[PATCH] [SCSI] qla2xxx: improve error returns in qla2x00_process_vendor_specific

[PATCH] [SCSI] qla2xxx: improve error returns in qla2x00_process_vendor_specific

[Stephen Rothwell]

These changes mean that the request sg will be unmapped if the response
sg map fails, both the dma_allocs are freed and the dma_allocs only get
freed if necessary.  There are still two places that return from this
routine without doing either the unmaps or the dma_frees.

This gets rid of these warnings:

drivers/scsi/qla2xxx/qla_attr.c: In function 'qla2x00_process_vendor_specific':
drivers/scsi/qla2xxx/qla_attr.c:2150: warning: 'req_data' may be used uninitialized in this function
drivers/scsi/qla2xxx/qla_attr.c:2152: warning: 'req_data_len' may be used uninitialized in this function

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
---
 drivers/scsi/qla2xxx/qla_attr.c |   20 ++++++++++++--------
 1 files changed, 12 insertions(+), 8 deletions(-)

I have only build tested these changes as I do not have the hardware
involved.  Also, I am not sure if some of the infrastructure does the
dma_free_coherent() for the rsp_data.

diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c
index 5a19aea..d8c5bf3 100644
--- a/drivers/scsi/qla2xxx/qla_attr.c
+++ b/drivers/scsi/qla2xxx/qla_attr.c
@@ -2173,7 +2173,7 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
 	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
         if (!elreq.rsp_sg_cnt) {
 		rval = -ENOMEM;
-                goto done;
+                goto done_unmap_sg_req;
 	}
 
 	if ((elreq.req_sg_cnt !=  bsg_job->request_payload.sg_cnt) ||
@@ -2224,7 +2224,7 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
 				DEBUG2(qla_printk(KERN_INFO, ha, "scsi(%ld)"
 				"loopback option:0x%x not supported\n", vha->host_no, elreq.options));
 				rval = -EINVAL;
-				goto done_unmap_sg;
+				goto done_dma_free;
 			}
 
 			DEBUG2(qla_printk(KERN_INFO, ha,
@@ -2304,17 +2304,21 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
 	}
 	bsg_job->job_done(bsg_job);
 
+done_dma_free:
+	dma_free_coherent(&ha->pdev->dev, rsp_data_len,
+		rsp_data, rsp_data_dma);
+	dma_free_coherent(&ha->pdev->dev, req_data_len,
+		req_data, req_data_dma);
+
 done_unmap_sg:
+	dma_unmap_sg(&ha->pdev->dev,
+	    bsg_job->reply_payload.sg_list,
+	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
 
-	if(req_data)
-		dma_free_coherent(&ha->pdev->dev, req_data_len,
-			req_data, req_data_dma);
+done_unmap_sg_req:
 	dma_unmap_sg(&ha->pdev->dev,
 	    bsg_job->request_payload.sg_list,
 	    bsg_job->request_payload.sg_cnt, DMA_TO_DEVICE);
-	dma_unmap_sg(&ha->pdev->dev,
-	    bsg_job->reply_payload.sg_list,
-	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
 
 done:
         return rval;
-- 
1.6.6.2

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

Re: [PATCH] [SCSI] qla2xxx: improve error returns in qla2x00_process_vendor_specific

[Giridhar Malavali]

Thanks for pointing this. 

The patch does not apply cleanly with the latest BSG changes. I will make the necessary changes and re-send the patch.

-- Giri 


On Feb 17, 2010, at 8:11 PM, Stephen Rothwell wrote:

> These changes mean that the request sg will be unmapped if the response
> sg map fails, both the dma_allocs are freed and the dma_allocs only get
> freed if necessary.  There are still two places that return from this
> routine without doing either the unmaps or the dma_frees.
> 
> This gets rid of these warnings:
> 
> drivers/scsi/qla2xxx/qla_attr.c: In function 'qla2x00_process_vendor_specific':
> drivers/scsi/qla2xxx/qla_attr.c:2150: warning: 'req_data' may be used uninitialized in this function
> drivers/scsi/qla2xxx/qla_attr.c:2152: warning: 'req_data_len' may be used uninitialized in this function
> 
> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
> ---
> drivers/scsi/qla2xxx/qla_attr.c |   20 ++++++++++++--------
> 1 files changed, 12 insertions(+), 8 deletions(-)
> 
> I have only build tested these changes as I do not have the hardware
> involved.  Also, I am not sure if some of the infrastructure does the
> dma_free_coherent() for the rsp_data.
> 
> diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c
> index 5a19aea..d8c5bf3 100644
> --- a/drivers/scsi/qla2xxx/qla_attr.c
> +++ b/drivers/scsi/qla2xxx/qla_attr.c
> @@ -2173,7 +2173,7 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
> 	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
>         if (!elreq.rsp_sg_cnt) {
> 		rval = -ENOMEM;
> -                goto done;
> +                goto done_unmap_sg_req;
> 	}
> 
> 	if ((elreq.req_sg_cnt !=  bsg_job->request_payload.sg_cnt) ||
> @@ -2224,7 +2224,7 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
> 				DEBUG2(qla_printk(KERN_INFO, ha, "scsi(%ld)"
> 				"loopback option:0x%x not supported\n", vha->host_no, elreq.options));
> 				rval = -EINVAL;
> -				goto done_unmap_sg;
> +				goto done_dma_free;
> 			}
> 
> 			DEBUG2(qla_printk(KERN_INFO, ha,
> @@ -2304,17 +2304,21 @@ qla2x00_process_vendor_specific(struct fc_bsg_job *bsg_job)
> 	}
> 	bsg_job->job_done(bsg_job);
> 
> +done_dma_free:
> +	dma_free_coherent(&ha->pdev->dev, rsp_data_len,
> +		rsp_data, rsp_data_dma);
> +	dma_free_coherent(&ha->pdev->dev, req_data_len,
> +		req_data, req_data_dma);
> +
> done_unmap_sg:
> +	dma_unmap_sg(&ha->pdev->dev,
> +	    bsg_job->reply_payload.sg_list,
> +	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
> 
> -	if(req_data)
> -		dma_free_coherent(&ha->pdev->dev, req_data_len,
> -			req_data, req_data_dma);
> +done_unmap_sg_req:
> 	dma_unmap_sg(&ha->pdev->dev,
> 	    bsg_job->request_payload.sg_list,
> 	    bsg_job->request_payload.sg_cnt, DMA_TO_DEVICE);
> -	dma_unmap_sg(&ha->pdev->dev,
> -	    bsg_job->reply_payload.sg_list,
> -	    bsg_job->reply_payload.sg_cnt, DMA_FROM_DEVICE);
> 
> done:
>         return rval;
> -- 
> 1.6.6.2
> 
> -- 
> Cheers,
> Stephen Rothwell                    sfr@canb.auug.org.au
> http://www.canb.auug.org.au/~sfr/