2_razwi_rr_hbw_shared_printf_info(hdev, mstr_if_base_addr, false, "PCIE", true, + NULL); + WREG32(razwi_happened_addr, 0x1); + } + + razwi_happened_addr = mstr_if_base_addr + RR_SHRD_LBW_AW_RAZWI_HAPPENED; + if (RREG32(razwi_happened_addr)) { + gaudi2_razwi_rr_lbw_shared_printf_info(hdev, mstr_if_base_addr, true, "PCIE", true, + NULL); + WREG32(razwi_happened_addr, 0x1); + } + + razwi_happened_addr = mstr_if_base_addr + RR_SHRD_LBW_AR_RAZWI_HAPPENED; + if (RREG32(razwi_happened_addr)) { + gaudi2_razwi_rr_lbw_shared_printf_info(hdev, mstr_if_base_addr, false, "PCIE", true, + NULL); + WREG32(razwi_happened_addr, 0x1); + } +} + static void gaudi2_print_pcie_addr_dec_info(struct hl_device *hdev, u64 intr_cause_data) { int i; - for (i = 0 ; i < GAUDI2_NUM_OF_PCIE_ADDR_DEC_ERR_CAUSE; i++) - if (intr_cause_data & BIT_ULL(i)) - dev_err_ratelimited(hdev->dev, "PCIE ADDR DEC Error: %s\n", - gaudi2_pcie_addr_dec_error_cause[i]); + for (i = 0 ; i < GAUDI2_NUM_OF_PCIE_ADDR_DEC_ERR_CAUSE ; i++) { + if (!(intr_cause_data & BIT_ULL(i))) + continue; + + dev_err_ratelimited(hdev->dev, "PCIE ADDR DEC Error: %s\n", + gaudi2_pcie_addr_dec_error_cause[i]); + + switch (intr_cause_data & BIT_ULL(i)) { + case PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_LBW_ERR_INTR_MASK: + break; + case PCIE_WRAP_PCIE_IC_SEI_INTR_IND_BAD_ACCESS_INTR_MASK: + gaudi2_print_pcie_mstr_rr_mstr_if_razwi_info(hdev); + break; + } + } } static void gaudi2_handle_pif_fatal(struct hl_device *hdev, u64 intr_cause_data) diff --git a/drivers/misc/habanalabs/gaudi2/gaudi2_masks.h b/drivers/misc/habanalabs/gaudi2/gaudi2_masks.h index 0239d118abc5..e9ac87828221 100644 --- a/drivers/misc/habanalabs/gaudi2/gaudi2_masks.h +++ b/drivers/misc/habanalabs/gaudi2/gaudi2_masks.h @@ -144,4 +144,17 @@ #define DCORE0_SYNC_MNGR_OBJS_SOB_OBJ_SIGN_SHIFT 15 #define DCORE0_SYNC_MNGR_OBJS_SOB_OBJ_SIGN_MASK 0x8000 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_ERR_INTR_SHIFT 0 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_ERR_INTR_MASK 0x1 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_LBW_ERR_INTR_SHIFT 1 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_LBW_ERR_INTR_MASK 0x2 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_BAD_ACCESS_INTR_SHIFT 2 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_BAD_ACCESS_INTR_MASK 0x4 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_ERR_INTR_MASK_SHIFT 3 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_ERR_INTR_MASK_MASK 0x8 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_LBW_ERR_INTR_MASK_SHIFT 4 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_AXI_LBW_ERR_INTR_MASK_MASK 0x10 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_BAD_ACCESS_INTR_MASK_SHIFT 5 +#define PCIE_WRAP_PCIE_IC_SEI_INTR_IND_BAD_ACCESS_INTR_MASK_MASK 0x20 + #endif /* GAUDI2_MASKS_H_ */ -- cgit v1.2.3-58-ga151