[PATCH 10/12] habanalabs: skip read fw errors if dynamic descriptor invalid

From: Oded Gabbay
Date: Sun Nov 28 2021 - 14:37:13 EST


From: Ohad Sharabi <osharabi@xxxxxxxxx>

Reporting FW errors involves reading of the error registers.

In case we have a corrupted FW descriptor we cannot do that since the
dynamic scratchpad is potentially corrupted as well and may cause kernel
crush when attempting access to a corrupted register offset.

Signed-off-by: Ohad Sharabi <osharabi@xxxxxxxxx>
Reviewed-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
Signed-off-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
---
drivers/misc/habanalabs/common/firmware_if.c | 17 +++++++++++++++--
drivers/misc/habanalabs/common/habanalabs.h | 2 ++
2 files changed, 17 insertions(+), 2 deletions(-)

diff --git a/drivers/misc/habanalabs/common/firmware_if.c b/drivers/misc/habanalabs/common/firmware_if.c
index bafadcc65497..760ca139d5cf 100644
--- a/drivers/misc/habanalabs/common/firmware_if.c
+++ b/drivers/misc/habanalabs/common/firmware_if.c
@@ -1772,6 +1772,9 @@ static int hl_fw_dynamic_validate_descriptor(struct hl_device *hdev,
return rc;
}

+ /* here we can mark the descriptor as valid as the content has been validated */
+ fw_loader->dynamic_loader.fw_desc_valid = true;
+
return 0;
}

@@ -1828,7 +1831,13 @@ static int hl_fw_dynamic_read_and_validate_descriptor(struct hl_device *hdev,
return rc;
}

- /* extract address copy the descriptor from */
+ /*
+ * extract address to copy the descriptor from
+ * in addition, as the descriptor value is going to be over-ridden by new data- we mark it
+ * as invalid.
+ * it will be marked again as valid once validated
+ */
+ fw_loader->dynamic_loader.fw_desc_valid = false;
src = hdev->pcie_bar[region->bar_id] + region->offset_in_bar +
response->ram_offset;
memcpy_fromio(fw_desc, src, sizeof(struct lkd_fw_comms_desc));
@@ -2317,6 +2326,9 @@ static int hl_fw_dynamic_init_cpu(struct hl_device *hdev,
dev_info(hdev->dev,
"Loading firmware to device, may take some time...\n");

+ /* initialize FW descriptor as invalid */
+ fw_loader->dynamic_loader.fw_desc_valid = false;
+
/*
* In this stage, "cpu_dyn_regs" contains only LKD's hard coded values!
* It will be updated from FW after hl_fw_dynamic_request_descriptor().
@@ -2412,7 +2424,8 @@ static int hl_fw_dynamic_init_cpu(struct hl_device *hdev,
return 0;

protocol_err:
- fw_read_errors(hdev, le32_to_cpu(dyn_regs->cpu_boot_err0),
+ if (fw_loader->dynamic_loader.fw_desc_valid)
+ fw_read_errors(hdev, le32_to_cpu(dyn_regs->cpu_boot_err0),
le32_to_cpu(dyn_regs->cpu_boot_err1),
le32_to_cpu(dyn_regs->cpu_boot_dev_sts0),
le32_to_cpu(dyn_regs->cpu_boot_dev_sts1));
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index c2129c9fe9e4..77ac4bb98137 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -1034,6 +1034,7 @@ struct fw_response {
* @image_region: region to copy the FW image to
* @fw_image_size: size of FW image to load
* @wait_for_bl_timeout: timeout for waiting for boot loader to respond
+ * @fw_desc_valid: true if FW descriptor has been validated and hence the data can be used
*/
struct dynamic_fw_load_mgr {
struct fw_response response;
@@ -1041,6 +1042,7 @@ struct dynamic_fw_load_mgr {
struct pci_mem_region *image_region;
size_t fw_image_size;
u32 wait_for_bl_timeout;
+ bool fw_desc_valid;
};

/**
--
2.25.1