habanalabs: add an option to delay a device reset

Several H/W events can be sent adjacently, even due to a single error.
If a hard-reset is triggered as part of handling one of these events,
the following events won't be handled.
The debug info from these missed events is important, sometimes even
more important than the one that was handled.

To allow handling these close events, add an option to delay a device
reset and use it when resetting due to H/W events.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index d52381d..dc9341a 100644 (file)
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -13,6 +13,8 @@
 #include <linux/pci.h>
 #include <linux/hwmon.h>
 
+#define HL_RESET_DELAY_USEC            10000   /* 10ms */
+
 enum hl_device_status hl_device_status(struct hl_device *hdev)
 {
        enum hl_device_status status;
@@ -980,7 +982,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
 {
        bool hard_reset, from_hard_reset_thread, fw_reset, hard_instead_soft = false,
                        reset_upon_device_release = false, schedule_hard_reset = false,
-                       skip_wq_flush = false;
+                       skip_wq_flush, delay_reset;
        u64 idle_mask[HL_BUSY_ENGINES_MASK_EXT_SIZE] = {0};
        struct hl_ctx *ctx;
        int i, rc;
@@ -994,6 +996,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
        from_hard_reset_thread = !!(flags & HL_DRV_RESET_FROM_RESET_THR);
        fw_reset = !!(flags & HL_DRV_RESET_BYPASS_REQ_TO_FW);
        skip_wq_flush = !!(flags & HL_DRV_RESET_DEV_RELEASE);
+       delay_reset = !!(flags & HL_DRV_RESET_DELAY);
 
        if (!hard_reset && !hdev->asic_prop.supports_soft_reset) {
                hard_instead_soft = true;
@@ -1043,6 +1046,9 @@ do_reset:
                hdev->reset_info.in_reset = 1;
                spin_unlock(&hdev->reset_info.lock);
 
+               if (delay_reset)
+                       usleep_range(HL_RESET_DELAY_USEC, HL_RESET_DELAY_USEC << 1);
+
                handle_reset_trigger(hdev, flags);
 
                /* This still allows the completion of some KDMA ops */

diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index cef4717..1edaf6a 100644 (file)
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -142,6 +142,9 @@ enum hl_mmu_page_table_location {
  *
  * - HL_DRV_RESET_FW_FATAL_ERR
  *       Set if reset is due to a fatal error from FW
+ *
+ * - HL_DRV_RESET_DELAY
+ *       Set if a delay should be added before the reset
  */
 
 #define HL_DRV_RESET_HARD              (1 << 0)
@@ -151,6 +154,7 @@ enum hl_mmu_page_table_location {
 #define HL_DRV_RESET_DEV_RELEASE       (1 << 4)
 #define HL_DRV_RESET_BYPASS_REQ_TO_FW  (1 << 5)
 #define HL_DRV_RESET_FW_FATAL_ERR      (1 << 6)
+#define HL_DRV_RESET_DELAY             (1 << 7)
 
 #define HL_MAX_SOBS_PER_MONITOR        8
 

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 0eed0ef..21c2b67 100644 (file)
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -8199,7 +8199,7 @@ reset_device:
                                        | HL_DRV_RESET_BYPASS_REQ_TO_FW
                                        | fw_fatal_err_flag);
        else if (hdev->hard_reset_on_fw_events)
-               hl_device_reset(hdev, HL_DRV_RESET_HARD | fw_fatal_err_flag);
+               hl_device_reset(hdev, HL_DRV_RESET_HARD | HL_DRV_RESET_DELAY | fw_fatal_err_flag);
        else
                hl_fw_unmask_irq(hdev, event_type);
 }