habanalabs: fix completion queue handling when host is BE

This patch fix the CQ irq handler to work in hosts with BE architecture.
It adds the correct endian-swapping macros around the relevant memory
accesses.
Signed-off-by: Ben Segal <bpsegal20@gmail.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

drivers/misc/habanalabs/irq.c

@@ -80,8 +80,7 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg)
 	struct hl_cs_job *job;
 	bool shadow_index_valid;
 	u16 shadow_index;
-	u32 *cq_entry;
-	u32 *cq_base;
+	struct hl_cq_entry *cq_entry, *cq_base;
 
 	if (hdev->disabled) {
 		dev_dbg(hdev->dev,
@@ -90,29 +89,29 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg)
 		return IRQ_HANDLED;
 	}
 
-	cq_base = (u32 *) (uintptr_t) cq->kernel_address;
+	cq_base = (struct hl_cq_entry *) (uintptr_t) cq->kernel_address;
 
 	while (1) {
-		bool entry_ready = ((cq_base[cq->ci] & CQ_ENTRY_READY_MASK)
+		bool entry_ready = ((le32_to_cpu(cq_base[cq->ci].data) &
+					CQ_ENTRY_READY_MASK)
 						>> CQ_ENTRY_READY_SHIFT);
 
 		if (!entry_ready)
 			break;
 
-		cq_entry = (u32 *) &cq_base[cq->ci];
+		cq_entry = (struct hl_cq_entry *) &cq_base[cq->ci];
 
-		/*
-		 * Make sure we read CQ entry contents after we've
+		/* Make sure we read CQ entry contents after we've
 		 * checked the ownership bit.
 		 */
 		dma_rmb();
 
-		shadow_index_valid =
-			((*cq_entry & CQ_ENTRY_SHADOW_INDEX_VALID_MASK)
+		shadow_index_valid = ((le32_to_cpu(cq_entry->data) &
+					CQ_ENTRY_SHADOW_INDEX_VALID_MASK)
 					>> CQ_ENTRY_SHADOW_INDEX_VALID_SHIFT);
 
-		shadow_index = (u16)
-			((*cq_entry & CQ_ENTRY_SHADOW_INDEX_MASK)
+		shadow_index = (u16) ((le32_to_cpu(cq_entry->data) &
+					CQ_ENTRY_SHADOW_INDEX_MASK)
 					>> CQ_ENTRY_SHADOW_INDEX_SHIFT);
 
 		queue = &hdev->kernel_queues[cq->hw_queue_id];
@@ -122,8 +121,7 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg)
 			queue_work(hdev->cq_wq, &job->finish_work);
 		}
 
-		/*
-		 * Update ci of the context's queue. There is no
+		/* Update ci of the context's queue. There is no
 		 * need to protect it with spinlock because this update is
 		 * done only inside IRQ and there is a different IRQ per
 		 * queue
@@ -131,7 +129,8 @@ irqreturn_t hl_irq_handler_cq(int irq, void *arg)
 		queue->ci = hl_queue_inc_ptr(queue->ci);
 
 		/* Clear CQ entry ready bit */
-		cq_base[cq->ci] &= ~CQ_ENTRY_READY_MASK;
+		cq_entry->data = cpu_to_le32(le32_to_cpu(cq_entry->data) &
+						~CQ_ENTRY_READY_MASK);
 
 		cq->ci = hl_cq_inc_ptr(cq->ci);