[PATCH 01/16] habanalabs/gaudi: scrub HBM to a specific value

From: Oded Gabbay
Date: Wed Aug 18 2021 - 09:39:46 EST


From: Ofir Bitton <obitton@xxxxxxxxx>

In order to enhance debuggability, we will scrub the whole HBM to
a specific value, in case HBM scrubbing is enabled. Scrubbing will be
performed after reset and after user closes the FD.

Signed-off-by: Ofir Bitton <obitton@xxxxxxxxx>
Reviewed-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
Signed-off-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
---
drivers/misc/habanalabs/gaudi/gaudi.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 4ea45bdb9a71..6852bb8e55db 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -4194,6 +4194,11 @@ static int gaudi_hw_init(struct hl_device *hdev)
goto disable_msi;
}

+ /* Scrub both SRAM and DRAM */
+ rc = hdev->asic_funcs->scrub_device_mem(hdev, 0, 0);
+ if (rc)
+ return rc;
+
/* Perform read from the device to flush all configuration */
RREG32(mmHW_STATE);

@@ -4758,8 +4763,8 @@ static int gaudi_hbm_scrubbing(struct hl_device *hdev)
"Doing HBM scrubbing for 0x%09llx - 0x%09llx\n",
cur_addr, cur_addr + chunk_size);

- WREG32(mmDMA0_CORE_SRC_BASE_LO + dma_offset, 0);
- WREG32(mmDMA0_CORE_SRC_BASE_HI + dma_offset, 0);
+ WREG32(mmDMA0_CORE_SRC_BASE_LO + dma_offset, 0xdeadbeaf);
+ WREG32(mmDMA0_CORE_SRC_BASE_HI + dma_offset, 0xdeadbeaf);
WREG32(mmDMA0_CORE_DST_BASE_LO + dma_offset,
lower_32_bits(cur_addr));
WREG32(mmDMA0_CORE_DST_BASE_HI + dma_offset,
--
2.17.1