Re: [PATCH net] staging: Remove set but not used variable ‘status’

[maowenan <maowenan@huawei.com>]

On 2019/5/25 20:59, Sven Van Asbroeck wrote:
> On Sat, May 25, 2019 at 12:20 AM Mao Wenan <maowenan@huawei.com> wrote:
>>
>> The variable 'status' is not used any more, remve it.
> 
>>      /* do the transfers for this message */
>>      list_for_each_entry(transfer, &m->transfers, transfer_list) {
>>          if (transfer->tx_buf == NULL && transfer->rx_buf == NULL && transfer->len) {
>> -            status = -EINVAL;
>>              break;
>>          }
> 
> This looks like an error condition that's not reported to the spi core.
> 
> Instead of removing the status variable (which also removes the error value!),
> maybe this should be reported to the spi core instead ?
> 
> Other spi drivers appear to do the following on the error path:
> m->status = status;
> return status;

I have reviewed the code again, and it is good idea to store m->status in error path, like below?

@@ -374,7 +374,7 @@ kp_spi_transfer_one_message(struct spi_master *master, struct spi_message *m)
     list_for_each_entry(transfer, &m->transfers, transfer_list) {
         if (transfer->tx_buf == NULL && transfer->rx_buf == NULL && transfer->len) {
             status = -EINVAL;
-            break;
+            goto error;
         }

         /* transfer */
@@ -412,7 +412,7 @@ kp_spi_transfer_one_message(struct spi_master *master, struct spi_message *m)

             if (count != transfer->len) {
                 status = -EIO;
-                break;
+                goto error;
             }
         }


...

     /* done work */
     spi_finalize_current_message(master);
     return 0;
+
+ error:
+    m->status = status;
+    return status;

 }


> 
>>
>> @@ -370,7 +368,6 @@ kp_spi_transfer_one_message(struct spi_master *master, struct spi_message *m)
>>
>>              if (count != transfer->len) {
>> -                status = -EIO;
>>                  break;
> 
> Same issue here.
> 
>