2020-09-16 19:09:18

by Oded Gabbay

[permalink] [raw]
Subject: [PATCH 2/3] habanalabs/gaudi: fix DMA completions max outstanding to 15

This is a workaround for H/W bug H3-2116, where if there are more than 16
outstanding completions in the DMA transpose engine, there can be a
deadlock in the engine.

Signed-off-by: Oded Gabbay <[email protected]>
---
drivers/misc/habanalabs/gaudi/gaudi.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 2c10e3f92c86..86650be5898f 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -1918,6 +1918,9 @@ static void gaudi_init_dma_core(struct hl_device *hdev, int dma_id)
WREG32(mmDMA0_CORE_RD_MAX_OUTSTAND + dma_offset, 0);
WREG32(mmDMA0_CORE_RD_MAX_SIZE + dma_offset, 0);

+ /* WA for H/W bug H3-2116 */
+ WREG32(mmDMA0_CORE_LBW_MAX_OUTSTAND + dma_offset, 15);
+
/* STOP_ON bit implies no completion to operation in case of RAZWI */
if (hdev->stop_on_err)
dma_err_cfg |= 1 << DMA0_CORE_ERR_CFG_STOP_ON_ERR_SHIFT;
--
2.17.1