From: Christoph Hellwig <hch@lst.de>
Date: Thu, 11 Oct 2018 14:17:14 +0000 (+0200)
Subject: scsi: dc395x: fix dma API usage in srb_done
X-Git-Tag: rel_imx_5.10.35_2.0.0-somdevices.0~5973^2~53
X-Git-Url: https://git.somdevices.com/?a=commitdiff_plain;h=3a5bd7021184dec2946f2a4d7a8943f8a5713e52;p=linux.git

scsi: dc395x: fix dma API usage in srb_done

We can't just transfer ownership to the CPU and then unmap, as this will
break with swiotlb.

Instead unmap the command and sense buffer a little earlier in the I/O
completion handler and get rid of the pci_dma_sync_sg_for_cpu call
entirely.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
---

diff --git a/drivers/scsi/dc395x.c b/drivers/scsi/dc395x.c
index 8705730b2e5a..8d7a26a33951 100644
--- a/drivers/scsi/dc395x.c
+++ b/drivers/scsi/dc395x.c
@@ -3349,14 +3349,12 @@ static void srb_done(struct AdapterCtlBlk *acb, struct DeviceCtlBlk *dcb,
 		}
 	}
 
-	if (dir != PCI_DMA_NONE && scsi_sg_count(cmd))
-		pci_dma_sync_sg_for_cpu(acb->dev, scsi_sglist(cmd),
-					scsi_sg_count(cmd), dir);
-
 	ckc_only = 0;
 /* Check Error Conditions */
       ckc_e:
 
+	pci_unmap_srb(acb, srb);
+
 	if (cmd->cmnd[0] == INQUIRY) {
 		unsigned char *base = NULL;
 		struct ScsiInqData *ptr;
@@ -3408,7 +3406,6 @@ static void srb_done(struct AdapterCtlBlk *acb, struct DeviceCtlBlk *dcb,
 	} else {
 		dprintkl(KERN_ERR, "srb_done: ERROR! Completed cmd with tmp_srb\n");
 	}
-	pci_unmap_srb(acb, srb);
 
 	cmd->scsi_done(cmd);
 	waiting_process_next(acb);