[PATCH 2/2] net: Add Keystone NetCP ethernet driver

Santosh Shilimkar santosh.shilimkar at ti.com
Thu Apr 24 14:04:48 PDT 2014


On Thursday 24 April 2014 12:47 PM, David Miller wrote:
> From: Santosh Shilimkar <santosh.shilimkar at ti.com>
> Date: Tue, 22 Apr 2014 17:21:15 -0400
> 
>> +struct netcp_tx_pipe {
>> +	struct netcp_device	*netcp_device;
>> +	void	*dma_queue;
> 
> Indent *dma_queue the same as the other struct members.
>
sure 
>> +	unsigned		dma_queue_id;
> 
> Use explicit "unsigned int".
> 
>> +	unsigned		dma_chan_id;
> 
> Likewise.
>
ok
 
>> +struct netcp_addr {
>> +	struct netcp_intf	*netcp;
>> +	unsigned char		addr[MAX_ADDR_LEN];
> 
> If this is just an ethernet driver, ETH_ALEN is more appropriate here.
> 
Yep. Will use ETH_ALEN

>> +	unsigned		tx_compl_qid;
> 
> Explicit "unsigned int" please.  I'm not going to point out all of the other
> instances, audit your entire submission for this problem please.
>
Thats true... Will fix all instances of those. 

 
>> +static inline u32 *netcp_push_psdata(struct netcp_packet *p_info,
>> +				     unsigned bytes)
>> +{
>> +	u32		*buf;
>> +	unsigned	 words;
> 
> Do not use tabs between the type and the variable name in local variable
> declarations.
> 
ok

> Please audit for and fix this in your entire submission.
>
Will Do.
 
>> +static inline u32 hwval_to_host(bool big_endian, u32 hwval)
>> +{
>> +	if (big_endian)
>> +		return be32_to_cpu(hwval);
>> +	else
>> +		return le32_to_cpu(hwval);
>> +}
> 
> You're much better off having a set of methods, one for big endian and
> one for little endian, that just straight line codes the appropriate endian
> accesses.
>
good idea.
 
> These conditionals peppered all over the place are just ugly.
> 
Agree. Will try to fit into some logical functions.

Thanks for the review David !!

regards,
Santosh



More information about the linux-arm-kernel mailing list