On Sun, Oct 17, 2010 at 07:16, Mike Frysinger wrote:
> On Thu, Sep 16, 2010 at 23:46,  <[email protected]> wrote:
>> Revision 9145 Author lliubbo Date 2010-09-16 23:46:22 -0400 (Thu, 16 Sep
>> 2010)
>>
>> Log Message
>>
>> [#5553] spi_bfin5xx: fix bits_per_word check
>>
>> If a spi_transfer has a bits_per_word isn't (8,16),it will be treated as 16
>> by
>> mistake.
>> This commit fix it with more check and reject that transfer with a err msg.
>>
>> Modified: trunk/drivers/spi/spi_bfin5xx.c (9144 => 9145)
>>
>>               drv_data->len = transfer->len;
>>               cr_width = 0;
>>               drv_data->ops = &bfin_transfer_ops_u8;
>> -     } else {
>> +     } else if (bits_per_word == 16) {
>>               drv_data->n_bytes = 2;
>>               drv_data->len = (transfer->len) >> 1;
>>               cr_width = BIT_CTL_WORDSIZE;
>>               drv_data->ops = &bfin_transfer_ops_u16;
>> +     } else {
>> +             dev_err(&drv_data->pdev->dev, "transfer: unsupported 
>> bits_per_word\n");
>> +             return;
>>       }
>
> reviewing this again, i noticed another issue ... this would
> incorrectly reject GPIO CS transfers which are multiples of 8 bits.
>
> check out the setup func to see how the bits_per_word field there is 
> validated.

well, scratch that.  i think we need to do more work before GPIO CS
can support more than 16bits.  i'll open a tracker item and make sure
the setup func rejects non-8/16bit transfers.
-mike
_______________________________________________
Linux-kernel-commits mailing list
[email protected]
https://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits

Reply via email to