scsi: lpfc: Add new RCQE status for handling DMA failures

A new RCQE status value indicating DMA failure when transferring
asynchronously received data to an RQE is introduced.  Such errors are
unexpected and handlers are updated to log KERN_ERR and dump lpfc's debug
trace buffer to kmsg.
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Link: https://lore.kernel.org/r/20230417191558.83100-6-justintee8345@gmail.comSigned-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/lpfc/lpfc_hw4.h

@@ -536,9 +536,9 @@ struct sli4_wcqe_xri_aborted {
 /* completion queue entry structure for rqe completion */
 struct lpfc_rcqe {
 	uint32_t word0;
-#define lpfc_rcqe_bindex_SHIFT		16
-#define lpfc_rcqe_bindex_MASK		0x0000FFF
-#define lpfc_rcqe_bindex_WORD		word0
+#define lpfc_rcqe_iv_SHIFT		31
+#define lpfc_rcqe_iv_MASK		0x00000001
+#define lpfc_rcqe_iv_WORD		word0
 #define lpfc_rcqe_status_SHIFT		8
 #define lpfc_rcqe_status_MASK		0x000000FF
 #define lpfc_rcqe_status_WORD		word0
@@ -546,6 +546,7 @@ struct lpfc_rcqe {
 #define FC_STATUS_RQ_BUF_LEN_EXCEEDED 	0x11 /* payload truncated */
 #define FC_STATUS_INSUFF_BUF_NEED_BUF 	0x12 /* Insufficient buffers */
 #define FC_STATUS_INSUFF_BUF_FRM_DISC 	0x13 /* Frame Discard */
+#define FC_STATUS_RQ_DMA_FAILURE	0x14 /* DMA failure */
 	uint32_t word1;
 #define lpfc_rcqe_fcf_id_v1_SHIFT	0
 #define lpfc_rcqe_fcf_id_v1_MASK	0x0000003F

drivers/scsi/lpfc/lpfc_sli.c

@@ -14682,6 +14682,38 @@ lpfc_sli4_sp_handle_rcqe(struct lpfc_hba *phba, struct lpfc_rcqe *rcqe)
 		spin_unlock_irqrestore(&phba->hbalock, iflags);
 		workposted = true;
 		break;
+	case FC_STATUS_RQ_DMA_FAILURE:
+		lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
+				"2564 RQE DMA Error x%x, x%08x x%08x x%08x "
+				"x%08x\n",
+				status, rcqe->word0, rcqe->word1,
+				rcqe->word2, rcqe->word3);
+
+		/* If IV set, no further recovery */
+		if (bf_get(lpfc_rcqe_iv, rcqe))
+			break;
+
+		/* recycle consumed resource */
+		spin_lock_irqsave(&phba->hbalock, iflags);
+		lpfc_sli4_rq_release(hrq, drq);
+		dma_buf = lpfc_sli_hbqbuf_get(&phba->hbqs[0].hbq_buffer_list);
+		if (!dma_buf) {
+			hrq->RQ_no_buf_found++;
+			spin_unlock_irqrestore(&phba->hbalock, iflags);
+			break;
+		}
+		hrq->RQ_rcv_buf++;
+		hrq->RQ_buf_posted--;
+		spin_unlock_irqrestore(&phba->hbalock, iflags);
+		lpfc_in_buf_free(phba, &dma_buf->dbuf);
+		break;
+	default:
+		lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
+				"2565 Unexpected RQE Status x%x, w0-3 x%08x "
+				"x%08x x%08x x%08x\n",
+				status, rcqe->word0, rcqe->word1,
+				rcqe->word2, rcqe->word3);
+		break;
 	}
 out:
 	return workposted;
@@ -15203,6 +15235,38 @@ lpfc_sli4_nvmet_handle_rcqe(struct lpfc_hba *phba, struct lpfc_queue *cq,
 		hrq->RQ_no_posted_buf++;
 		/* Post more buffers if possible */
 		break;
+	case FC_STATUS_RQ_DMA_FAILURE:
+		lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
+				"2575 RQE DMA Error x%x, x%08x x%08x x%08x "
+				"x%08x\n",
+				status, rcqe->word0, rcqe->word1,
+				rcqe->word2, rcqe->word3);
+
+		/* If IV set, no further recovery */
+		if (bf_get(lpfc_rcqe_iv, rcqe))
+			break;
+
+		/* recycle consumed resource */
+		spin_lock_irqsave(&phba->hbalock, iflags);
+		lpfc_sli4_rq_release(hrq, drq);
+		dma_buf = lpfc_sli_rqbuf_get(phba, hrq);
+		if (!dma_buf) {
+			hrq->RQ_no_buf_found++;
+			spin_unlock_irqrestore(&phba->hbalock, iflags);
+			break;
+		}
+		hrq->RQ_rcv_buf++;
+		hrq->RQ_buf_posted--;
+		spin_unlock_irqrestore(&phba->hbalock, iflags);
+		lpfc_rq_buf_free(phba, &dma_buf->hbuf);
+		break;
+	default:
+		lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
+				"2576 Unexpected RQE Status x%x, w0-3 x%08x "
+				"x%08x x%08x x%08x\n",
+				status, rcqe->word0, rcqe->word1,
+				rcqe->word2, rcqe->word3);
+		break;
 	}
 out:
 	return workposted;