diff options
| author | Ofir Bitton <obitton@habana.ai> | 2022-04-25 19:58:37 +0300 |
|---|---|---|
| committer | Greg Kroah-Hartman <gregkh@linuxfoundation.org> | 2022-05-22 22:01:19 +0300 |
| commit | 2db04a68265519f6d3783f4cb99175fb6923aa2d (patch) | |
| tree | a9fb46e5be658acadd84d5ebe962f9e80ec39990 | |
| parent | ff086c186bf2b2237410ce9145c16d2141a881ff (diff) | |
| download | linux-2db04a68265519f6d3783f4cb99175fb6923aa2d.tar.xz | |
habanalabs/gaudi: set arbitration timeout to a high value
In certain workloads, arbitration timeout might expire although
no actual issue present. Hence, we set timeout to a very high value.
Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
| -rw-r--r-- | drivers/misc/habanalabs/gaudi/gaudi.c | 27 |
1 files changed, 11 insertions, 16 deletions
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c index 7828337eedce..e74ea4c2d008 100644 --- a/drivers/misc/habanalabs/gaudi/gaudi.c +++ b/drivers/misc/habanalabs/gaudi/gaudi.c @@ -95,7 +95,7 @@ #define GAUDI_NUM_OF_QM_ARB_ERR_CAUSE 3 -#define GAUDI_ARB_WDT_TIMEOUT 0x1000000 +#define GAUDI_ARB_WDT_TIMEOUT 0xEE6b27FF /* 8 seconds */ #define GAUDI_CLK_GATE_DEBUGFS_MASK (\ BIT(GAUDI_ENGINE_ID_MME_0) |\ @@ -2811,9 +2811,8 @@ static void gaudi_init_pci_dma_qman(struct hl_device *hdev, int dma_id, WREG32(mmDMA0_QM_ARB_ERR_MSG_EN + dma_qm_offset, QM_ARB_ERR_MSG_EN_MASK); - /* Increase ARB WDT to support streams architecture */ - WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, - GAUDI_ARB_WDT_TIMEOUT); + /* Set timeout to maximum */ + WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, GAUDI_ARB_WDT_TIMEOUT); WREG32(mmDMA0_QM_GLBL_PROT + dma_qm_offset, QMAN_EXTERNAL_MAKE_TRUSTED); @@ -2990,9 +2989,8 @@ static void gaudi_init_hbm_dma_qman(struct hl_device *hdev, int dma_id, WREG32(mmDMA0_QM_ARB_ERR_MSG_EN + dma_qm_offset, QM_ARB_ERR_MSG_EN_MASK); - /* Increase ARB WDT to support streams architecture */ - WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, - GAUDI_ARB_WDT_TIMEOUT); + /* Set timeout to maximum */ + WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, GAUDI_ARB_WDT_TIMEOUT); WREG32(mmDMA0_QM_GLBL_CFG1 + dma_qm_offset, 0); WREG32(mmDMA0_QM_GLBL_PROT + dma_qm_offset, @@ -3127,9 +3125,8 @@ static void gaudi_init_mme_qman(struct hl_device *hdev, u32 mme_offset, WREG32(mmMME0_QM_ARB_ERR_MSG_EN + mme_offset, QM_ARB_ERR_MSG_EN_MASK); - /* Increase ARB WDT to support streams architecture */ - WREG32(mmMME0_QM_ARB_SLV_CHOISE_WDT + mme_offset, - GAUDI_ARB_WDT_TIMEOUT); + /* Set timeout to maximum */ + WREG32(mmMME0_QM_ARB_SLV_CHOISE_WDT + mme_offset, GAUDI_ARB_WDT_TIMEOUT); WREG32(mmMME0_QM_GLBL_CFG1 + mme_offset, 0); WREG32(mmMME0_QM_GLBL_PROT + mme_offset, @@ -3261,9 +3258,8 @@ static void gaudi_init_tpc_qman(struct hl_device *hdev, u32 tpc_offset, WREG32(mmTPC0_QM_ARB_ERR_MSG_EN + tpc_offset, QM_ARB_ERR_MSG_EN_MASK); - /* Increase ARB WDT to support streams architecture */ - WREG32(mmTPC0_QM_ARB_SLV_CHOISE_WDT + tpc_offset, - GAUDI_ARB_WDT_TIMEOUT); + /* Set timeout to maximum */ + WREG32(mmTPC0_QM_ARB_SLV_CHOISE_WDT + tpc_offset, GAUDI_ARB_WDT_TIMEOUT); WREG32(mmTPC0_QM_GLBL_CFG1 + tpc_offset, 0); WREG32(mmTPC0_QM_GLBL_PROT + tpc_offset, @@ -3412,9 +3408,8 @@ static void gaudi_init_nic_qman(struct hl_device *hdev, u32 nic_offset, WREG32(mmNIC0_QM0_ARB_ERR_MSG_EN + nic_offset, QM_ARB_ERR_MSG_EN_MASK); - /* Increase ARB WDT to support streams architecture */ - WREG32(mmNIC0_QM0_ARB_SLV_CHOISE_WDT + nic_offset, - GAUDI_ARB_WDT_TIMEOUT); + /* Set timeout to maximum */ + WREG32(mmNIC0_QM0_ARB_SLV_CHOISE_WDT + nic_offset, GAUDI_ARB_WDT_TIMEOUT); WREG32(mmNIC0_QM0_GLBL_CFG1 + nic_offset, 0); WREG32(mmNIC0_QM0_GLBL_PROT + nic_offset, |